When can I stop?

Processing Rules in SiteCatalyst version 15

2012-02-06T23:22:00.000Z

For those of you who follow my blog and were expecting a post detailing which if of my six things that I wanted for SiteCatalyst version 15 actually got implemented in version 15 have been disappointed for the last 14 months. They're going to continue being a little disappointed as well, because I'm only going to discuss two of them in this post. Sorry. That one is number one on the list and number two, which have both been done through a similar new feature called processing rules.

What are Processing Rules?

Processing rules allow you create rules on your data in between it being collected and presented in reports. More specifically for those who use them, they process data before it gets to the VISTA rules.

This is very important as previously if you wanted to change the way your set up worked you'd have to change the code on the page before you could do anything. This meant you might have to wait for quite a long time before you could start collecting data. Now you can set the rule up and it will start processing the next day.

How Do I Get Processing Rules?

It's not easy. No, Adobe don't just want anyone to get access to the rules because they could cause all sorts of chaos with the account.

Firstly you have to have version 15. These rules don't work in version 14.

Secondly you need to pass a little test, which is slightly more complicated than it sounds. Be sure that you read the instructions on the test centre. My advice, without wanting to give away what the exam is about, is to scrub up on context data variables.

At the moment the test is free (look up knowledge base 10655 for the detailed instructions), but don't bank on it always being free!

What are Context Data Variables?

This is a very convenient sub section of this blog post, isn't it? Context data variables are variables that you can set in your on page code that don't have any effect.

Sounds odd, doesn't it? Taken out of context of processing rules they make little sense (apart from in VISTA rules), but in the context of processing rules they make a lot more sense.

So the rules are that they are set like:

s.contextData['key']="Value"

in the html on the page (or in the s_code through an automated route). So for this blog post I could have:

s.contextData['Author']="Alec"

and now I have a variable that is being passed through, but not populated in any reports at the moment.

This allows me to decide at a later date which variable I want to put that data into through processing rules.

How Do I Create Conditions For My Rules?

Conditions for the rules are vital and you should think about this long and hard before you create rules about the conditions that you put on them. It's rare that you'll want to blanket your rules on all of the values that you have associated with them, so think about when and where.

You have many more options for conditions than you do for actions. There are a whole list of variables that you can create your conditions on, that you can't take action on (more of that in a second). Most notably you can take information from the referrers without being able to do any action to change the referrer.

This won't help solve the problem that some of your twitter referrers are t.co now instead of twitter.com, for instance, but it might help you isolate them into another variable by creating a condition that if the referrer is t.co or twitter.com then it puts the value into another report.

What you can create your conditions on is the value of any of your existing standard or custom conversion and traffic variables, plus any of your hit attributes (referrer, user agent, IP address, etc). Interestingly you can also use query strings of urls and query strings for referrers - this is useful where you've forgotten to put campaign codes on links into the site.

Note that you have a number of options for you conditions, as you'd expect. You can use it for situations where the value is equal to something - this will create specific conditions on an only option. But you can also set it when values start, end or do not start or end with your option. There is also an option for if the variable has been set or not. This is useful for populating a report where you've forgotten to do it in your s_code (particularly important on your pathing reports!).

How do I Create Actions For My Rules?

The actions you can do are slightly limited. You can change any of your custom variables that you could set in the code on the page is a good rule of thumb (with the exception of the products variable).

This rather conveniently includes setting events. Going back to my original post from last year was the annoyance that I couldn't set events based on rules in the background and this effectively solves that problem. I also suggested that I'd like to default custom traffic variables being copied to custom conversion variables - you can effectively do this through the processing rules as well.

On a side note, I'd recommend doing both these things through the coding on the page if you can (or via these rules until you can do it through the code). You are limited to 50 rules, so you could only do 50 of your custom variables.

Your action rules allow you to do a number of things:

Overwrite data with new data. Although technically this is called overwriting, if your variable isn't set in the first place you are overwriting a blank value. Note that these processing rules happen in order, so if you overwrite a variable in rule one, in rule two you are going to be looking at the variable you have overwritten, not the original variable set on the page. You can overwrite conversion variables, traffic variables and standard page related variables (page names, urls, hierarchy, etc). They don't have to be the same variables that you set your conditions on in the first step. You can overwrite them in the following ways:

A value from another variable (so you could choose to overwrite prop1 with the value from eVar 1 for example)
A concatenation of other variables (so you could overwrite prop1 with the value from eVar1 and eVar2)
A custom value (so you could set your prop1 to be 'Alec' if you really wanted to, based on your conditions)

Delete the value of the data. Again this doesn't have to be the same variable as the one you set your condition on, but you can remove the data point if you decide that you've set it in the wrong place.
Set an event. You can set any event that you like and you can tell it to be whatever value you like, including making it equal to any other conversion variable, traffic variable or event.

Some Examples of Processing Rules

Off the top of my head here are some common examples of places you might want to use the variables:

Creating a custom conversion variable for your referring domains, keywords, etc report. This sounds slightly ridiculous - copying one report to another, but custom conversion variables have one slight advantage over the standard referrer reports. That difference is the ability to do SAINT classifications on them.
Capturing different campaign parameters in the same report. One common mistake is using the wrong campaign parameter on inbound links and it being too late to do anything about it after it has been created. Using this report you can capture your campaign variables in the campaign report irrespective of how they were set up. There is also the option of creating campaigns where none existed before by using the rules to create a new campaign
Creating conversion events in a similar manner to Google Analytics. If you want to create some rules for your conversion events then this is possible here.
Split your campaign variables into different reports. If your affiliate time out window is 90 days, but your paid search time out window is only 30 days you'd probably want them in different eVar reports so that you can match them with how the third parties report with their expiry rates. However you'll also want them in the same report so that you can monitor their performance in a standard way.

And many, many more. I'm sure I could go on for ever!

5 Tips for Using SiteCatalyst Version 15 segments

2012-01-26T22:40:00.001Z

With an ever increasing number of people going onto version 15 of SiteCatalyst I thought that it would be a perfect opportunity for me to write a blog post about segmentation, which is the major functionality improvement. It looks like it is something that should be easy to do, but if you follow my easy instructions below, you'll get the best out of it. Remember, the data you get from it is worthless if you don't do anything with it!

1. Page Views Based Segments

It seems to make sense to start with page view based segments first, because they are the lowest denominator in the world of web analytics.

But in the case of segments, I find these things the most difficult to deal with and I'll try and explain why in the hope that it will make it easier for you to understand, as perverse as that sounds.

The main reason that I have trouble with page views based segments is that I can't really relate them to visits, which is the metric that I use most frequently, but I think I have some diagrams that will help explain the situation:

I'll explain in a minute how to create the segments themselves, but lets get our heads around the values first. Your segment only includes a subsection of all the pages on your website. The visit and visitor do not get counted unless they land in your segment. This has a couple of implications:

The entry page to the whole site may not be in the segment, so the segment's entry page will be the first page in that segment that the user arrives at
This also means that the referrer when the user does eventually get to a page in the segment is going to be from an internal source (not the original entry source)
The user may leave to pages outside the segment before returning - in pathing reports SiteCatalyst will not represent the leaving part
The exit page to the whole site may not be in the segment, the exit page of the segment is the last page the user viewed in their visit in that segment
A visitor may have visits to the site which don't touch the segment affecting the visit number reports - a New Visitor is only new to the segment

The best way to think about this is if you have two completely separate websites - one for your segment, one for everything outside your segment. If you were looking at analytics reports for one of the sites, you wouldn't expect any visits to the other to be included, nor would you expect to know about what the user did before getting to the website.

We have three ways that we can create segments:

Creating a defining rule to show a number of pages to include (or exclude)
Creating more than one defining rule to show a number of pages, of which the rules have overlap
Creating more than one defining rule to show a number of pages, some of which may not have overlap

Creating the segments is very simple itself. Once you've clicked the add segment button you get a screen a bit like the one below:

You just need to drag your container from the left hand side and put it in the box on the right hand side. You have two options here - you can either have an include or an exclude. If you use an include then it will only include pages that you specify in your filter, excludes obviously do the opposite.

One note - try not to have double negatives because SiteCatalyst doesn't like them.

You can then click on the 'Page View' or the little pencil icon in the corner to edit the details about your container to choose your options:

Here you have the choice of using any of your standard or custom traffic or conversion variables. Any pages where the values in these pages match your selected filter will be included in the segment (don't forget to give it a name!).

There are several options for how you set your filters:

Use greater than and less than for numeric values
Contains, starts with and ends with are great for situations where you want to do a search type filter
is null is great for situations where you can't work out which pages don't have certain tags

2. Visit Based Segments

Visits based segments are much more intrinsically obvious. You deal with them all the time in the world. There the staple of Google analytics segments, so I won't dwell too long on them.

However there are some things that you should know about them. If you create a segment based on visits, then it will include the whole visit, not just from the point of that it qualifies for the segment.

The other thing to note is that this may affect the way that your conversion reports work. If your conversion report is set to a length of time different to a visit, then it's possible that your reports may be slightly misleading. Remember that in conversion reports, the value of the conversion is related to the time of the event, not of the time the value was written to the report. So you could have a report that is showing you values from previous visits in your report, if you event is included in the segment.

For example, my segment could be people who viewed the home page. But I didn't view the home page on my first visit when I came through a campaign, so my first visit isn't included in the segment. In the second visit, when I am included in the segment, my conversion on my campaigns report, set to 30 day time out, converts against the campaign from the first visit.

3. Visitor Based Segments

Visitor based segments work in a similar way to visit segments, but instead of including just the visit, they will include everything that the visitor has done. This leads to some interesting scenarios:

The visitor might have done the action that puts them in the segment during the time period of the report. For example I set up a segment for people who have bought one of my products as a visitor segment. If I am looking at a report for December 2011, it's possible that I am looking at visits and visitors where they bought that product in January 2012.

It's also worth remembering that the rules for the custom conversion reports still hold - if you create a visitor segment based on the value in one of your eVars, but the time out for that report is set for visits then subsequent visits by the visitor will convert against the value for that visitor.

This has some good effects. If you create a segment for people who come to the site through Facebook, you can look at the referring domains report to see what started the visit that actually caused the conversion. Don't forget that this visit from Facebook could have been after the conversion.

4. Event based variables

You can also add events in as your containers. This sounds a little odd, but remember that you can set events on any pages (or on clicks) that you like, so creating something that only shows when the event occurred could make sense, although it does bring us nicely into the next section...

5. Combinations of the above

Event based segments only really work where they have one of the other three containers involved first. You can add a container that is for visits and then specify that the only visits you want are the ones where they had an event in them. This is great for looking at just those that bought something or those who signed up to emails and so on.

Obviously you can also have visitor segments which only look at particular visits or particular pages and you can have visit segments that only include a number of page views.

However they don't work the other way around, you can't specify a subset of pages and then say only give me visits who came from search engines. Follow the heirarchy and you'll be fine.

2011's Predictions Revisited - How Wrong was I?

2012-01-03T22:48:00.000Z

At this time of year, it is traditionally that people write their predictions for the year coming up. They usually make a series of bold claims that subsequently start to look ridiculous or they make a series of claims that are so mind numbingly obvious that everyone gets bored after a couple of seconds. Of course the next year nobody bothers to check on their predictions to see whether they were wildly off or particularly astute. Given that I did the wild predictions last year, lets revisit those claims and see how stupid I really was.

1. The Fall of Google

Ah.

This is going well already, isn't it?

So Google didn't fall last year at all, in fact, it went from strength to strength. It is still the number one visited website according to anyone that matters (Alexa, Compete, Comscore, etc).

But it hasn't all been going Google's way. Why did I think that it would be their downfall last year? The reason I thought was spam. Particularly I was looking at spam that was 'user generated content' within affiliate websites. Sites which offered 'reviews' to give value add, which were becoming increasingly nonsense and no use to the user at all. I thought that users would soon start switching off because of this.

However what I didn't foresee was Google making some major updates to the way that the algorithm worked:

Google Panda was an algorithm update in February. One of the main differences was that newer content hit the top of the rankings sooner and pushed older content further down the rankings. It also seemed to adversely affect affiliate websites and websites that do scraping of content. It also reduces the impact of mass produced pages.
Google Plus and Plus one is the start of a major change to the way Google creates its personal algorithm. Individual users are going to have their own personal algorithm based on what their contacts have been doing.
Google changed the default search to secure. This will mean that they can transition those who have Google accounts to use their personalised data for search. Particularly this will impact those who are using Gmail, Google Plus and other systems that frequently run in the background.
More Personalised Paid Search. Whilst all the organic updates were going on, one big change that was often missed was the move towards personalised paid search. This was always the fall back of those sites with poor reputations, but it may not continue in that way as users push down the paid search ads from undesired websites.

So Google certainly haven't been quiet this year in making sure that my prediction hasn't come true. One could argue that they've made themselves much, much stronger than they were in the past (which was pretty damn strong).

2. The Death of Online Newspapers

Erm...

Yes...

I'm going well so far.

So it turns out it wasn't the death of online newspapers. The MailOnline hit 80m unique browsers in a month in November 2011. This is up 73% on the previous year. The Guardian was up 59% and the Independent was up 36%. So how did they do it?

Well the Mail decided that rather than being niche it was going to go as mainstream as it possibly could. 55% of their unique browsers now come from outside the UK, although 67% of their page impressions come from inside the UK suggesting that they are attracting a lot of people who are just floating past (probably through search). This is a pattern that is matched at the Guardian.

So maybe some more newspapers have gone behind a subscription model? Well if anything it has gone the other way around. The News of the World was News International's attempt to see if it could make a tabloid work behind a subscription model after moving the Times there. Unfortunately that experiment went by the wayside as they were forced to close amid a huge hacking scandal that hasn't quite reached its zenith yet. It remains to be seen whether the Sunday Sun will follow the same route when it arrives.

3. The Mobile Web's applications find wider use

I'm on safer footing with this one at least. This was an important year in mobile application building for one big reason. The big reason is that Adobe decided that they weren't going to update their Flash for mobile browsers. This doesn't sound like a particularly exciting move to the average bystander, but it has far reaching implications.

Firstly this is an admission by Adobe that they'll never get Apple to have Flash in Safari on the iPhone, therefore missing out on one of the biggest markets in mobile. Notably they are going to continue to help developers with Flash on applications.

Secondly they are going to start helping developers work towards developing for HTML5, which is standards compliant. This means that as mobile browsers get more sophisticated they will be starting to show what is on the actual web page, rather than developers having to code for different devices.

It's an interesting move because the shift in the mobile world is a significant one at the moment. The growth in the Mobile market over the last twelve months has been to Android:

Data from Comscore via fiercemobilecontent

Android of course being open source means that you can develop any applications you like for it and do whatever you want with it, including creating new browsers. My current browser of choice on the mobile is Dolphin.

4. The boom of use of Analytics tools

Last year at this time I said that there were 94 jobs in the UK for Web Analytics on Total Jobs. This year there are 119. Is that evidence of an increase in use of Analytics over the last year?

There was a continuation of the acquisitions, most notably as Adobe started buying companies that not only had technology, but also did consultancy. Google also introduced a 'Premium' version of the tool, which you can pay for to get support.

Who knows what will happen, but the industry is facing a major crisis in the coming year. As the rest of the world watches, Europe is introducing new regulations on cookies as the US watches on. If the US follows the EU example then it could well spell the end of Web Analytics as we know it. If the EU experiment fails then we're likely to find hundreds of US websites flooding the area that the EU companies lose competitiveness in.

Certainly it is an interesting time to be part of the industry! It's make or break time.

Time is running out: get your cookie policy right

2011-12-14T21:08:00.000Z

Yet again I am sitting here writing another post about cookies, when I'd much prefer to be out eating some. I think that this will eventually get me down to the point where I will just flop on my keyboard and the post will just a be a garble based on what my ear lands on and where I dribble. But we should at least try and look at this in an objective manner, because the ICO has been issuing more guidelines this week on what we should be doing to comply with the cookie laws that the UK has been handed by the EU.

Just a recap for those of you interested: In July 2010 the legislation was ratified at an EU level to make it so that cookies had to be opt-in. The IAB and the web analytics association decided to look into what the new cookie legislation meant with the hope that it would go away. Meanwhile companies were coming up with new ways of doing tracking of users (such as the one that I work for). Then in May this year, the ICO confirmed what the new laws meant, the press reaction was not pretty. Nobody really wanted to do opt in (I even offered my own solution!), when in fact it looks suspiciously like the ICO is missing the point of data privacy as it fails to prosecute those who break regulations.

So where does the new documentation fit in?

Christopher Graham at the ICO says:

This report is to be read alongside an updated version of our advice (the earlier version was published in May this year).

You can find the report linked to in the report by Christopher, but the long and the short of it is that this is regulation already and having a 12 month window until May 2012 doesn't mean you can just ignore it an it will go away. Six months in, he says:

The report can be summed up by the schoolteacher’s favourite clichés: “could do better” and “must try harder”. A report that listed the URLs of sites that were perfectly compliant from day one would be very short indeed.

So you don't have to read it, I'll help you with some of the summaries. Dave Chaffey over at Smarter Insights has a very useful piece showing what he thinks will be allowed and what won't be allowed under the new legislation. The long and short of it is that things that won't be exempt:

Analytics cookies - you will have to ask users if they want to opt in (don't forget that the ICO saw a 90% drop in their traffic when they implemented theirs).
First and third party advertising cookies - you will have to ask users to opt in to your advertisers if they do any tracking of their ads (or somehow get them to in ad).
Affiliate cookies - you will have to ask users before putting these on their computers

Of course this is all a bit ridiculous because of the following situation. You visit Facebook and get a cookie when you log in. This cookie follows you around the web when you look at a page with a 'like' button. You've already accepted it, so there is no reason for them to prompt you to accept again. Facebook can continue providing all the tailored adverts that they like. Even if you log out, you still have the cookie.

This is of course true for Google, Microsoft, Yahoo!, etc. So anyone who issues a first party cookie on their site when you log in to one of their services suddenly has a massive advantage in providing tailored content. Your Publishers will just switch to these companies and the users will be no better off.

The key point is not who obtains the consent but that valid, well informed consent is obtained.

The report also doesn't state where the users should get consent - presumably that is at the point of setting, not the point of accessing (otherwise you'd have to get consent on every single page, which would be a nightmare for website and user):

An organisation with several connected websites could in theory obtain consent for cookies set on each site in one place, for example when the user logged in on one site. In order for this consent to be valid it would have to be absolutely clear which websites the cookies in question were set on, what those cookies were used for and exactly what the user was agreeing to.

So the point of the process is twofold. Firstly you should aim to give users more information about cookies and this seems obvious. Privacy policies have long been vehicles to get you out of trouble legally, rather than to inform users. Creating a page about cookies and how you user them would seem a simple task, as would making it prominent on the page.

However I disagree with the assertions of the document. Even if it is more prominent on the page, it doesn't mean that people are more likely to read it, especially given their four solutions:

For those of you who can't read that - they suggest you make it a different colour or give it a different name, but effectively keep it hidden out of the way.

So lets look at the techniques for getting opt in consent that they suggest:

Pop ups and similar techniques (such as message bars and splash pages)

You'll be reminded, of course, that the ICO itself uses this method on their website - their is a little message bar at the top of the page:

This was so successful that 10% of the visits actually signed up to it.

Personally if I was going to suggest one of these, I would go with the splash page. Many advertisers do it and actually it would be a relatively simple implementation. You could add a bit of JavaScript to the bottom of the page that would load on the first page view to cover the whole screen and invite the user to accept cookies.

Get Consent when the User Signs Up or Buys

This of course seems a bit late in the process. You've already lost your users information about how they got to the site, what caused them to sign up, or buy. You won't be able to make any marketing decisions about this user from how they got to the site. You're adwords data is effectively useless now because you don't know what the return on investment is (unless of course you've got a Google Conversion tag on your conversion page, the user already having signed up to Google's cookies just by viewing Google).

Settings Led Consent

For websites where the user can make some settings that personalise the website for them, this is an option. When the user says "I live in London" you can tell them that to remember this, you are going to store it in a cookie and that you are also going to use the cookie to track the user across the website to make it easier for them to use.

It's interesting to note that the ICO has repeatedly responded to arguments that this isn't practical:

People say this law just isn’t practical – what happens if I do nothing and wait for it all to go away?This isn’t going away. It’s the law. The UK Regulations come from a European Directive that was passed in 2009. The requirements cannot easily be changed and cannot just be ignored. Many organisations are making a lot of effort to comply. The Information Commissioner has been clear that he will take a practical and proportionate approach to enforcing these rules where organisations are making the effort to comply.

Personally I see only three options for you and your website:

A Splash screen asking the users to accept cookies. I don't see this as a good option for users or website owners: users universally hate splash pages and won't know the difference between this and an advert.
Ignore your analytics and remove your competitive advantage over non-European websites.
Ignore the ruling and hope everyone else does too. The ICO can't fine everyone, right? RIGHT?

There is lots of data to analyse despite Google's Secure Search update

2011-11-29T19:57:00.001Z

A lot has been written recently about the introduction of Google secure search means for websites. In fact, I've been at the fore of writing about it myself on the Digital Transparency blog that I've been contributing to recently, so I know all about the subject. Well, I know a little bit about it, which is what I'm going to share with you now.

In case you've missed the memo, Google recently changed the way that it did search quite drastically. They now make it so that if you are logged in to one of their products then your search will be done through a secure environment. You may remember that when Google plus first came out I theorised that they were going to use it to make search more personalised based on what sort of stuff you have plus oned. I still think this will happen, but the first thing that Google needed to do was obfuscate the url in the search bar, as this is the way that they are going to make it personal. If you put it in the url then you decrease the security for the user because this information gets passed over when you click on a link. They obfuscated it by making it so that the referrer doesn't have a search term associated with it at all.

So what does it mean? Well it means that in Google Analytics you now have (not provided) as a search term and in SiteCatalyst you have Keyword Unavailable.

In fact if you look at the graph above, it is quite a lot of visits that it is providing - up to 10% of the search visits for this blog, but many others are seeing more and it does really depend on your audience. My audience is slightly more technical (I think!) and more likely to be logged in to something like Google Plus or Gmail than 'average' audiences. But remember that no audience is average and the figures you are going to get will vary.

As Avinash Kaushik points out in his blog post on the subject - there is a wealth of information that you can find out about the (not provided) data in Google Analytics. In fact there is a wealth of information that you can find out about the visits from all your tools. You can find out exactly the same set of information about the visits as you would be able to from any other segment of visits.

Of course one of the points that Avinash makes is that it seems like it is a large value because it appears near the top of your keyword list, but it isn't actually because it is comprised of lots of small keywords. It is worth remembering though that this data could be slightly biased because of the segment of the users. This is called 'sample bias' and is something that you should all know about.

Of course this isn't the information that we want from the data. We want the original stuff back. Analysing it in this manner is useful, but it is only providing a bit of detail that we knew we could get anyway.

Given that we're not going to get the original stuff back, what else can we do? We can still use the data that we have to continue doing conversion analysis (with the data bias assumed) as Avinash says. We can do it for the terms that we have and we can do it for those that we don't have.

What we can do is start looking at how we combine together some different data sets. A recent change to the way that Google Analytics works was an introduction of some of the web master tools reports. You may be sitting there and thinking 'So what' at this point, because this is data that you already had, but this isn't quite true. In webmaster tools you only get one months historical data, whereas in Google Analytics it will be for as long as your connection has been set up.

This much longer data set means that you can compare it to your visits in a much more efficient manner. For example I can see here that the number of queries that my search terms appear in doesn't seem to have gone down recently, despite a small dip in the number of visits from searches.

The other advantage of this is that you can filter the reports by your search terms as well to produce some graphs showing the data for just a sub set. You can try to improve for a particular search term and then see how often you appear in results because of that. You can also see the effect of news or changes in search popularity.

The advantage of these reports is that I can quite handily also look at not just keywords, but also of the pages and how often they appear. This can provide me useful information on which types of story are generating search terms over a period of time. Those that are doing well can have more posts written about them and vice versa. This is useful when writing new posts - I can tell if they are gaining traction in search engines or not. My last post on Conversion Funnel Analysis in Google Analytics didn't seem to be getting much search traffic to start with, but I can see with this report that it is starting to appear in search queries now.

But of course you shouldn't treat these reports in isolation. What search terms are generating queries to a certain page? Well Google Analytics has hit its limits here, but you can go back into your webmaster tools and start looking at that data too. This has the ability to break the search phrases down by page or vice versa (over a smaller time period):

So we have two different data sets that will give us lots of different bits of information that we can use in lots of different ways. It is a shame that we've lost some data, but there is still loads for us to analyse.

How to do Conversion Funnel Analysis in Google Analytics

2011-11-14T20:30:00.001Z

Last time I suggested that I should write a blog post about funnel analysis in Google Analytics, because all my conversion funnel analysis blog posts had been successful. So this week I am writing a blog post about conversion funnel analysis. It's almost like this blog has a plan week on week, isn't it? It doesn't, I can assure you that next time I'll be sitting with the computer in front of me wondering what on earth I should be writing about again, but there you go!

In case you've never come across this blog before, one of my early blog posts was on conversion funnel analysis: When, How, What, where I discussed what conversion funnel analysis is and what you should do with the data. I then followed it up with a post about how you do conversion funnel analysis in SiteCatalyst, which not only showed you how to do it in SiteCatalyst using fallout reports and so on, but it also expanded a bit on the principles to show that it wasn't as simple as it may seem. If you are interested, I also wrote a blog post on the sister website to this one, Digital Transparency, on how conversion rate analysis is about targeting as well as testing, which touches briefly on funnel analysis.

How To Create Funnels In Google Analytics

Creating funnels in Google Analytics is relatively easy. Relatively easy in the sense of the process is easy to set up, but it does require a little bit of thought. More of that later. Firstly though - this is how you set it up:

Click on the settings button in the top right hand corner of your page
Click on Goals in the second horizontal menu (next one to assets)
Click on one of your free 'goals'
Choose URL destination
Check the box for the funnels at the bottom
Add your goal and funnel stages

This process means that you need to set up your goals and funnels in advance and you need to think about the funnels that you are going to want. There is no retrospective data processing involved in this.

Note that you will be asked whether you want the first step in the funnel to be compulsory. If you check this box it effectively means that the goal will not convert unless the user has gone through this step before converting. It's a bit of an oddity, because whilst the users have to go through step one to get to the goal, they don't have to go through step two or step three. This is a vital bit of information because it relates to not just your conversion funnel but all your goal reports as well. As we'll see in the next section it isn't vital that you do it for your conversion funnel.

What Do The Reports Look Like?

Getting on to the more exciting part, you really want to know what the reports look like. Well you can access them in the left hand menu in the Conversions tab.

So these reports don't really visually represent the way things move through in the way that SiteCatalyst's fallout report does, but it does give you some interesting information. Because I didn't select that my steps were required in the report we have the options of seeing how many visits enter into the process halfway through as well as how many drop out halfway through.

This relates back to our option for having the first step as compulsory. If you selected it then your process will be limited to just those who also went through the first as well. The second step in this case will only include those that went through the first step. An oddity of how the funnels works means that your second step could quite possibly be lower than your total for the goal (if the users bypass it).

This is an interesting concept that I expanded upon on in my last post on Conversion funnels. Not all those visits that get to the goal will have come through your funnel route. This is especially true where there are large volumes of content that your users could arrive at halfway through.

But this process is very useful when you have a payment process (shopping cart, credit card details, delivery address, confirmation, etc) to be able to look at how people drop out of your process. This can help inform you where in your process people are dropping out and you can look at improving this and monitoring it over time. This is the vital part of the process - you need to change something and monitor how it changes over time.

What Should You Do With The Reports?

Well normally I'd sit here and tell you to segment, segment and segment some more, but in this case we can't do that because there isn't an option to put your advance segments on here. What we can do though is put in some custom filters to create alternative reports. It's a bit of a fix for something that should be done automatically in Google Analytics.

So if you create a new profile in Google Analytics for the top segments that you are going to be interested in using the profile builder. Don't forget to make sure that you have a standard filter that doesn't exclude any of your data, otherwise something may go wrong and you won't have the raw data in the background.

Using a profile of your data is important when looking at funnels. Each of your segments is likely to react in a different way when going through funnel and one of your jobs is to try and work out a way of getting each of your segments to get through the process as best as possible. This may mean that you need to have more than one process for each of your segments, but this is all about targeting as well as testing.

I'd also recommend creating a diagram like the one I suggested in my last post on this subject:

Creating a visualisation like the above will allow you to work out if there are steps in the process that you are ignoring. Are people moving in and out of the process all over the place or is it really a traditional funnel? Does one of your sources into your goal provide more of the traffic than others. Should you be focussing on a different journey for your users or the one that you've set your funnel up for?

Don't forget that you can set up several funnels for the same conversion point (if you have enough spare goals). Don't be afraid of setting up more than one that are very similar. You'll get lots more data that will give you a better clue as to what you should do to improve the process.

4th Birthday for whencanistop

2011-10-30T22:53:00.000Z

Happy birthday to me, happy birthday to me, happy birthday dear whencanistop, happy birthday to me. Yes, four years ago the humble blog that you are reading at this precise moment was born (well it was born four years ago since I wrote this post, you might be reading this months or years since I wrote it!). So the question on your lips, is presumably how has it all been going? In case you were wondering, yes there was a similar post when the blog was two years old, along with a few posts before that about how the blog was doing.

In the four years of this blog my life has changed significantly too. When I started doing this blog I was working for RBI, a company that owned a series of 'news' websites aimed at the business to business sector. Blogging there was something that was relatively new in a business sense there and so this started as a bit of an experiment (firstly internally, then on this blog) to show some of the organisation about how it could be done. The start of a decline in advertising revenue meant that I decided the time to seek new employment to further my career and I started working for Serco on BusinessLink - a Government website aimed at businesses. This was a slightly different world from that of RBI - much more risk averse and that meant my blogging was a bit more conservative (little talk about work!). Now in my current role as a Business Analyst at a consultancy firm the shackles are back off again, but not quite as much as at RBI (you've got to protect the customer's privacy!).

So how has the blogging been going? The aim was always to write two posts a month and make them of at least 1,000 words each. So how has that been working out? I've included this one, just to massage the figures a little bit:

Two blog posts a month seems to be working at the moment. I'm also working well towards 1,000 words per post. This year I have only written one post that is under 1,000 words and two that have been over 2,000 words (averaging 1,350 words).

So how have the visits to the site been working out? That is as much a measure of the success of the blog as anything else:

I think that I could have expected a slightly larger increase year on year, following on from a record month in October 2010. There have been three months that have had over 1,000 visits in the month and that suggests that this is a successful blog.

Also useful information to know is that in that time my search traffic has increased month on month more or less throughout the entire time. In fact, most of the peaks and troughs that I have in the visits to the blog are caused by situations of large numbers of visits directly or from referrers (typically twitter, but I do also get a large number of visits because of the blogger tool bar up there, I don't know what causes me to appear when people click on 'next blog' but I'm sure it is something exciting).

One of the advantages of writing lots of long blog posts is that you get lots of long tail search traffic. This is particularly true of this particular blog. 8,757 different search terms have generated 14,480 visits. It's difficult enough to work out what is going on with a few keywords, but when half of your visits from search engines come from a phrase that was used to visit your blog just once, it becomes increasingly difficult.

In fact, just by looking at the visits generated by my number one search term, funnel analysis, give you an indication of how difficult it is:

Rarely do I get more than 10 visits in a month for the search term and I haven't got more than 20 a month at all. In fact, I did the same graph with all phrases that included 'funnel analysis' and it was only about double that. So maybe a more efficient way of looking at the data is by looking at the landing pages that people landed on:

So given this piece of information (I've included the top 11 here, because the home page is in the list and that is mainly brand related traffic), it is interesting to note that my conversion analysis post is the most visited. Maybe it is also interesting that I followed that post up with a post on conversion analysis in SiteCatalyst, because of the success of this post. It is also number 7 on this list. A wise decision you may feel. My next project should be to create a post on conversion analysis in other tools (such as Google Analytics). And actually, if you look at the data there was a marked uplift in traffic to those posts from search at the time of the publication of the second post:

So this makes the last part of this post interesting. Lets look at a drilldown of content based on the time of the post was put up:

Now this looks interesting that 2008's posts have generated the most traffic (remember that my conversion analysis post was from 2008). But it is an interesting way of looking at the data. Of course 2008 was the most viewed content since 2007 - it had the most posts and the most time to generate traffic. It is hardly surprising that if you look at just this year 2010 and 2011 have generated the most visits.

Finally looking at the actual content that people are looking at:

There is a post from 2011 in the list (all be it at the end of the list). This is the sort of thing that I will use to help build the next couple of posts, given the interest that the readers have in the subjects. It won't stop me from producing posts that are more news related, but I am mainly going to be keeping those to the new blog that I've been writing at digital transparency. The top ten for those of you who want to click:

Six Tasks for creating a new blog

2011-10-17T21:07:00.001+01:00

A lot of my time in the last couple of weeks has been taken up with a new blog that I'm working on for work. In case you haven't realised, I am going to be a regular contributor to the website Digital Transparency. What I thought that I'd do today, is give you some general pointers on it that I've learnt along the way and things that I still need to do, to give you a good impression of what to think about when starting a blog.

Task 1 Decide which domain to use and redirect any others

Obviously one of the first things that we had to do was make sure that we had the right domain name. Whilst we owned digital-transparency.com, we didn't own digitaltransparency.com (at the time). We've since purchased the other domain name off someone who wasn't using it and was open to offers. One of the first things that you'll notice if you click on the links above is that they both resolve to the blog. That's good in a way, but really we need to make sure that one redirects to the other for the best possible SEO. We haven't done it yet, because it was a relatively recent thing and we haven't had the conversation about which we'd prefer.

Incomplete

Task 2 Decide which platform to use

Whilst this blog is on blogger, we went for WordPress on the other blog. When I set up this blog I didn't really know what I was doing. Almost four years later and I haven't really made many updates to templates in the background, bar the odd bit to make it a bit wider and more usable. The background is still the same one that you'll find on many blogs and the colour scheme isn't great. I don't really have many plugins going on either because this is such an archaic template they don't work that well. I could transform it into a more recent version, but it serves its purpose (you can see my posts!), I don't have any graphic design skills and I don't really have that much time.

WordPress on the other hand is a tool that is easily customisable. With a large technology company behind me I felt I would be more able to adapt that blog, especially as we want to do many more things with it than the occasional post every other week. Moreover we have our own hosting set up already (one of the advantages of blogger is that they do it for you), reducing an additional cost that you can often get. So installation was easy and set up wasn't that hard. Plus I got our technical guys to build a branded banner, so it is all looking nice.

Complete

Task 3 Make sure the post has friendly URLs

This is a much easier task and I'm glad it is something that I did right from the off. Firstly one of the problems that I've had in the past with blogs arise from situations when we've tried to move them or tried to do some more advanced SEO. Frequently what happens is that you lose all your existing urls and you have to set up a whole load of new ones. This is fine if your blog is days old, but not so fine if has an established set of links pointing to the old urls. Sure you can do some meta refreshes for those that accidentally find the old pages, bur really you want to 301 to the new urls for the benefit of the search engines.

So I've made the posts so that they have the title in the url, something that is quite simple to do in WordPress. In addition, I installed a little plugin called HeadSpace2 that allowed me to state what I wanted as the meta description of the post. This is really useful for blog posts where you don't necessarily have an 'excerpt' that makes any sense as a descriptor (the first 140 characters of this post wouldn't be very useful to the person who finds this blog through search).

Complete

Task 4 Link to Social Media

When I started this blog almost four years ago Twitter and Facebook weren't really being used for much. Now I'm active on Twitter, Google Plus and there is a Facebook page (which I do keep meaning to do something about). They could have been invaluable when I'd first started, so I've created a new Twitter persona called @digtran. I'm going for a slightly different model with this feed. I'm not going to follow anyone else. At all. Nobody. It is solely going to be a news feed of all the blog posts that are coming out.

That may sound like a crazy decision - but we already have channels to do that through Twitter. As well as virtually all of us having our own personal account, there are localised versions for the Corporate name as well (I occasionally post stuff through @adversitementuk, but Nick does that more frequently). The other thing that I have learnt from my own username on Twitter is that is has to be short - too many characters makes it difficult for retweeting - hence the short name.

We're also looking at what we do with Facebook, Google plus and Linked In. At the moment we're not quite there with these, partly because the organisation has lots of different types of facebook pages and we want to group them all together.

Partially Complete

Task 5 Set up Analytics on the pages

Obviously I had to set up Analytics on the blog. In fact Google Analytics was probably the first thing that I put on there. We also have access to SiteCatalyst within the organisation (obviously), so I've started the process for putting that on as well.

As you can tell, I've started that process as well - I wrote a blog post about the things that I should be tracking in SiteCatalyst. However I haven't quite managed to put any of that into action yet. One of the reasons that I haven't done so yet is that there is still limited content, so it doesn't matter too much (I've been the only one to post so far). Also I haven't done anything about making sure that the cookies that tool uses are first party cookies. That is definitely a job for after we've decided which domain to use.

One question you may ask is why I can't just make do with Google Analytics. I don't have that much exciting going on that I couldn't just use it, but the rest of the company uses SiteCatalyst for all their websites and we need to make sure we can have reporting that comes through the same systems for all the websites. It also needs to be comparative - Google and Adobe both run different processing in the background which means that whilst their figures may be similar they won't be the same.

Partially Complete

Task 6 Start Writing some posts

Having set up the blog, the next thing to do was start writing some posts. I've been doing it daily since launch and plan to continue doing that. But we have a longer term strategy for the blog, because me posting 'news' style updates isn't going to accomplish the plan of driving more business to the company.

So we have a plan for the next couple of weeks and months to start adding to the blogging people in the organisation to get to the point of between two and three posts per day on average. This should be accomplished with one blog post by me, which will be news related and another longer, more educational blog posts by the 50 odd other people who work at Adversitement. To do this it will mean that each person will have to write about a post every month. We think that this will be doable. In the long run the plan is that we'll have guest posters who will blog under their own name (for exposure for them) but on our Digital Transparency blog.

The next step in this process therefore is to start putting more profiles of users on the site and also to expose those people in a more efficient manner (a custom home page is the first step).

The start of a long road

An SEO review of the UK Government Websites

2011-09-30T23:18:00.000+01:00

I've written about SEO in the past and I have written about the UK Government websites in the past (although, admittedly not on this blog!). So it seems sensible that I should write about the two of them together. The trouble with writing about SEO is that it is very easy to sit here and claim what could be done better, but unless you are looking at the data of the number of visits generated and the value that they provide. Therefore the following should in no way be miscontrued as a the Government is wasting our tax money post. It is more of a if they did these things then it might be better post.

Before I go any further, I'd like to point out some other blog posts I've written in the past. Firstly I wrote about how to measure a Government website's success (on eConsultancy), using BusinessLink as an example. That having been successful I followed it up with another one about DirectGov. I also wrote a post on what the Government should do with online Marketing, the number one point was to become an SEO centre of excellence.

Along with the plethora of stuff I've written on how to measure SEO, I've also written the odd post on how to do SEO. This is going to be one of those posts. You can tell, because it has been very dry so far.

One of the big changes in web at the Government was the consolidation of all of the department's websites into three big ones. The idea being that this would reduce replication of content and allow the websites to link related stuff together more efficiently. In practice many of those prior Government department websites are replicated on DirectGov and BusinessLink in silos with the information still not particularly well linked together, so they are very reliant on Google being able to direct people to the right places. Along with this, massive budget cuts in Marketing have meant there is very little way to tell the public how to get to the sites or what they contain. This is why SEO is so important and particularly why I'm going to look at all three websites BusinessLink, NHS, DirectGov), rather than just one.

Domain Name

Domain names for SEO have been a decreasing factor for years. The increase in long tail searches (especially for content websites) mean that your domain isn't going to be very high in the searches that people make. Like branded websites, the only point of the domain name is for those who know about the brand.

Direct.gov.uk wouldn't have been a great brand name in a non-advertising world. Fortunately they started when there was advertising (high profile tv, radio and national press as well as online) (8/10)
BusinessLink.gov.uk was a good idea if only because of the existing BusinessLink 'shops' meaning that many people knew about the Government's help with Business already. Now that they no longer exist it isn't a reason to change domain name as people will still know about the shops (9/10)
nhs.uk kind of makes sense because it is the website for the NHS (even if there are large parts of it devoted to health and well being) and is probably a bit more snappy than 'thenationalhealthservice.uk'. I'm docking them points though because they aren't a .gov.uk which would give them a massive boost in the search rankings (5/10)

Friendly URLs

Search engines like having friendly urls. The search phrases in the url will aid the search engine in working out what the page is about and is a good ranking factor. You'll notice, for example that the url of this post matches the title. Moreover if a users sees a link they want to be able to tell what the page is about so that they can decide whether to click on it or not (but that is another matter).

Direct.gov.uk works well with this because it manages to put the menu structure into the url, as you can tell with the example below - it sits in the education and learning section, is about university and higher education and is about life at university or college. I'll dock a mark because I think the ID that the database uses at the end rather than nearer the start, but it is probably much of a muchness (9/10)

http://www.direct.gov.uk/en/EducationAndLearning/UniversityAndHigherEducation/LifeAtUniversityOrCollege/DG_180817

With Businesslink.gov.uk it is a different matter, the urls are not friendly in the slightest. In fact they contain a series of query strings that get lots in the detail about where you are in the navigation structure (for the database) without being at all informative to the user or a search engine. I won't go into all the directories on BusinessLink (because DirectGov doesn't have any it would seem unfair) but they follow a similar ruling as well - IDs, but not words (0/10)

http://www.businesslink.gov.uk/bdotg/action/detail?itemId=1073791931&r.l1=1073858790&r.l2=1073858944&r.l3=1073981161&r.s=sc&type=RESOURCES

nhs.uk is a bit more complicated. It does have some of the same page types as the above two. For example see the page below. This clearly tells you what the page is about and contains useful keywords:

http://www.nhs.uk/Livewell/homehygiene/Pages/how-clean-is-your-home.aspx

(5/10)

http://www.nhs.uk/ServiceDirectories/Pages/GP.aspx?pid=9A71A5C7-2974-4B9F-B559-AF0EEC76A756

Duplicate Content

The effects of duplicate content have long been argued. The facts are that the more duplicate content you have, the more likely you are to confuse Google. More likely you are going to end up splitting up your links and causing problems with rankings. The general opinion is to avoid it as much as humanly possible and where you can't, introduce the canonical url parameter. At RBI we used to 301 rewrite all the urls back to the canonical version to avoid confusion. For this part I thought I'd take the three random links I took above (seeing as I can't do the whole website) for the test.

A search for the opening paragraph from article above results in 23 results. Only one of these is www.direct.gov.uk, however there are a number from m.direct.gov.uk mobile version of the website. This is a bit confusing for Google, normally you'd have something in your robots.txt to ensure that these pages aren't crawled or that there is a 'noindex' parameter on the page, however these don't exist. More importantly the canonical url parameter passes different urls for the same page on the mobile site, whilst there isn't one on the main site (6/10)
For BusinessLink the search produces 8 results, which is relatively good. Until you realise that six of the other seven links are skins of the main BusinessLink site (specifically for the Welsh, Scottish, Northern Ireland, Welsh language and Blackpool version of the sites). Fortunately there is a canonical link parameter on the site. Unfortunately so is there on each of the other sites that points to their own version. Google has clearly decided that the 'English' version is the one people want, so is the one you will see, so I don't know why the others are there (6/10)
The nhs.uk article has 136 instances of it being indexed. This is a large amount and is in no doubt due to the syndication that nhs.uk offers. But also it is because there are localised versions of the page - I found identical knowsley.nhs.uk and imperial.nhs.uk pages. This is annoying as Google treats subdomains in the same way it treats new domains. Worryingly there is no canonical url parameter (1/10)

Redirects

An important thing for Search engines is making sure redirects are done and that they are done properly. You can do them via 302 (temporary - in which case Google will keep indexing the original version), 301 (permanent - all links to the old version will be pointed to the new version), meta refresh (via the meta tags on the page) and javascript (Google won't follow them and you'll lose value):

DirectGov is a bit of a mixed bag. What they tend to do is archive old content, rather than redirecting it around to the new version of the content. This is an interesting way of doing it. You may need old advice to check to see if you were compliant several years ago. Wikipedia I always think is a good example that the old version will be given a new page under discussions whilst the new version takes the old url of the page. I think this tends to happen with DirectGov (although it is hard to tell). Also useful to know that their shortened urls (eg direct.gov.uk/taxdisc) do end up with a 301 redirect to the right place (9/10)
BusinessLink is not quite the same. They don't really do 301 redirects, they do most of their shortened urls via javascript redirects (eg businesslink.gov.uk/tax is a javascript redirect - turn off javascript and you'll end up on a blank page with a link). Not only that, but they also tend to not bother archiving old content and new content is regularly given new urls with the old urls just being lost (2/10)
nhs.uk doesn't seem to bother with shortened urls too often and when they do they use meta refreshes. These aren't as good as 301s, but are ok. NHS again archive data on the national archives, whilst appearing to ensure that the new content takes the old content's urls (7/10)

Meta Tags

Meta tags are the generic term (that I'm using anyway!) for a bunch of things that appear in the background of the page. Firstly you have the <title> tag that tells you about the page. It is the link that you have in the search engine and it is what appears at the top of the page. You then have <meta name="description"> tag that tells you what the page is about and is what appears under the title when you search. Finally you have h1, h2, h3, etc headers of decreasing rank of importance of headlines. Meta Keywords aren't used by search engines any more (but may be used by internal search or for tagging purposes). I've used the same pages that I linked to in the friendly url section here

For DirectGov there is a nice title that contains the title of the page, there is also a useful meta description that looks like it has been appropriately written for the page. The title contains the website name as well, which is always useful, plus it isn't first up so you get the value of the description of the page first. The h1 and h2 tags don't seem to be particularly related to the page. The h1 tag doesn't even appear on the page in text and the h2s are menus. Only when it comes to h3s and h4s does it actually appear ok (7/10)
For BusinessLink we have a useful title, the brand name coming at the end and a description that matches the content of the page. The h1, h2 and h3 in this case work really well with the h1 being the first title on the page (the section header) and the h2 being the sub title (the page header), with h3 being important sections and right hand nav headers (10/10)
For nhs.uk it is a bit different. The title doesn't really match the title of the page, plus it contains the title of the section as well as the website name. The description is very useful though, in true journalism style it matches the standfast (the introductory paragraph). The h1 in this case works well, but the h2s and h3s don't seem particularly related to the page (6/10)

Overall Score

Before I do the overall scores (and remember these are just for fun), I haven't taken into consideration a huge host of things. I haven't looked at the keywords they should be ranking for, I haven't looked at all page types, I haven't looked at bespoke landing pages and I haven't looked at inbound links. These are all things to consider when it comes to SEO. This score is just for fun:

DirectGov - 39/50
BusinessLink - 27/50
NHS - 24/50

The trouble with sampling data

2011-09-14T23:28:00.000+01:00

Well it is that time of the month again. Yep, the UK has just released it's official figures on the number of people who are employed. Apparently the jobless total is now at 2.5m for the first time as the political commentators keep telling us. The Guardian has a very nice section on their website that deals with the data so that you can see how it has been going. It's great to be able to build your own mashups.

Why am I mentioning this? Well last month in his weekly column, Ben Goldacre pointed out that these figures aren't statistically significant and that the increase could just have been statistical noise. Have we learnt from this? Well it appears not because the official Office of National Statistics points out that:

The difference between the adjusted LFS and WFJ estimates (351,000) is within the bounds of the sampling variability of the difference. The approximate sampling variability (95% confidence interval) is roughly ± 300,000 to ± 400,000 (based on the estimated coefficients of variation published on page 83 of the Final Report of the Review of Employment and Jobs Statistics).

Whilst this is pointing out the difference between two sets of data, it also raises the point - unless the figures go up by 300,000 each month, then it is possible that any month on month change is just statistical noise.

This is why it is important to look at the statistics over time. Look at the data on the BBC:

With this graph we can clearly see that the numbers are going up or down over time. A one month change of a square up or down could be within the realms of statistical significance. By taking a two month average you can see that you are going increase the precision because the sample size is bigger (remember my post about statistical significance?) because you're ending up with a group of samples. Obviously a three month average gets more and more precise.

So this brings up the next question - why are we looking at sampled data here? Well it turns out that the Government doesn't really know the number of people unemployed because it's based on a couple of guesses. The claimant count - well that is a real figure. The Unemployed figures are based on a number of other factors including how many people are self employed, how many have two jobs, how many people are working abroad and how many non-UK people are working here which the Government simply doesn't know. So it has a survey to find out.

House Prices

Another area where you are going to see samples is in house prices. Here is a nice report from the BBC showing you how much houses cost according to Nationwide and Halifax. It uses the preferred method of year on year change because house prices are seen to be cyclical. You can even look at the Guardian's year on year changes for those two and the land registry.

Yet we still have reports saying:

House prices fell by 1.2% in August, according to the latest Halifax index, ending three months of price rises.

So why am I including these figures in a post about samples? Well it turns out the figures here are samples as well. The Halifax and Nationwide only produce figures on houses where their mortgages are used. Even the Land registry version is slightly limited because it is based on matching up data from different years for the sale of the same house. Really this data should be reported with error bars.

Inflation

Inflation is a term that's coined to talk about the increase in the price stuff costs. There are two favoured choices of how to measure it: using the consumer price index and the retail price index. How do they work? Well they both work in more or less the same way, one of them has a list of commonly bought stuff and monitors its price in shops over time. The other one is the same but with houses included as well.

Firstly you can't measure all the stuff that is on sale over time. New stuff comes on sale so frequently that the list would just get larger and larger. Plus some stuff you buy more frequently than others. I buy apples virtually every week, but I buy toilet bleach about once every six months. So the list takes all of these things into consideration and changes on a regular basis.

What that means though is that the data represents the sort of stuff that people buy, but if lots of people get made unemployed then will that make people buy fewer things? Will the list of stuff that people buy be smaller and hence affect the measure of inflation? Suddenly it doesn't look so ridiculous to start including error bars on the inflation values. Not least of all because the head of the bank of England has to explain every month why it is above a certain level.

Samples

Which brings us to an interesting point about samples. They exist in so many different forms of life's data that we need to think about it all the time - especially when looking at whether data is accurate or precise. So if you're planning survey data and you are going to be taking samples you need to think about the following when doing your analysis:

Seasonal changes in the data - this will affect whether you do year on year, month on month, or even three month moving averages over time to see the difference
What bias is introduced into the data because of the sampling - even full data sets sometimes don't include the entire population
What is the sample size - smaller samples will be less precise than larger samples
Has any normalisation been added to the data - whilst this may seem like a good idea, it may mean that one sample is not quite the same as another one decreasing your precision
Is the result of the sample the data you need - your sample may be asking the wrong question giving you the right answer but not to the analysis that you're interested in
Correlation does not equal causation - just because your data says that two things happen at the same time it doesn't mean that one caused the other

All of these things are true whether you are looking at official Government data, website survey data or even web analytics data. You should always quote error bars!

Tips for Building your Google Adwords to help Analysis

2011-08-29T21:36:00.000+01:00

In a perfect world either all your prospective customers would know about you or they would be able to find you in a search engines because you’d rank for all those search terms that prospective customers (even those that don’t know it) search for. Of course we don’t live in a perfect world. So in an imperfect world you need to pay the search engines money to artificially promote yourself in Search rankings.

The idea of Google Analytics was always that you could integrate your paid search data to get a better impression of how well your advertising pounds with Google were working. So what I thought I’d do is show you a bit about how it all works. Because that’s what I do!

Creating an adwords account couldn’t be easier. All you need is a credit card and a website. In fact, you don’t even really need a website, you could spend your money driving your traffic to another website.

When you’ve created your account you are ready to create your adwords. Many people jump into this part without any preplanning, but it is important to see how campaigns are structured in Adwords so that you don’t regret some of your early decisions.

So the way that Google groups is adwords is described using the diagram below. You can create many campaigns with many ad groups associated with a campaign . Keywords are then associated to ad groups, of which you can have many. To help with your future optimisation you really want to build this structure as flat as possible so you have the greatest control over keywords, however in a practical world it is difficult to manage large numbers of keywords in a flat structure, so you’ll need to think about how you can sensibly group these keywords together to manage them on a useful manner.

It’s also important to think about what each of these groupings allows you to do, so that you can manage it on a more efficient basis.

Campaigns

At a campaign level you can do the broad level of interesting stuff:

Targetting of location
Choosing types of ad (content network, mobile devices, Google Search, etc)
Set a budget for each day of your campaign
Ad Extensions (eg locations, extra links to the site, product details, phone numbers, etc)
Start and end dates for your campaign
Optimisation for clicks or conversions
Demographic bidding for the content network

Ad Group

At the Ad Group level you can get into a bit more detail about your adverts and start building them:

Create your ad text (headline, description, url displayed
Destination url
Cost per click (if you didn't choose that before)

Keywords

This is the stage where you get the least options of things you can do. You're essentially just choosing the keywords you want to bid on!

Keywords bid for
Broad or exact match
Exclusions from the list
Cost Per Click (if you didn't choose that before!)

Of course putting it like this it can be fairly obvious how you'd want to set up your campaigns! If you're in a need of creating lots of geo targeted adverts, then you're going to need to split them between campaigns (for example). If you're creating adverts with lots of different destination urls, then you will split them between Ad Groups within a campaign. If you're going to come up with a list of keywords that you want to specify exact match or exclusions, then you're going to do it at the keyword level.

Goals and Events

The next thing that you are going to have to think about with your paid search is what you want your users to do. This is critical - you don't want to just think getting people to the site. To make it worthwhile you need to look at the return on investment. So you want to be looking at goals and events. What is it that you want your users to do when they get to the site.

This would be a useful point to mention the Google Adwords metrics. They look a bit like this:

Clicks - how many different clicks you have had on your adverts (note that they may not have got to your site, so they may not match your visits)
Impressions - this is a measure of how many times your advert was shown to users
Click through rate (CTR) - this is clicks divided by impressions (ie how many of those people who saw the advert clicked on it). It is a great measure of whether your ad wording is catchy
Average Cost per click (CPC) - this shows you how much it has cost you per click
Cost - rather self explanatory, this is the total cost of all the clicks
Average Postion - this shows you the average position that your advert has appeared on the page

As with all averages, you need to be slightly wary. If your ad appears nine times in position ten and once in position one then you are going to have an average position of nine and you'll probably have a click through rate of 10% (CTR = 10%, average position = 9). Whereas you could also legitimately be in position eight one hundred times in a row and only get one click (CTR = 1%, average position = 8)

Creating Reports

So you noticed earlier that you can put your destination url in your ad group. This means that you can use your campaign tracking at this level of the process. This is useful because it allows you an extra level of reporting that you may not have had at a higher up level.

What would be useful is if you came up with a way of naming your ad groups in a way that easily allows you to group them at the campaign level (whether you do that via SAINT or with Google Analytic's built in functionality or however your tool does it). This means you can create some very handy top level reports of visits and conversions. From this point you can choose which campaigns, ad groups and eventually search terms that you want to concentrate on for conversions!

Of course the beauty of analytics is that you can find out what the keywords people typed into Google! The url that you come from in Google contains a useful little query string with the keywords in it! Of course your analytics system doesn't know if you clicked on a paid advert or a normal link - this is why it is vital that you tag your campaigns in a way that allows the tool to know (usually with a campaign code on the landing page url).

Having this level of detail is brilliant if you want to work out if an individual keyword is driving enough revenue to be worthwhile spending it on getting there in the first place! If it isn't, then you can put it on an exclude list so that you won't bid on it. And if it is, then you can try and work out how to get more visits and more revenue by pushing your position upwards with a higher cost per click!

Really, you've got no excuse not to go and give it a go now.

The ICO has its Data Protection Priorities Wrong

2011-08-14T20:26:00.000+01:00

It's very difficult writing a 'let's get things in perspective' blog post, when half your country has been rioting, hence I've waited a bit before putting this together. I've been quite vociferous on the cookie legislation that has been introduced here in the UK and I've been keeping myself up to date on the other blogs on the subject. Whilst I agree with many of their sentiments, I thought maybe it would be useful to bring in a comparison with wider legislation around privacy.

For a bit of history - you can read a description of what cookies are and why they affect your privacy and other ways that you can track users. You can read about the reaction in the popular press to the new laws and you can read about my proposed solution.

If you want to see the impact of asking users to opt in to cookies, then the ICO has quite a good case study themselves. The ICO introduced an option to opt in to cookies at the top of their pages.Vicky Brock asked them to tell us about their visits using their analytics tool through an FOI request (the freedom of information act is a useful British law that allows you to ask the Government for some of their data for a modest fee - as long as it is anonymous). The result was stark - they had a 90% drop in traffic. Clearly this isn't representative of their user base any more - it'll be so biased that it will be unusable, especially for long tail keyword analysis.

Source: Vicky's Flickr

Following on from this, I've been reading some articles by other bloggers. Notably Thomas Baekdal who wrote a post entitled 'Privacy Smivacy' in which he states that Tracking is not a privacy issue, because:

Your privacy is being violated when other people decide to share what they know about you. It is not a privacy issue that people know something about you to begin with.

I think this is a fair point. And he highlights it by using a couple of examples in a follow up post. Although I think his examples are a little bit limited because whilst there should be no reason to worry about someone noticing what I do when I'm on their lawn, a better example might be having a party on someone else's lawn. You know the person, but you don't know who they have invited to the party and they're all watching what you do. Not only do you not know who they are, you don't know when they are next going to turn at a party on someone else's lawn.

However, this would be fine if it wasn't for very loud accusations of wrong doing by some organisations. Kissmetrics have been in the press because they were using 'etags' as the unique identifier for users. Kissmetrics of course were very vociferous in their response to this claiming that whilst they collected this information, they didn't share it across customers or share it with third parties. In fact there was nothing that you could do with the information. Spotify and Hulu, two of the biggest clients subsequently stopped using KissMetrics. It brings an interesting case for the 'analytics code of ethics' and I'm not sure that Kissmetrics passed this test. Users should always have the option of opting out if they don't want to participate.

But Thomas has a very good point - your privacy is being violated when other people decide to share what they know about you, so lets look at some of those situations when this has happened and the general reaction.

Council contractors lose Government data, including bank account details

In case you're not familiar with this case - a council contractor copied a load of data on to an unencrypted USB stick which was subsequently 'lost' in a pub. This is clearly against ICO's privacy guidelines - any personal information that is transported should always be encrypted. It's part of the ICO's guidelines.

What is the end result of this case? Well the ICO haven't fined anyone because they think that no harm was done as the USB was returned.

Sony Loses Customer Data from Playstation Network

Earlier this year Sony closed down their much used playstation network for several weeks. The rumour was that a beta version of the network 'accidentally' allowed subscribers to have access to all 77 million users details in an unecrypted format. 77 million users had their personal data breached and copied, including bank account details. Users were requested to change their passwords for logging into the network, but also told to check with their banks to ensure that they hadn't been charged.

What is the result of this case? Well it's still being investigated, but it would appear to be an open and shut case. Sony have even admitted that it was their fault. Will they get the maximum £500k fine?

Lush lose Customer Data

Lush, over a five month period, managed to lose 5000 customer's bank account details. There was some code that was added to the website of the company that allowed a third party to access their bank account details as users bought stuff on their site.

What was the result of the case? Well Lush have said that they'll add extra security to their site. The ICO have said that this is sufficient (obviously the 5000 who had their bank account details stolen will agree).

ICO only issues 6 fines

In fact, rather than looking at all the times that the ICO hasn't issued fines, it is easier to look at the situations where it has. In fact it has only issued six fines so far, since it has been allowed to do so. Four have been to councils and one of them was to an organisation that no longer exists.

So what we have here is a number of companies that have lost customer data and they haven't been fined. What we are going to have when they start instigating the cookie ban is an organisation that is going to fine people who collect information which isn't personally identifiable.

The ICO should be given more power to prosecute those organisations (however big they may be) through larger fines or prison sentences for those responsible through corporate liability if they do not store personal data in a secure way. They should not be looking at handing out fines to those who collect data which is not personal and which will never be shared with third parties. The priorities here are wrong.

4 Tips to get more out of SiteCatalyst Marketing Channels

2011-07-28T21:51:00.000+01:00

In the past couple of weeks I've had a couple of clients ask me about the Marketing channels report in SiteCatalyst. I was part of the beta programme in my last job (do I get in trouble from Omniture for saying that?) so I've had access to it for a lot longer than everyone else and assumed it was common knowledge about how it all worked and what you could use it for. In fact, I even wrote about the report when it first came out.

So lets get the basics out of the way. Ben Gaines wrote a very good piece over at the Omniture blog on how to set it up. I've been using a bit of a different analogy of how it works because I don't think the waterfall one makes an awful lot of sense. What I've been comparing it to is a coin sorter. You put the coins in one end and the thing sorts the coins into different types based on the size of the coin.

As the coins fall through, you get small ones being taken out first, then progressively up to the larger ones. If it was the other way around all of the coins would fall into the holes for the larger coins. The Marketing Channels works like that too. You choose (via selection rules) what traffic/conversions you want to take out first and then the second rule only applies to what is left.

Then once you've chosen what channels you want, you go to the next level of channel detail. This is also worth thinking about when setting up, because you'll want to go to that level of detail if you actually want to use the data to optimise.

Tip 1: What First Touch and Last Touch Mean

Technically First touch is the way that the user got to your site the first time they came to the site and last touch is the way that they lost got to the site.

With Last touch you should note that it doesn't necessarily mean that they did this at the beginning of their visit. If you set up one of your channels to be responding to an internal campaign, then that will overwrite the previous last touch. Of course in your settings you can choose so that it doesn't overwrite the most recent marketing channel, but I can't think of a situation where you would only want the first touch channel recorded.

This is annoying in situations where you use other sites payment systems (eg Paypal) and the user leaves the site before coming back at the end for the thank you page and conversions. Obviously you don't want them to count in your channel and whilst you could make it so that they only count for the first touch or your touches only count at the beginning of the visit, really they're mucking up all of your reports and you want to exclude them in the internal url filters.

Also note for practical purposes direct visits and internal visits will only count if it is the first page of the visit by default (otherwise they may end up getting counted lots of times where they shouldn't). This is at odds with how the others are defaulted, so you may want the others to only count for the start of the visit (which is a rule you can add in at creation).

Interestingly First touch is a bit misleading. It is the first touch as long as they haven't been inactive longer than the visitor engagement expiration. This setting allows you to say how long you want your user to be inactive for before you start counting them as a 'new' visitor for the purposes of this report. Note they'll only be new in this report and it is defaulted to 30 days of inactivity. To combat this issue there is a new metric called 'New Engagements' which shows you when a visit is coming to the site with a new first touch.

Confused? Here is a handy guide:

Tip 2: Visits may be misleading

It's not often that I say that you shouldn't be looking at visits in these reports, but bear with me on this one.

So the example that I gave above is a good one. If you're looking at your whole time period for the five visits above, you have five visits. One for each of the last touches of A, B, C, D, E and two first touch visits for A and two first touch visits for C.

And if you converted on day 35 then you have one conversion. Last touch allocation E, first touch allocation C. With me so far?

What's the conversion rate?

Well the conversion rate for last touch Channel E is 100% (1 visit, 1 conversion). For all the others it is 0%, even though they contributed, they weren't the last touch, so you wouldn't expect them to.
First touch channel conversion rate? Well that is 33%. Why? Because you had three visits with a first touch allocation of C, but only one conversion.

But wait, all is not lost. Just because you can't do conversion rates like you want to for your first touch, doesn't mean that you should ignore the data. Remember the new engagements metric? You can use those to work out your conversion rates for your first touch. The downside is that your new engagements are disconnected from your conversions over time.

Tip 3: Break down your last touch by your first touch

I'm sorry if I'm teaching you to suck eggs here (what a strange expression!), but this is the point of the whole Marketing Channels!

It seems to be quite obvious to me that lots of people who buy stuff do research first. They'll search around using Google, find the one they liked and then go there at a later date by typing the url of the site in. Or at least it feels like it should work like that. Does it? Well you can find out by breaking down those conversions who came directly to the site as their last touch by the first touch channel they had.

So what do you do with this information? Well you accept for a start that a load of people have deleted cookies, swapped computer or it is their first visit, so their first touch is the same visit as their last touch.

Then you go off and try and use this information to help you optimise your website. Change stuff and then monitor it over time. Does your natural search frequently have a first touch of paid search for conversions? Try altering the landing page for your paid search to encourage people to find you through appropriate keywords when they come back. Does it happen the other way around? Update your marketing messages in your paid search to reflect the methods that they came through in their first touch.

Do all that stuff that you do to try and make more money, then use this data to show that you've made more money because of the changes that you've made!

Tip 4: Break down your last touch by your campaign reports

A couple of weeks ago I was harking on about the plugins that the clever guys at Adversitement head office have created and a few weeks before that I was talking about the various different ways to set up your campaigns for first, last and linear campaigns.

What would be useful is if you could break down your last touch or first touch marketing by one of these guys. What? Marketing Channels are on full sub relations? You can break down one by another?

I like this particularly when you break down your Marketing Channels by the campaigns that they responded to linearly. You'll end up with some at the bottom that look like they have 0, but actually because SiteCatalyst rounds that number is a fraction of a conversion. It's a nice quirk of statistics that makes me feel like something is working.

Use this information to help build your campaigns, your campaign messages and your landing page messages to get better conversions.

Google+ and it's impact on Search Engine Optimisation

2011-07-11T20:39:00.004+01:00

The big news of the last couple of weeks is the launch of Google+. Or should we be calling it G+? Or Google Plus? The fact that nobody quite knows what to call it is just adding to its allure. Well it is for me anyway. So what is going to be the impact of Google+? Well lots of people are talking about its impact on Facebook, but I think we need to look elsewhere.

What is Google Plus?

Google+ is being described as the Facebook killer and the similarities are there for everyone to see. It's a system whereby you add 'friends' and then you share updates, pictures, links, locations and videos about yourself with your 'friends'. It even looks remarkably similar:

All you really need to be able to do is to invite all your facebook friends to Google+ and you are away.

Then when your 'friends' post updates, pictures, links, locations and videos about themselves and their 'friends' you can comment on them to create an interactive experience.

Why is Google Plus over Facebook?

Well Google is telling us that the unique selling point of Google+ are circles, sparks and huddles. Of course none of these are particularly unique to Google.

Google Circles may be the most unique part of Google+. It is a very simple system that is actually in Facebook (lists allow you to filter your wall and you can use those lists to publish updates to groups), but wasn't there at the start. Essentially they are a system that allows you to put your 'friends' into groups that you can choose to share updates, pictures, links, locations and videos. And of course it works vice versa, you can choose what you want to read about other people depending on your mood.

Google Sparks is effectively a way of creating a news feed of updates about particular subjects. It's a way of getting Google News with a bit of personalisation, as far as I can tell.

Google Huddles is a way of creating conversations in a very similar way to facebook chat. As with the new chat you can do it with multiple friends, although I can't quite work out how to do it online and offline.

Of course from this point onwards it is a very short step to be able to get your gmail and Google+ linked together, something that Facebook has not been able to do (despite Microsoft attempting to build integrations - Hotmail and facebook allow you to export and import contacts, but it isn't truly integrated).

The other interesting thing about Google+ is that it is in the navigation bar of every single Google page. You can't avoid it. I'm logged in when I'm searching, when I'm in gmail (if I had one), any time I'm sitting on a Google.com domain in fact. Half the people in the web are going to be using it all the time.

What about Google Plus one?

What I see next for Google+ is the integration with Plus one. Funnily enough the introduction of plus one, Google+ and the Panda updates have coincided quite well.

The Panda update for those of you who are interested was an update to the Google algorithm that pushed websites down the search engine results that were a bit spammy. However more than being spammy, they were sites that would take other sites content and conglomerate it into one big site. They would then create lots of interlinking between pieces of content to help them push up the rankings. The idea was that they didn't produce any new content of value, so they shouldn't be high up the rankings.

So what is Plus One? Well Plus one is a new way for Google to allow you to promote particular domains higher up in your search results.

Because I have plus oned my website, it will now appear higher up in search results than it would do normally. It's a clever way of allowing me to choose what sort of results I want to see in Google.

Of course, everyone else that searches and has my blog appear in search engines will also see my plus one. It's clever like that. Having seen that someone has promoted my website in search results, it might meant that you are more likely to click it because it has been endorsed.

You're going to start seeing it more often in your search results.

Incidentally, if you login to your webmaster tools, you can find out how many people have plus oned you.

Google Plus One and Google+

This is where is going to start getting a bit more exciting. My plus one of my website has my Google+ ID associated with it. I'm now going to start seeing who my friends have plus oned. We're moving away from a personalised experience based on what I want, to a personalised experience based on what my 'friends' like.

This is a major move from Google. It's going to change the way that people do search. We're going to be first looking at what our friends are promoting, then what our peers are promoting, then what other people are promoting.

We're going to start seeing a new wave of Search Engine optimisation.

What should SEOers do about Google+?

So if the things at the top of search engines are going to be what individuals want to see, we're definitely going to be in a new movement. It wasn't that long ago that websites started making you 'like' them on facebook to let you see them. I think initially this is going to be what websites will try with Google Plus One. I don't think it will work though. It is too easy to take your plus one back again and won't give you the commercial gain. Plus we've seen no evidence that the website will get any user information to market to in the future with a plus one as opposed to a like.

So how do you get the plus ones? Inevitably it will be in the usual ways of making your content relevant, interesting and informative. The best articles will be promoted in the same way they are now, but the results will be less generic.

The number one thing that SEOers are going to be insisting on is a Plus one link on every article though. Now how do I get one on this blog...

PS - I'm on Google+ if you want to contact me. Once I've worked out how to get a page for my blog on there as well, I'll update this post.

Six Omniture SiteCatalyst Plugins for better data

2011-06-29T22:49:00.000+01:00

We've been doing lots of work recently with clients recently where we have installed plugins. So I thought "Hey, why don't I write a blog post about plugins." So I am. This blog post is going to tell you what they are, what you need to do to get them, some popular ones and the sort of data that they'll give you at the end. The most important thing, of course is always what you do with that data in the end, so I'll give a couple of examples of those as well.

What is a plugin?

There are two parts of any SiteCatalyst implementation. One part is putting a bit of HTML on your page that describes a series of variables that you wish to set: your custom traffic variables (props); your custom conversion variables (eVars); Your custom Products; and your custom events. Campaigns, PageNames, etc are just standard versions of these four things. You can do the entire lot by setting them in the HTML on the page.

The second part is that you need a bit of JavaScript. When the JavaScript runs it picks up all those variables that you set in the page code and sends them off to Omniture's servers. Inevitably you might decide that you want to use the JavaScript file to set some of those variables so that you don't have to code them on to every page. Some of these things will happen automatically (like your browser version, for example), others you will need to set in the code.

The plugins for SiteCatalyst get put into this JavaScript file. It's a way of setting variables without having to code them on the page. The advantage of putting them in the JavaScript file is that you can do some clever stuff by setting values in the cookie and then recalling it or grouping variables together.

If you read through Adam Greco's post on the Omniture website from a couple of years ago (on my birthday no less!), then you'll be able to see all the different types.

I suggest, as does Adam, that you do two things when you think about putting one of these plugins into your JavaScript:

Get someone who knows what they are doing to help you (talk to Omniture, hire a consultant, etc)
Test thoroughly. I've also discovered recently that even this doesn't always work so make sure that your release schedule will allow you to make fixes easily and quickly if something goes wrong

There is plenty of help in the knowledge base, but I would still suggest the two steps above as there are probably pitfalls that you will come across that a consultant or Omniture would know about that will save you time and money.

Previous Page Plugin

One of the best plugins that you can install is the previous page plugin. It does what it says on the tin really - it sets the value of the current page on the cookie and then when the user loads a new page it loads the previous page in the variable of your choice.

What do you do with this information? Well if you have your previous page in prop1 and your internal search variable in prop2 you can set up a correlation between the two of them and get some very valuable information. You can find out what people searched for on a particular page. When you've found out what people are searching for on a particular page you can make it so that it is easier for them to find that thing by creating links to those things. Better user journeys should mean more sales. But of course you can measure that using an eVar and your events.

Previous Search Term Plugin

Of course the next step in this situation is to create a previous search term report. This report showing the search terms that people used on the previous page, if correlated with the current page, will tell you the sort of thing that people are clicking on when doing a particular search.

With this information you can start altering your search results to give your users better information when they do their searches. This in parallel with an eVar report telling you the conversions that were made for searches is particularly useful to improve your internal search.

Campaign Analysis Plugin

Talking of events and eVars that measure your conversions, there are a few plugins that they've created over at head office can create reports that will give you a bit more data than before. What they've done is created a couple more variables by tracking your campaigns over a longer time period. I talked a couple of months back about tracking your campaigns using a first, last and linear conversion method. Measuring like this will help you work out which campaigns are helping with your conversions.

Of course using a plugin you can write all of those campaigns into your cookie so that when your user comes to convert you have a chronological list of the order they went through them. If you were really clever you could put the dates in as well, but I think that you'd end up with a bit too much information. What you'll end up with in this situation is a nice way of working out how your users interact with campaigns. Knowing this information will help your Marketing guys build their campaign messages based on the stuff that users have responded to in the past. This will help you improve your conversion rates and make more money.

Of course you can do a number of other things with this plugin, including counting the number of campaigns that the person responds to. Remember that if your campaigns report is set to Full Subrelations (as all will be in version 15) then you'll be able to create a report showing for a particular campaign that converted how many other campaigns the users responded to.

Conversion Analysis Plugin

Another part of the conversion analysis plugin actually makes use of two different parts of the plugins. Firstly you use the ability to pick up existing variables to create your own 'marketing channels' report. It sounds a bit odd doing this in the code, but the next step makes it all worthwhile. Every time a user comes to the site you write to the cookie what channel they responded to, so that when the user converts you can write events to each of the other channels showing that they 'assisted' the sale.

This report is invaluable in showing all channels that helped with a sale, not just the first and last as your Marketing Channels does. It may enable you to spend your money in the areas that help with the conversions and not just those that initiate or end it.

Form Analysis plugin

A popular plugin in helping to build a better user journey to get users to convert is the form analysis plugin. This is a helpful plugin that says for a particular page did the user manage to complete it correctly, complete it incorrectly or abandon halfway through. These values are set by using the javascript to 'monitor' what the user is doing. Of course, for the privacy people out there, it will only send signals back to the server in the case of those three events.

So you end up with a report that has the name of your form followed by the field that it was abandoned or incorrectly filled out at along with the number of correct completions. Using this information you can help build better forms for improved conversion. An improved conversion rate for a form will improve conversion for all your campaigns.

Percentage Page Viewed Plugin

As a nice one to finish on, this is something I learned about at 2010's London Omniture summit. This plugin shows the amount of a page that a user has viewed. It allows you to work out if people are getting all the way down to your call to action or reading all of your content. Of course it can only be set on the page afterwards, so you'll need to run it in conjunction with the previous page report. There is a nice post from Adam on this from last year.

Your designers will be the most grateful people in the world if you can give them this report.

Have I missed any that you particularly like out?

5 Tricks for better SiteCatalyst Dashboards

2011-06-14T22:06:00.000+01:00

About a year ago I wrote a post on the new Omniture SiteCatalyst Dashboards. This time around I'm going to give you a few hints and tricks of things that you can do to make it all work a bit better. I have been expecting to write about SiteCatalyst version 15 at some point, but there haven't been enough people who have moved to it yet, so it is going to have to wait. You don't want to read a post with examples from a dummy account, do you? In case you were wondering, the dashboards in SiteCatalyst 15 are the same as in version 14. Obviously with the added improvements of things you can do in the normal reports. So lets get going.

1. Put your company logo on the reports

Ok this isn't going to make your dashboards more useful in any way possible, but it does make them look more professional.

You don't have to have an Adversitement logo

The more professional your reports the more likely the management are to listen to the recommendations that you make because of them!

Want to go a step further? Why not put a front page on your reports with your name on it so that people know to contact you in case they get stuck. That is easy too because you can add text and images to the front of reports!

You can add in images, text, html pages, RSS feeds and XML feeds so you can customise your pages how you want. If you're going to have a front page to describe your dashboard, you might want to include stuff that will work when it gets published out (which the HTML won't). You can also upload images for free to Flickr to link to for these reports.

2. Add Gauges

Ok this isn't really adding some specific data to your report either, but it does allow you to present it in a way that will mean something to your management. It is adding that vital ingredient of context.

So what do you put in as your maximum and minimum values? Well you have a couple of options:

Put in your average monthly targets as a top target and 90% as your bottom target (or put in your target as your bottom one and 110% as your top one)
If you don't have targets, why not put in your highest and lowest values from the last 12 months - this gives nice context as to whether it is within a range or doing better than expected

These targets are for your standard metrics and your custom events, but they are also for your calculated metrics as well. If you want to show how your conversion rate is doing then this is a very good way of showing it.

The downside of these reports are that they are only for metrics - you can't have a split down for the values of one of your variables (eg you can't have conversion from a particular referring domain).

3. Have two graphs

When you set up your reports you quite frequently want to have your graph of two different metrics, but they aren't in the same scale. For example how often have you wanted to have your entries for a page and bounce rate on the same report, but because bounce rates are so much lower than entries the thing looks ridiculous and you can't see the difference of your bounce rates per page.

The annoying thing is, if you create two graphs on the same report then you end up with an annoying situation in your dashboard of only one graph (unless they are pie charts - but that isn't appropriate in all situations).

So your solution is a bit of an annoying way around of creating the same report twice in your dashboard. Do it how you would do normally for just one metric. Then when you come to add your second metric, you need to leave the first metric in your report, but put it at the bottom. Then move your first metric to your right hand graph, but make sure that you are still sorted by the first metric (like so):

You might have to alter your report again to be sorted by entries once it has loaded (why does it do that? Why bother asking if it then doesn't pay attention to it?).

Now when you add to your report they'll be in the same order as before. This is important because you don't want your top pages and your top bounce rates - they won't match up.

4. Copy your dashboard to another report suite

So you've got two websites being tracked through Omniture. The websites come out of the same template of your CMS, so you've done identical SiteCatlayst implementations on them. That's great, it makes your life a lot easier. It makes it a lot easier if you've created one report that you want to replicate in another report suite.

So it is as simple as clicking on 'More Actions' and then 'Copy Dashboard'. When in this menu you should choose to change the report suite and you'll be given a list of them all!

As a little aside to this one - you can quite easily change your reporting cycle at this point as well. If your reports are set up to be monthly reports then you can copy your dashboard to be identical and then change your date rang. Remember when choosing the date range that if you choose 'last week' or 'last month' from the drop down then the reports will roll over at the end of each month/week. If you choose a particular date range using the free form boxes or by clicking on a date then your date ranges will roll over by a day every day. You probably don't want this.

5. Use Classifications to get better graphs

Classifications are useful ways of transferring one variable that has no meaning to the average person to another with a sensible name. I've written about how to do SAINT before - it is really, really simple.

But the other way you can use SAINT is to help you create graphs for your dashboards that would normally have been unavailable. For example one client had split out the different types of a conversion in an eVar so that they could easily report on each type. It turns out that the ten different values related to two or three different types. We classified these into the three different types so that we could turn them into a graph.

The other option you can use for SAINT is the rather useful ability in Excel of turning text into columns. This is especially useful if you have set up, for example, your page names to be 'Site Section : Page Title'. You want to have entries and bounce rate of your site sections, so you can export the list into excel, use text to columns with your colon as your delimeter.

Another clever way of using the classification is to create a line in a graph that is 'everything else'. An annoying state of SiteCatalyst is that it doesn't have this option. But you can create it using SAINT. Remember that you can only have five values in your graph, so if you are going to have an 'everything else' you need to classify into just four others.

So there you have it. Have I missed out the golden egg that you've been using all this time?

An open letter to the ICO: A solution to the new EU cookie legislation

2011-05-31T22:51:00.000+01:00

It was a law that no UK business wanted, the Government didn't want to implement, the media didn't understand and the people knew little about. So how did a near-on technically impossible law get implemented 5 days ago on the 26th May (albeit with a year to implement)?

Just in case you were reading the fourth link I put up there and couldn't be bothered to look at the whole pdf, let me enlighten you:

Respondents recognised that they had limited knowledge and understanding of internet cookies: only 13% of respondents indicated that they fully understand how internet cookies work and 45% indicated that they had some understanding of them.

Testing of respondents’ knowledge of internet cookies confirmed their limited understanding: Only for one out of sixteen internet cookies related statements a majority of respondents knew the correct answer with other respondents either selecting the incorrect answeror indicating that they did not know the answer.

Go to page 24 of that document. There are 16 questions that were asked of over a thousand random people, with their responses. A third of them thought that cookies "let people send me spam".

With this in mind, how are we ever going to come up with a system that works? The Government wants to implement something to help users, but the users don't understand in the first place. Not only that, there is no common solution - every single company that creates solutoins was going to come up with its own.

Fortunately there is a helping hand out there. Here is an option that is easily implemented across every single website. It is guidelines that they could easily hand out to the general public so that they'll understand what is going on. Users will be able to make informed decisions about whether they want to use a website based on the cookies that they are giving out.

Note that I think that websites should do housekeeping and should only have cookies for services that they still use.

So here are the rules. Each icon should be placed at the top of the page (I don't believe that the average website can't find somewhere at the top of the page next to their logo). I think they only need to provide the icon with the most cookies. Each icon has appropriate alt text and should be linkable to a page that tells users how they use the information from the cookies and how to opt out.

This site uses cookies to remember your preferences and to perform vital site functions such as remembering your shopping basket through your visit.

This site uses cookies to anonymously track what you do across the site. This information is used to improve the site and improve the site's Marketing.

This site uses cookies to alter its content based on your past browsing experience on this site.

Third parties who have sent users to this site from their own use cookies to ensure that they get paid for passing on business.

This site has taken content from other websites (eg YouTube) and put it on their own site. Those sites use their own cookies to monitor usage of their content across the web.

Third party advertisers are allowed to put adverts on this site. They use their own cookies and may take your actions on other sites or information that you have entered from other sites to target the advert to you.

Have I missed anything out? I did buy a whole packet of biscuits so that I could make these icons.

More importantly - does this satisfy what the ICO is trying to do:

Educates users on the ways that websites use cookies to monitor what they do
Gives users an easy option to opt out if they want to continue using the service (assuming the icon's link points through to ways of doing this)
Provides a scalable solution so that users aren't asked to do the same thing in hundreds of ways depending on what tacking company the website happens to use
Is easily implemnetable by all websites
Is not intrusive of the users' journey

I think it does. Now all we need to do is get the ICO on board and get them to produce another set of guidelines in a short space of time. Ed Vaizey, if you are reading this, you are more than welcome to steal it and pretend it is your own.

Not the actual buscuits though. They're mine. Nom, nom, nom.

The New Cookie Law: Reaction round up

2011-05-10T19:41:00.003+01:00

So this week the Information Commissioners office published their guidelines on cookies following their ratification into law. The outcome is that they've essentially said that you should get explicit consent from your users for all but the 'essential' cookies. Analytics and advertising cookies they explicitly point out are not essential, so there is some work needed (possibly). Here is a link to the report (pdf warning).

An analytic cookie might not appear to be as intrusive as others that might track a user across multiple sites but you still need consent

Shameless plug: Here are some links that will give you the full details based on stuff I've written:

How Slashdot saw the new Cookie ruling (August 2009)
EU cookie laws verified (July 2010)
What, Why and How of Cookies affecting your privacy (August 2010)
The UK needs to change its cookie policies (November 2010)
Cookies are the best way to track users Or are they? (February 2011)

So you don't have to, I've decided that I'm going to give you a media round up. Aren't I nice:

Google Analytics blog - No comment
Adobe Omniture blog - No comment. EDIT (25th May) - Adobe have put an updated post about this on their blog explaining the ruling with some suggestions on what to do.
Webtrends blog - No comment (Although I'll grant them a blog post from July 2010 - which in conclusion gives "Three Alternatives, All Bad" of which the new law is the second). EDIT (25th May): Webtrends released a short post on this several days ago which I missed (sorry).
Yahoo! Analytics blog - No comment
Core Metrics blog - No comment
IAB blog - No comment. EDIT (26th May) - The IAB have updated us following Ed Vaizey's 'clarification'.

(I'll update this table as and when they do make comments - I'm well aware we haven't really put a full thought into it yet)

So that is fairly conclusive what the big analytics providers think. They think it is your problem. Your website, you sort it out.

Ok, ok, it was only a day ago that this was announced, but its not like hasn't been coming (see the five blog posts I've posted) that they couldn't have at least come up with some thought on the situation. They could have helped come up with some solutions, suggestions, hints. There will be a session to discuss at the Adobe summit in London next week (8:30am in the morning!) to discuss the new laws, so at least it is starting to move (anyone turning up, by the way, can come and say hello to me - I'll be on the Adversitement stand in the main hall for much of the day!).

Well lets look at the mainstream press (who are going to be one of the industries screwed by this as they have to start revealing how many different types of advertisers and get consent - this could seriously affect their revenue streams):

The BBC

Typically of an institution that is state funded - gives the facts with little opinion.

... from 26 May the ICO is obliged to investigate any complaints it gets about the use of non-compliant cookies

and

One solution, brokered by the Internet Advertising Bureau, might be the use of an icon on adverts that, when clicked, reveals information about data being gathered.

caught my eye as useful comments.

The Telegraph

The Telegraph also just appears to deal in fact, in some cases not particularly useful fact either.

Many websites collect user data in the form of ‘cookies’ - small files that store information to help websites recognise regular visitors, sometimes including their name and address.

says the article, although it would probably be worthwhile caveating that statement to say that it only stores information given to it by the user.

This could mean significant changes for many websites, which typically publish a privacy page telling visitors to get in touch if they want to opt out of receiving cookies.

The article goes on. I'd contend that it affects all websites, not just many.

Jim Killock, executive director of the Open Rights Group, said: “The key thing is that people who are being tracked by cookies for purposes such as online advertising absolutely should give their their prior and informed consent. The ICO needs to help to draw a distinction between cookies that enable websites to work and cookies that are there to track people.”

Presumably Mr Killock dislikes using free websites and more to the point dislikes you using free websites.

The Wall Street Journal

The WSJ is no stranger to controversy from the analytics world, so they've taken a far more pragmatic approach:

If the U.K. Information Commissioner thought that publishing guidance on the implementation of the European Union privacy law was going to calm things down, he seriously miscalculated.

It appears that the actual website owners are fighting back too:

“Frankly that’s not good enough,” said Scott Allison, CEO of Teamly. “As a small self-funded startup we just don’t have the time to analyze and interpret the legislation, especially when given just over two weeks’ notice.”

At least it appears that the WSJ gets it (maybe the Telegraph should have attempted to keep hold of Ben Rooney!):

The most controversial area of the guidelines cover so-called “3rd party cookies”, cookies placed on a user’s computer typically by advertisers on a site. Almost all ad-supported sites use third party cookies.

Although I wonder if the author has mentioned this to his bosses who run the WSJ! Overall this was an informative read looking at the issues from both sides, although not really providing solutions. More please!

ZDNet

I think ZDNet see this as the start of something, rather than the end of it:

Businesses cannot yet rely on consent via browser settings, so must find alternative ways of gaining consent for cookies that store information on users' machines, the advice stated.

But they also state that:

Although the European Union set a deadline of 25 May for the implementation of its new web-tracking rules, and the UK transposition of those rules is supposed to take effect the following day, the Department for Culture, Media and Sport (DCMS) said in April that enforcement would be delayed "in the short term" while technical solutions are established for the gaining of consent.

I suspect this is many people's views currently "Unless someone comes up with a good way of doing this without pissing off the user, the Government won't do anything about it."

The Guardian

From the rather intriguingly titled 'Guardian Government Computing' there was very little information.

"But we want to spread the net as wide as we can and would welcome further comments from others who have practical examples to share. This advice is very much a work in progress and doesn't yet provide all of the answers."

Says Information Commissioner Christopher Graham. It also states that:

Advice for the general public on what the new law will mean for them is currently being drafted by the ICO.

Which is interesting, because it suggests that different information is being given to businesses than to the general public. Why is this? How is it going to be different? Are they going to tell you that cookies are bad and that you should block them? I'm considerably more worried about this.

The FT

It amused me that the first time I tried to view this article I couldn't (and you may not be able to - Google the title and you should be able to get in) because you have to have signed up to their website:

“The more privacy intrusive your activity, the more priority you will need to give to getting meaningful consent,” the ICO said. Only functional cookies – for example used to transfer a selected purchase to an online shopping basket – will be exempt.

And

The ICO has powers to fine companies up to £500,000 for breaching the rules, although Mr Vaizey has indicated that companies will be shown a short period of leniency after May 25 as they adjust to the new regime.

Fining people £500k seems a bit excessive. How much will I get fined for not having anything on my blog telling you that you are being tracked?

The Inquirer

I wasn't going to include this, but their first sentence made me laugh a bit:

SHOPPER STALKING ONLINE FIRMS in the UK have been given guidance on how to use tracking cookies.

That's told me.

The Out-Law

You'll have noted that I've frequently linked to the Out-Law in the past around this issue and I'm bound to continue in the future.

"We would advise anyone whose website allows or uses third party cookies to make sure that they are doing everything they can to get the right information to users and that they are allowing users to make informed choices about what is stored on their device," the ICO said.

Is probably the most sensible advice around

"The guidance leaves it up to organisations to decide how to get users’ permission for cookie usage, which means different companies will use different methods. Only once enforcement action starts will we really know which of these methods the ICO thinks are within the law and which are not. The guidance does list possible methods though, which will help companies, and may be updated when browser-based technical ways of giving permission emerge," [Claire] McCracken [OUT-LAW lawyer at Pinsent Masons, the law firm behind OUT-LAW]said.

Interesting that a law firm would make these statements. It seems a bit pointless (to me) to have a law that is so ambiguous that we won't know if every business is complying or not until one of them is fined and the rest have to follow suit

SCL Blog

My friend Bob Mitchell has also posted on his blog about this. Bob is also the country manager for the UK for the Web Analytics Association (although they are yet to announce their stance):

Based on this I think that our own use of cookies on www.scl.com needs to be eliminated and other, passive, methods used until visitors hit a site feature where they’re fairly committed already. Furthermore I can see no effective way to make Google Analytics compliant right now.

Which is probably going a lot further than anyone else has suggested so far.

David Naylor

David has amused me in the past with his take on this and a page with multiple pop ups:

The new European Cookie Law comes into effect towards the end of this month, I’ve read it is the 25th May (26th May according to the BBC) and I still don’t know anyone who:

A) Is prepared for the change
B) Knows how they will technically deal with the problem.

Which probably sums it up very well.

Sorry if I've missed anyone out, but I think this has gone on long enough. Post your links in the comments if you find anything else!

Edit: 11th May 16:44 GMT:

Wired

A guest post from Milo Yiannopoulos, the Telegraph's Technology editor, this is a little bit of a rant. But you can see why, because he is coming at it from a different angle to the rest of us:

Enforcing this legislation -- if it can even be enforced, so unworkable and technologically illiterate is the basis on which it has been conceived -- will have disastrous effects on innovation and entrepreneurship throughout Europe and it will dramatically affect the livelihoods of publishers whose incomes come from advertising.

I can see his point here. I can see many, many companies completely dissappearing as this destroys their business model. Alongside any business that runs advertising on its site, any business that runs an affiliate network (eg Money Supermarket and Confused). Milo also looks at the effect of foreign websites on the UK:

It's not beyond the realms of possibility that the Wall Street Journal or New York Times will decide it's simply not worth serving pages to the UK when it's impossible to monetise them and the user experience is so poor. We should also expect British advertising technology firms -- one of the hottest sectors in British tech -- to decamp to the US, where the law is less restrictive.

So poor for the users and poor for British Tech. This appears to be a solution that doesn't suit anyone.

It has been fuelled by campaigners and journalists who are apparently ignorant of both the mechanics and the economics of publishing and advertising.

Sadly this is true. My former colleague Ciaran Norris put it quite well:

... it’s still annoying that Google & Facebook target ads and emails at me, right? Well, no. Because that’s the cost of us receiving free content...

Social Media Integration with Web Analytics: How, Why, What

2011-04-26T19:39:00.001+01:00

We've been talking about social media in the office quite a lot recently. And why not, it is not just starting to take off in the sense that people are using it, but that Marketers are wondering how they make use of it all. As we have been discussing that this is going to be the start of Marketers seeing social media platforms as being vital for the advertising of their products.

However, as Google Analytics built a bridge between site analytics and advertising analytics for those doing Search Marketing, there is going to be a growing trend of Analytics bridging the gap between site analytics and social media analytics. So here I am going to explain why it is a good idea, how you do it (roughly) and then what you need to do at the end of it all. But before we can do that, we are going to have a brief history lesson:

Old School Marketing

Back in the day (and in fact, still today) offline Marketing has been a big mystery. William Lever once said "I know half of my Marketing isn't working, I just don't know which half." This is kind of true for most offline Marketing measuring.

What essentially happens is that there is some research done by asking people what they think (about the product, advertising, etc). When people talk about research, effectively what they are trying to do is work out what people are talking about in the real world relating to your product/brand/subject.

This will help inform the advertising type, message, timing, channel, etc to make sure that it is pushed towards the correct demographic. This stuff is sophisticated and is the reason that you'll get adverts for cars and computer games during the Champions league semi final tomorrow, but you won't see them during the Pop Idol final.

Someone with a degree in Statistics will then analyse the results of the campaign (whether they are sales results, brand awareness ratings, etc) to look for minor fluctuations caused by the advertising. They'll then be able to give you an ROI for the campaign. This will lead you to try and come up with ways of optimising it, leading you to go back to the research stage again.

Don't let anyone in the online industry pretend that in the offline world this doesn't happen and that the offline industry is awash with people throwing money away. I sat through some excruciating meetings whilst at an insurance company listening to the offline people claim responsibility for increases in sales from the website caused by the television adverts - proven by what appeared to be minute changes in weekly sales cycles.

The emergence of online Marketing

In the new world of online Marketing it is of course much easier to do your analysing. Instead of having a statistics degree, you can now track your campaigns in a much more orderly fashion using Analytics tools. Google Analytics was one of the first who linked all of their data from their advertising directly to the data from the people using the website, but others were using campaign codes in the meantime.

So in the new world you could do so much more in one system. You can directly measure the impact of the advertising into minute detail using your analytics. No longer is it true that you don't know which half of your advertising is wasted, because you can see which half.

This doesn't remove from the fact that you still need to do research. Largely this research tends to be completely disassociated from your site analytics (however you can do voice of the customer integrations in SiteCatalyst). You still need to find out what they are talking about to help you build your campaigns in Google; where to put your banner adverts; who to build affiliate deals with, etc. It's a step that is a bit too frequently left out online.

An Online Social Web

In the social world, those people doing the talking about your product/brand are now doing it online. They're talking about you on your Facebook page, on other people's facebook pages, on your Twitter, on other people's Twitter, on your social media platform of choice by you and by others too.

At the moment there tends to be a giant wall between your research of what people are saying about you on Facebook (plus what their demographics are: age, sex, occupation, followers, friends, etc). Information is coming out of those systems, but you haven't done that clever bit of linking it into your analytics. You have a load more information about your users but you can't link them together.

It's annoying, because that information is online. Your website is online. Why can't you tell if they are the same people?

Breaking the Social Media Wall

What we need to do is break down that wall and start linking the data sources. However the word of caution should start here - it is really difficult to link two data sources together and not end up with loose ends. Two sets of data sources never match up in the online world. It's the reasons that your clickthroughs and your visits for a paid search keyword may not match in Google Analytics. It's the reason that ComScore claims you have half the visits that you think you did.

There are a few of ways you can do this:

You can build campaign codes into the links that you deliberately put into the social world yourself. This means you should get more accurate information on the clicks into your site from those sources. This is by far the basic level that you should definitely do.
One of the advantages of the Facebooks, Twitters, YouTubes of this world is that they build really nice APIs that allow you to transfer data from their system into yours. Great hey? Using those APIs you can transfer a host of information about what people are doing and talking about. You won't be able to link that directly (think correlation and causation), but you will be able to have the data sources side by side (assuming your analytics tool allows data insertion - if not you might need a third party system!).
With the more bespoke elements of Facebook and YouTube allow you to create your own 'channel' or page. These allow you to put custom javascript tracking code on the pages and will allow you to build your own analytics. Combining this with the results from the API for these tools will give you far more detailed demographic data about your users than you ever dared to imagine.

What do you do with the Social Media Data

So where does that leave us? Well it brings us back to the final stage of everything you should think about every time you do something involving analytics. Is the additional work worth the data in terms of increasing the bottom line?

So your questions should be:

Will knowing demographics about the users and what they are saying about us help our research process to help us build our Marketing? If the answer to this 'no' then you need to stick with number 1 on the list above. You just want to know the best way to put your money through by getting results of campaigns.

If the answer is 'yes' then you need to ask yourself the next question: Are we in a position to be able to advertise to the correct demographics and make use of a full integration? Only if the answer to this is 'yes' should you do the last stage.

Interestingly, just like it was within Google's interests to link Adwords and Analytics, surely it is within Facebook, Twitter, YouTube (owned by Google!) to create a simple integration. Those customers with simple integrations are going to encourage advertisers to spend money on their platform.

Credit for Pictures:
State Library and Archives of Florida
The Library of Virginia

Four ways to make use of the new Google Analytics interface

2011-04-19T20:14:00.000+01:00

This is my first post for a while on this blog, but I have mitigating circumstances - I've just come out of hospital having had some knee surgery. You'll be pleased to hear that this hasn't stopped me making bad puns, ridiculous comments and writing about stuff that has been going on recently. This week it was pointed out to me that there is a new version of Google Analytics that has come out. Dan Baker has a very good overview of the changes. Just to point out before going through the whole post, the change is mainly just a facelift - most of the stuff in Google Analytics is the same, it has just been presented in a different way.

In fact, it is interesting to note that with SiteCatalyst 15 coming out recently (more on that soon in the coming weeks - in the mean time here is a link to the 6 things that I thought should come in SiteCatalyst version 15) that the two tools are starting to converge towards each other. It's almost like they are copying each other's features as Google attempts to make its data presentation better and SiteCatalyst attempts to make its segmentation better. I suppose it was inevitable really - one tool recreating the best features of the other.

So let's have a look at it:

So it effectively is a bit cleaner looking and brings some of the features that they've added to the tool in the last couple of years to the forefront. And it's those features that they want you to use more because it'll give you that little extra in the fight to make more money from your website (or whatever it is you want to do with your website).

As Dan says in his blog post, for the 'power users' the left hand navigation is mainly the same, but it makes it easier to get to the reports because you don't have to go through each topic heading to get to them. For the 'basic users' the groupings should make it easier to understand and they may make more use of the reports.

Advanced Segments

Advanced Segments have been a staple of Google Analytics for a while now. Two and a half years ago I wrote about the new features of Google Analytics, of which advanced segments was one of them. What Google are attempting to do by putting these at the forefront of the new tool is to give the users as much use of them as possible:

By putting them as the main headline link, they are showing that you can do much more with the reports than you originally thought you could. As long as Avinash continues to shout 'segment, segment, segment', this will be given more and more prominence.

Previously the ability to create your custom segments was hidden away at the bottom of the left hand navigation in the middle of the help section. Now it is bold and bright when you put your segments in.

Custom Reports

Two and a half years ago I also wrote about custom reports when talking about the new features. Now these custom reports have been thrown in at the deep end, albeit with a bit of an upgrade to them. Personally I quite like the way that you can now build your reports:

As a 'super user' I also like the way that I can now add these custom reports to any of the different websites that I manage. In this case it is useful if you have created profiles for your website including, excluding or having different versions of the traffic. It makes it much easier to create custom stuff.

The other interesting thing about these reports is that you can add a tab on to the top to create a series of custom reports with lots of different pages. It is a bit like the advanced dashboard creation that you can do in SiteCatalyst.

One of the nice things about the custom reports is that you can apply your segments to them at the touch of a button because that advanced segmentation is still at the top of the screen. The downside is that you can't default this to be on with the report. What you can do though is apply a filter to your custom report. These filters work a bit like your segments, but with a bit less functionality.

Dashboards

The dashboards in Google Analytics have always been a bit rubbish (sorry!). Mainly they are used as a starting point for your reports and then you can move into the report when you want the actual detail when you want more than five lines of detail. The new dashboards are a little different as they are a little bit more customisable.

I hasten to add they're only a little better, but mainly because it is easier to customise them. The limit of your data is still quite high. In all probability, you will use the custom reports more than you'll use the dashboards.

So with the new dashboards you can have pie charts with up to six slices or tables with up to 10 rows in them. You can also add in and sub out whatever you want to make it a bit more useful. You can also create a few more of them and have them as something that you give to senior managers as the quick overviews.

The downside is that there is not yet the ability to export them, they are just screens that people can see. The other disadvantage is that you can't get the advanced segments on them as you would wish them to be. This is a bit of a shame, but hopefully it will be coming in a future release (it is still in beta at the moment - even if most of Google's products never make it out of beta).

Intelligence

Intelligence is something that Google has had for a while and used to be mixed up in the left hand navigation. They have now moved it out of the left hand navigation to the top navigation on its own, because it is more of a functionality than a report type.

I'll point out now that the alerts that these give out tend to be about as much use as Google Alerts. Maybe they are run on the same system. Needless to say mine are usually late, don't really bear much relation to reality, but the idea is really good, so I really want them to be better.

The trouble is the same as SiteCatalyst alerts - essentially you don't get them until it is too late (the beginning of the next day/week/month). So if I have a spike in my traffic I want to know now that I am having a spike in my traffic, not wait until the end of the day. I have a chance to be able to do something about it now, but tomorrow it may be too late. The buzz has gone and I've missed out.

Having said that, there are things that knowing the next day might be of genuine use. For example if you want to know if your adwords have run out of money, then it is very easy to do this on a next day basis. It is also useful if you want to find out when you hit a certain number of conversions so that you can do some capacity planning for your warehouses - set up an alert for 10, 20, 30 and 40 sales in a month, then you'll be able to do something about it.

So there are four quick ways of making a win out of the new Google Analytics interface. What are your favourite features?

Salesforce integration with SiteCatalyst through Genesis: Why, How and What

2011-03-25T19:56:00.001Z

This week we've been looking at how we do a SalesForce integration with SiteCatalyst, so I thought I'd write a post about the benefits and how it works in a no way predictable theme to the posts on this blog. Funnily enough though (in a no way predictable second sentence to this post), I have written about this sort of shenanigans before. In fact it wasn't long ago when I was listing five tips for Lead generation web analysis. It was a little while ago that I was talking about databases and SiteCatalyst, which is kind of essential to this process.

Why do a SalesForce integration with SiteCatalyst?

Did you read those five tips for lead generation? Why not? Go back and read it. I'll wait here.

Ok, well I came up with four tips and a fifth one that I said don't bother with. The fifth one was this one we're talking about.

Ah.

This is embarrassing.

Seriously though, most of the time this probably won't work.

So why would you do this integration? Well with SalesForce and SiteCatalyst it is possible that you can integrate your data in a way that will make it hugely beneficial for you if you can make use of the insight. It will give you the information to be able to work out if your marketing is making you money and will give you real insight as to whether changes to the website are making money otherwise you are just guessing.

Those of you who use Salesforce will be more aware of the benefits. SalesForce is a customer relationship managment tool that allows you to track all the interactions of your leads online and offline (if you set it up in that way). If your revenue doesn't come in a direct fashion from your website then it is a great tool to not just manage each of the interactions of the customer, but also to track how much profit you are making from each customer.

So if you have all your revenue figures in one system and all your traffic in another, it would be great if you could link them up to do that deep analysis. Otherwise you are going to have to either assume an average order value in SiteCatalyst (and if the order value varies wildly, then this can be very misleading) or you'll have to assume an average marketing cost across all your revenues (which could be equally misleading).

How does it work?

Well in those five tips on lead generation numbers three and four were to put more data in your analytics system and to put more data in your sales funnel. It turns out that if you do both of these then you can almost create a unique identifier for your data (called a key). This will allow you to link your two data sources together.

So what you do is you alter your code on your page to collect a couple of new variables which can also be picked up by your sales system. These are effectively unique identifiers that allow you to link together tables that previously didn't have common elements.

Yahoo Pipes shows how you can link up database fields to come up with a single, joined up output

So you create some new fields in Salesforce that allows you to do a bit more with the data from SiteCatalyst - however this is a difficult process in Salesforce. For it to work properly, you really need to collect information from the beginning of the visit. This may mean that you'll end up with more data in Salesforce than you know what to do with.

A better way of doing it is to use the Salesforce API to insert the data into SiteCatalyst. It's a much smaller dataset that you'll be adding, even if it does have to be dynamically updated by Salesforce over time. This is fine though, because SiteCatalyst can allow that to happen.

What about the different data filters?

One of the main difficulties is that the two systems have different filters set up. SiteCatalyst is set up to remove robots, spiders and general things that don't look like people. It will also be excluding a bunch of people who don't want to be tracked because of blocking cookies and/or javascript. You may also filter out your internal people.

In Salesforce you are probably filtering in a slightly different way. You may be excluding users who look fraudulent, those that sign up but never buy, etc. Those set up internally by sales support may still be included.

You are definitely going to end up with some of your salesforce that can't be linked to data in SiteCatalyst. You are also going to have to accept that you are going to end up with SiteCatalyst data that has subsequently been removed from Salesforce. The trick is working out how much of it and what to do about it.

So one of the early steps in the process is definitely data 'cleansing'. You need to go through the two data sets and work out if there is a massive difference. You'll need to look through those that don't match up and try and work out why. This is always easier in Salesforce, but you may have to look at any bespoke SiteCatalyst filtering you have set up.

Data Classification is vital

I think this is one thing that I constantly mention. And it is so easy to do. Classify your data in SiteCatalyst. Why? Because you don't want to have to create too many tables in Salesforce and SiteCatalyst capturing the same data in different ways. Capture it once, classify it into many different data sets.

The other advantage of classification is that you can collect data in a computer language and subsequently turn it into a human language that the user may be able to use.

Use Genesis

Any how could I forget this point? Omniture/Adobe have already done this, so the actual data integration is a 'simple' case of mapping the two data sets within Genesis. Of course you'll need to do a bit of work first to make sure that the integrations can work.

You'll need a Salesforce expert on hand to be able to tell you the names of fields.

You'll need a SiteCatalyst expert to work out the spare variables and events. They'll then be able to advise on the reports that you should look at, the points that you can do correlations between tables, etc.

You'll also need someone who can take action on all this insight so that you can make your company more money.

5 Tips to campaign success

2011-03-09T16:59:00.000Z

Whilst the world finally catches up with the Analytics industry to talk about cookies, I'm going to move off from cookies and go back to one of my educational pieces. That's just how I roll. Today I am going to write about campaign tracking, because I haven't really talked about it before - my posts are always a bit more specific, so this one is going to do some basics of things you should do and how you can do them.

1. Set up your campaign codes

I've written before about how to set up campaigns in Google Analytics and in HBX, so I'm not going to do it again. The long and the short of it is that you append a campaign code on the end of your destination url in the form of a query string. It's then picked up automatically by the tool and plonked into the correct report.

The same is kind of true of SiteCatalyst. I'll rephrase that - the same is true in SiteCatalyst, although it of course requires a little bit of set up (see part 2)! The great thing about SiteCatalyst is that whilst you have a built in campaign variable that you can use (s.campaign), you have the equivalent of 75 campaign variables that you can do anything with (your eVars) that all work in the same way. This means that you can set up your s_code to replicate your Google campaign codes quite easily - matching up one for your campaign, one for your source, one for your medium, etc. You can then use your sub relations to link them all together so that you can break down one by the other.

2. Configure your reports

Of course, one of the benefits of paying money for a tool like SiteCatalyst is the things that you can do in the background.

eVar Admin allows you to change how the eVar works

So you could use some of your 75 eVars for a campaign where:

Conversions occur only against the first campaign result, showing you the original campaign
You could use another one of your 75 eVars to convert against the most recent campaign
You could use one of your eVars to convert against each of the campaigns that the user has come through linearly
You can set your campaign to expire after the visit so that you can work out how the campaigns work in a visit
You could set up another eVar to expire campaigns against a particular conversion event allowing you to dissociate a second sale from the original campaign

All these things can be done using the same campaign code, just replicated in a different custom campaign variables. You need to work out if this data is going to be useful for you as well, but allowing you that ability gives you a bit more power.

Whilst being able to break down your most recent campaign against one with linear breakdown gives you some important information about how users interact with your campaigns before converting, you might also want to know how they perform in order. In which case you need to use an plugin that will allow you to stack the campaigns on top of each other in the cookie and then the value can be recorded from there.

3. Classify your Campaigns

I cannot stress enough that classifying your campaigns it the best thing that you can do. If you're not using SiteCatalyst, you can just as easily do your classifications using excel and pivot tables.

I've done a walk through of SAINT classification on here before and it isn't difficult to set up. There are two hard things about doing a SAINT classification and they are:

Deciding how granular you are going to make your campaign codes in the first place (this will be dependent on what you think you might be able to do with the data)
Maintaining the SAINT upload

The first of these will determine your SAINT classification right from the off. If you make your data very granular you can do much more with your classification - putting the data into more groups to be able to do better analysing. However you don't want your campaigns to be too granular as it may be difficult to maintain and you may never need to see the data at the level (or more importantly act on the data at that level).

Determining the level of granularity that you set your campaigns at the first place will make it easier to work out which classifications you need:

If you break your campaigns down just by the type (email, search, banners, etc) then you are going to have very little classification to do - you may just want to classify them based on who is going to be responsible for each of them, which should be easy to maintain
If you break your campaigns down to the minutest detail (search keyword, email link, banner creative on which site, etc) then you may have to classify the campaigns many times, possibly in a hierarchy which is going to be difficult to maintain going forward

My suggestion is to create a balance based on what you think you (and your colleagues) can use and maintain. Of course that is highly objective!

4. Measure your campaigns

But we've been talking about this all along, haven't we? Well not quite, the number of agencies that will tell you that they will be able to optimise your campaigns so that they give you the best click through rate, the best cost per click, the best rankings, etc and that they will then give you the best return, are missing one thing. The thing is that the CEO wants to see a spreadsheet like this:

They want to know how much it cost and more importantly, how much money you made from it afterwards! The ultimate metric for your campaign is not the number of visits it generated, but the amount of profit it generated. It will allow you to go to your boss and say "Look at this, I'm making you more money than you spend. If you give me more money, I'll show you how much more profit it will make you."

This is why when you set up your SiteCatalyst or your Google Analytics you need to clearly define those goals and make sure that they are tagged correctly. Make sure those events are firing at the right points so that you can work out how much money you are making from your campaigns.

Of course it works better if you've done all the things above - you can analyse your campaigns in conjunction with other campaigns if you have them stacked or set up in many eVars. You can group your campaigns together and work out which ones you need to drill down into if you have them classified. Remember that you are trying to increase profit not just by decreasing cost (what your agencies are doing) but by increasing revenue.

5. Test, test, test

More importantly, your campaign return on investment is all subjective.

Is it good?

Is it bad?

It doesn't really matter, what matters is whether you can make it better. Using your analytics data you should be able to identify those that you can put your time and effort into to make them better. You do this by looking at your micro-conversions, your engagement, your follow up with the user, etc. Then you change it and see if it makes a difference. Measurement is only useful if you take an action on the data that you've collected.

Cookies are the best way to track users. Or are they?

2011-02-28T23:02:00.000Z

It’s not unusual for me to write about cookies. So I am going to continue doing so! Last week we had a web analytics Wednesday event where we looked at data and privacy issues which the poor chap who was presenting (Alan Meneghetti) must have wished he’d never turned up! He was from a company (Clyde and Co) that specialised in International law, however as we’ve seen in the past online, the law and what users want rarely match up. And in this case it looks like the users are revolting to pick their own standards because the Government won’t.

The trouble is, as I’ve talked about on my guest post at eConsultancy, the problem isn’t with the technology that is being used (it’s the opposite to the problem with those that like to torrent!), the problem is with the way that businesses use the technology to exploit the user. Users don’t like it. In the same way that I don’t like someone wandering around the shops behind me telling me that I tried on something similar in a shop earlier and I could go back there and buy it.

Does it matter that there is a clear indication in every website’s privacy policy what they are using cookies for? Who reads privacy policies anyway? Do you read the privacy policy of a website as soon as you visit? When you login to a website do you read its privacy policy?

So without turning this post into another rant in a similar vein to the eConsultancy post (which had a series of recommendations I might add), I thought I should point out how futile the approach of blocking cookies is.

Browser Cookies

Browser cookies are currently the best way of tracking someone’s user behaviour. Or are they? Notoriously I already tell the people when I do my analytics training that you users are a fickle bunch. You:

Delete cookies frequently
Use different browsers (or even computers/devices)
Use the same browser for multiple people
Block cookies

They’re not very accurate. “Get over it.” I tell the people on the training. Unique users in web analytics tools do not equal people. They never have and never will. Treat them as such.

Logging In

But are browser cookies currently the best way of tracking someone’s user behaviour? What happens if I could tie up your visits another way? How could I do that you ask? Well there have been ways that organisations have been doing it for a long time. The most simple of which has been to get you to login and use that information as the unique identifier.

This was the basis for Redeye when they ran their analysis many years ago on a website that users had to login to. They worked out that on average the number of users calculated was about twice the number of people over the space of a month if you used cookies. It was about seven times as high if you used User agent and IP address and this pretty much ended anyone using user agent and IP address as a measure of a person.

Add into this fact that you are probably logged in to Google all the time (and/or Microsoft or Yahoo!). Although they claim they don’t Google could quite easily link up what you were doing on all those websites that they have tracking code on by your Google login ID and then use that to do targeting on you. Microsoft on the now defunct Gatineau used the information in your MSN passport to give those websites with their tags more detailed demographic information.

Whilst you can’t see a big company doing something they shouldn’t really, I am sure there are enough companies out there that would want to exploit the users in the hope that they’ll make a small amount of money out it by having lots of tags on lots of site and allowing you to login once. Especially on an ad network.

User Agent and IP address

In may appear that we are going back in the ages here, but bear with me. What happened in the UK was it became abundantly obvious that some analytics systems were underestimating people on a daily basis because they only counted those that accepted cookies and ran javascript. A common trick in the last five years was for newspapers to bump up their user figures by adding in a noscript tag and/or using and IP address/useragent lookup to count those other people. Never mind that this was deemed inaccurate because of the difficulties in excluding robots and spiders, this was a good way of making it seem like you were doing better than you were.

Flash cookies

Controversially a company recently was caught using Flash cookies to resuscitate the browser cookie and allow them to continue tracking you even if you’d deliberately deleted your browser cookie to stop this happening. Flash cookies are computer dependent rather than browser dependent so they can allow you to monitor users from various different browsers (if they use them).

Flash cookies are far more hidden than browser cookies and so the average user won’t know how to delete them. This makes them far more dangerous, especially considering the proliferation of flash used in online advertising – 95% of the time you won’t know whose website you are connecting to, to get your content.

User Behaviour

What would happen if it turned out that you could identify me by the way that I browsed? That’s a controversial subject that the psychologists have been debating for a long time. Maybe some people hold the mouse in a certain place on the screen, or they automatically run it over links. Some of us read at a similar sort of speed and use similar types of search terms to get to websites. There could be a whole host of unique behaviour that could identify us in a better way than user agent and IP address ever could.

All of these linked together

My new company (Adversitement) has recently just announced a new technology called ActCorrect. What it does is use an algorithm to link together visits using some or more of the above technologies (don’t ask me which ones, I didn’t do it!). We reckon it will allow you to track a visitor on a far more accurate level than ever before.

However as mentioned, this capability isn’t only as good as those who are using the data. We’ve been quite specific that we won’t use this technology to identify individuals and target them. The point of this technology is to get a better insight of the aggregated use of the website over time and a better handle on how well campaigns work and how the website is used. This information shouldn't be used to identify individuals.

However it will only be a matter of time before a company comes up with something similar and then uses the technology for targeted, personal and real time advertising that users are going to see as an invasion of privacy. To stop this there is no point continuing the banning of useful technology. The process has to start with the Government creating laws that limit how you can use the data. What’s more, it has to happen fast.

The telephone sales industry has been forced to provide a clear message stating that they are going to record messages at the start of a phone call and for what purposes it is going to be used. Maybe we need it so that if you turn up at a website it tells you what it is going to do with your data (rather than asking for your permission to do so).

Page Names in Analytics and why they are important

2011-01-27T20:14:00.000Z

For those of you who are unaware, I have recently changed jobs and in those situations you always end up thinking about new things in different areas. Recently one of the things that I have been thinking about is how you set up naming conventions in your Web Analytics tools (for those that need naming conventions). It's something that I've thought about before (obviously), but only consequentially. So here I thought I'd give some generic advice on what you should be looking at when you come up with some naming conventions.

Page Names Must be Unique

Sometimes I'm not convinced that people get this one. Of course if you are using Google Analytics all your page names are already unique because they are the url of the page by default, so this makes life a little bit easier.

Of course there are downsides to this approach that aren't immediately obvious by looking at the urls of this blog:

Firstly there is a distinct possibility that there may be additional query strings that are associated with a page that you aren't aware (for example session IDs). It's unlikely with a blog, however just as simple search for the most visited url on this blog brings back two other entries in my Google Analytics for versions of the page with strange query strings that Google hasn't ignored.
Secondly it is possible that the same page may appear in various different forms of a url (eg www.mysite.com/Sports/Socks/Sportssocks.jsp and www.mysite.com/Socks/Sports/Sportsocks.jsp). This is obviously bad from an SEO point of view (you only really want to have one url per piece of content), however it is possible to exclude one from the search indexes (or use the canonical url parameter), whilst still having the same url.
Thirdly it is possible that your CMS can populate pages without changing the url of the page. This is particularly true for situations where the user submits a form and the next page says that you've completed the form successfully. This is somewhat annoying because these are normally the pages that you are most interested in!

There is a way around this, of course, because you can get your developers to use the _trackPageview() parameter so that you can specify what you want to collect as your Page Name. If in doubt you could use your canonical url parameter.

In other tools, the Page name variable is something that you are advised to set in the code (even if you can leave it blank in tools such as SiteCatalyst to pick up the URL). So this gives added incentive at the start of your implementation process to ensure that you are capturing it in a unique way.

When in a situation where you are having to specify the name of the page, then you can also come up with a couple of problems:

You may end up with two identical pages having different names depending on how the page is built (using our crude sports socks example above you can have them both way rounds and end up with a Sports:Socks and Socks:Sports page name with the same results).
Capitalisation tends to be very important and you should do everything you can to put all the Page Names into one case.
If you are using Page Title tag, many pages may have the same title. This will mean that your pages names are no longer unique. In fact this is one of the areas that it is sometimes the easiest to give pages unique names as you can assign CMS IDs to page names.

Of course there are ways around all of these problems, but you might need to make sure that you look at some real examples before you aimlessly push these things live. Testing in environments which don't mimic your live environment in terms of content may mean that you don't pick up one or more of these.

What information can you get out of these reports?

Pages reports tend to be the reports that allow you to do the most pathing analysis. This is an important bit of knowledge when it comes to looking at your site scent. Having the ability to know what people did having arrived at the site is one of the most important things you'll be able to do to increase conversion (or page views, or whatever it is that you want your users to do).

Don't forget that you'll probably also want to use unique pages in your funnel reports. This is pathing at its beautiful best. You get to take a series of users' paths and work out how far they went through your process before giving up (or going elsewhere). This can allow you to focus your optimisation skills on the areas where you get the biggest benefit.

But it is not just pathing reports that are important, it is those bookends of your pathing reports that are really important: Your Entries, Exits and Single Access (or Bounces). Having the ability to know where a user entered the site and where it exited is really important in working out what you can do to improve the site. Everyone has to Enter the Website and Exit it, but they don't have to do all the other stuff in between!

Without having unique page names (or too unique if you suffer from having different page names for the same page) you may end up misinterpreting your reports and spending too long trying to optimise the wrong page or not enough time trying to optimise those pages that would benefit most.

The other thing that these reports are almost always linked to are any link tracking technologies. What your system will do is track links against the page that they are on. If you suffer from non-unique pages you may end up confusing these reports and wasting your time trying to work out how users clicked on non-existent links or why data looks different today than it did yesterday (when you were looking at a different page name in your Analytics, but the same content).

What Should be in a Page Name?

The million dollar question about page names is always: "What should I include in the page name?" and there is no easy answer to this question.

I think it is always useful to remember what the reports look like and what metrics are available. Remember that you will have visits (or unique page views in Google Analytics) which shouldn't be added together. Just because you can't add them together doesn't mean you can't count them elsewhere in a different custom variable in a different grouping! A large proportion of the time you will be looking at the page individually (hopefully you've made them unique!).

Sometimes however, you may want to group some pages together to look at other metrics that you can be added together. With entries, exits and single access, you can add your values together to get a sensible answer (you can only do one of these per visit so there is no need for deduplication). So you'll need to think about what you can add where you would need the grouping. This might mean that you want to look at particular sections, types of pages, etc.

So hopefully that gave you some food for thought. It's important to make sure you think about these things early on in the process to ensure that you don't have to revisit your development processes to get the tagging right - you'd prefer to be able to spend that development resource making the website!

4 Predictions on the Future of the Web and Analytics

2011-01-04T18:18:00.001Z

Well it is this time of year that the world and his son comes out and does some predictions on the next couple of years, so I who am I to buck the trend? Admittedly I am going to look stupid when someone comes and has a look at this post in a couple of years time (or will they?), as everyone who ever makes predictions has ever done. Of course, these predictions are likely to be so far fetched that if any of them come true then I'll be masked out as a prophet and if they don't I'll be derided as a charlatan (or more likely, not read at all). Otherwise it'd would be pointless making them.

1. The fall of Google

Google was launched in 1996 and first started offering its Adwords service just over a decade ago in 2000 as its main revenue stream. The world wide web was first used in the 80s, so we're not looking at something that has been around for a long time comparatively. Technologies and websites come and go quite quickly in this fast paced world; it only takes a couple of slip ups to suddenly become outdated and unused (cf myspace.com).

Why do I think there will be a fall for Google? Spam.

Spam is going to cause the downfall of Google and they know it. Search for any particular product and you'll be faced with a series of websites that will just point you to places that you can buy the product from along with reviews. Go on. I'll wait for you. If you want some help, here was my search for a Daewoo television.

Google likes this because back in the day only dedicated customers posted reviews because they wanted to tell other people about the product purchases. It showed that Google was adding value by not just giving you websites that you could buy televisions from (in the paid search listings) but also reviews to help you decide (and eventually images of them, videos of them, news about them, etc).

Now however most of these are little more than spam sites whose main purpose is to make money through affiliate schemes. So they get people to make up reviews to make it seem like they have many more reviews than they actually have. Review sites of products are near on useless. In the same way that Amazon's reviews can often be near on useless.

The more users realise this, the less likely they are to use Google to search for products. The fewer product searches, the fewer the click throughs on Google adwords, the less money Google makes.

How does Google get itself out of this mess that it has made for itself? A major algorithm change? That seems unlikely. It seems more likely that things like 'instant' are being designed not for shorter keyword searches, but to encourage longer keyword searches in users. In real terms it means that SEMers and SEOers are going to have to go to even greater lengths with bespoke landing pages and keyphrase specific searches for much longer search terms. Years ago you'd hope to optimise for Daewoo. Now you're going to need to oprimise for "Daewoo DUB-2850GB reviews" (yes I know I've got an old tv).

Or users are going to go elsewhere: straight to Amazon, Kelkoo, ebay, etc. This will require you to make sure that your site is optimised for a whole range of different traffic types. Hard times ahead for those content creators (especially as they'll be asked to optimise the journey whilst doing that).

2. The Death of Online Newspapers

The trouble with newspapers is that they always tended to be for quite niche areas. The Washington Post, the New York Times, The Sutton Guardian. Unfortunately the world wide web doesn't have the limited distribution models of the old print models meaning that they are no longer just local to their states (or small towns). As they try to cater for their world wide audience of users, the local advertisers are losing interest (especially when they can use other methods online to get straight to their audiences) and the national/global advertiser baulk at the atrocious click through rates money is draining away.

What this effectively means is that there are too many online newspapers for the world. Some of them are going to disappear down a very bad financial hole (probably dragging their print editions with them). Those with more sustainable business models are going to survive, those who are just chasing audiences aren't.

That means that we're probably going to end up with a few very large news websites, lots of very small celebrity websites (linked together through social bookmarking sites like Digg, Stumbleupon, Reddit, etc or through bespoke blogging websites like Perez Hilton) and a series of websites who follow Murdoch behind a paywall (and there is no guarantee that will work). These will all be backed up by niche websites that either report on small areas in location or subject matter.

What does that means for us? It creates a tough job for the analysts who work for those newspapers who are going to have to change their life from optimising the website for all audiences to optimising for those that will provide the most value (either for the advertisers or the products that the website sells one). Newspapers are very dynamic and these can change over time very quickly. Report makers are going to have to be on their toes to ensure that those in charge get the best information and measurement is put to the front of all business and technical decisions.

3. The Mobile Web's Applications Find Wider Use

I've talked about the web on mobile in the past. We're not at the point of making our websites work on mobiles any more, we're at the point where people carry around mini browsers on their phone that can view the internet like any computer. They do it on the go and if it doesn't come quickly, they get annoyed and go somewhere else. At the moment (due to my broadband provider having a bit of a fault), my mobile phone's 3G connection is quicker than my broadband on my computer. Yep, you heard that right.

My old Sony phone - it needed a mobile version of your site

Does your website have a mobile version? It doesn't matter any more. We're already at a point where if you are going to buy something from your mobile phone, you are only going to do it if you have a smart phone that is capable of seeing the full website.

Or alternatively you might want to do it through an app. I predict that every website that sells something is going to have to create an app, otherwise they are going to be left behind. At the moment the standard with these apps is that you should create one for Apple (and then possibly Android/Blackberry secondly) that anyone can use. But apps in this sense of the word are probably going to disappear. Why should I have to download your full app to see your products.

What will happen in the future is that you will create open source applications that can be run within existing web pages. I want to be able to give my users of this blog access to (for example) web analytics books - I should be able to plug and play a right hand panel that loads (for example) an Amazon application that shows the top 5 best selling web analytics books. Users can then interact with Amazon, buy books, do searches, etc through my web page. They'll never need to go to Amazon at all any more.

This brings a whole new world to internet measurement. Not only do we need to think about where people came from when they bought things (think referrers), they'll also have to think about what website they were on when they bought things. And not only that, the suits upstairs are going to have to think up some way of encouraging websites to have their apps embedded on their sites (cost per acquisition seems the most sensible, but retailers will push for a 'rent' style model to start with is my prediction).

4. The boom of use of Analytics tools

Well, I probably would say this, wouldn't I? I mean, it's my job.

Unfortunately the boom of analytics tools has been and gone. Omniture has been and bought half of its rivals before eventually being bought out itself by Adobe. Google and Yahoo! both have their own free tools, whereas Microsoft gave up with theirs. IBM have got rid of their old tools and bought out CoreMetrics. WebTrends seems to be the only major tool not to hawk itself out.

The trouble was that many of these companies were owned and run by technical people and venture capitalists. Those dedicated to using the tools have never really had much of a say in how the tools are developed and designed for the users (yes, yes, I know about Avinash, Omniture Ideas, the Yahoo! analytics blogs, etc). Possibly quite rightly the developers of the tools have been too focussed on ensuring the accuracy of the tools and ensuring that they stay up to date with emerging technology.

Now is the time of the analyst. They say we are in the information age and nowhere is this more true than in an analytics world. Who can make sense of all this data and turn it into something that is worthwhile to the website owner? Well actually lots of people can, but they just need to be taught how to do it. This is why I see that the changes in the use of analytics tools isn't going to come from more analysts (there is a distinct lack of them already - there are 94 jobs in the UK on TotalJobs open at the moment for people wanting Web Analytics skills), it is going to come from more people learning basic analytics skills.

What does that mean for your analyst? It means they are going to have a hard time linking up the work that they do in analysis (in depth stuff), dashboard and KPI creation/evaluation (not quite so in depth stuff) and teaching those less skilled how they can use the information to do their job.

Whilst a normal consultant would see teaching others to do their job as the death of their industry, there will always be enough of the in depth stuff that the average person doesn't need (or even want) to do in their day to day job. There will also always be enough people coming through who will need to be taught the basics of web measurement and how to use the information to do their job that this will be a never ending job.

Personally I've already started buying my tweed jackets with leather elbow patches.

Five tips for Analytics for lead generation or offsite conversion websites

2010-11-22T19:57:00.001Z

I've just realised it has been a month since I last posted. That is too long, so my humble apologies. For those of you who aren't aware I have been guest posting on eConsultancy as well (follow me on facebook/twitter if you want to get all the updates!), so I have been busy still. Anyway, moving onwards, I thought today I'd look at a little side project that I've been working on looking at a website where the conversion takes place off site. This adds a level of complication that you spend half your life looking at solving. This is especially pertinent for those sites whose sales process is completely offline (eg lead generation websites) or where the sale process is online but on another site (eg PayPal, eBay, Amazon, etc). Here are my top five tips on how to get better analytics:

1: Collect as much data as you can

This may sound blindingly obvious, but if your users complete the sale offsite, collect the last possible point that they're on your site as your conversion point. By that, I don't mean the last page (because many pages have a 'checkout' link that will point out of your site). What you can do instead is tag up the outbound link so that you can collect the data.

In Google Analytics you need to collect the information as a faux page view using the _trackPageview parameter. This effectively reloads your tag when a user clicks on a link with a custom value for your page url. I'd recommend you choose wisely as you don't want to use a link that is later going to be used as a proper page. Google has a very good example of how to do this on their support site:

href="http://www.example.com" onClick="javascript: pageTracker._trackPageview("/G1/example.com");"

This onclick event is something that can be used in SiteCatalyst as well to much greater effect. As well as being able to set up custom links on any page, you can also do the same in your downloads reports and your exit link report. You can also include in your onclick event LinkTrackVars and LinkTrackEvents (look them up in the help under codes 1452 and 1453 on how to complete these from a technical point of view).

I'm sure that all other tools have something equivalent you can use with an onclick event - talk to your account managers and they'll be able to tell you.

As with all of these work arounds you need to be wary that you aren't measuring your end goal. This is just the last available point you can collect data to. What you would see if you could look at the whole data set may be that different traffic sources, partnerships, etc convert in different manners - but you have less control over that. What you do have control over is your website and how many of them you can get across that divide.

2: Use all the data that is collected

Frequently I find people who have collected a whole host of data on one system not comparing it to a whole host of data they've been given from another system. The most frequently used response to this is that they don't match up and you are comparing apples with oranges. Those of you who have been paying attention may have noticed that I even said as much in my last post - you can't compare analytics systems.

Just because you can't compare analytics systems doesn't mean you shouldn't use both of them. They can both be used for various different insightful bits of information that can then be used together. The trick is to compare the insights and not the data.

Early Analytics experts showing that there is a link between conversion and rotation of the screen

That is to say that if you have a lot data about your sales process from the point that they leave the site you can easily compare this with the information that you have about what people do on the site. A promotion driving more people to click through to sale, but sale conversion has gone down in the same time then you may be able to link the two and do something about it. Think about the things that you'd like to do about it and see if they make an improvement. This stuff is all about insight - use as much as you have available in your arsenal.

When looking at lead generation, this should be much, much easier. You should be able to track using other systems what goes on between collecting the lead and making the sale. It may take several months of course and you may need to get your company at a stage where they are monitoring things like how many phone calls they are making and how many sales based on medium (web, phone, face to face, etc), but it should be possible.

3: Collect more information in your web analytics tool

For lead generation, you tend not to be able to collect information in your analytics tools about sales, but where the sale process is online, it may be possible. In fact, it is the way that affiliates and aggregators have been paid for years. The user does all the fancy stuff on one website and when they buy off the second website a cleverly inserted tag on the final page lets you know that they have bought. Everyone is a winner. Well you are.

Suddenly you have another step in the process because you can match like with like information on where people have come from, what they've done, etc. More importantly the website that you are selling through should be more than delighted to do this because it means that you're more likely to improve your site to be able to sell through them (thus making them more money). You may find certain difficulty in this though where big websites like Amazon and eBay insist that you use their information, but if you can negotiate with them then it may be more likely. Insurance websites, in my experience, are much happier to deal with super affiliates that affiliate networks because they think that it gets a bit closer to the customer.

Even doing this sort of thing as an affiliate has its possibilities. If you can auto generate a random code from your analytics tool when people click pass the details through (a 'Customer code') that you can pass into your third party tools then you should be able to link up the data sources directly. They won't match because of the different filtering you do, but you a mapping exercise will allow you to look at like for like.

4: Collect more information in your sales funnel tool

So you can't collect the information in your analytics tool. Can you collect the information in your sales funnel tool? This is probably more pertinent to lead generation websites, but why not pass into your sales funnel tool some information about how the user got to the site in the first place? It should be a simple process using cookies to store a users campaign code if they arrive at the site and then pass that through into your sales funnel tool. That's what analytics tools do anyway so inserting this simple step shouldn't be difficult.

This allows you to do something slightly different to the above - it takes away a huge amount of information about what the user did, where they came from etc, but for your campaigns you should get a whole host of information about how likely they are to convert. This will allow you to prioritise your spend on different areas. It won't give you everything that you wanted, but it should allow you to do more than you were previously able to.

5: Integrate all your systems

I suppose this wouldn't be complete if I didn't suggest that you integrate all your systems. It's highly unlikely this will work out for you, but I think I should mention it.

Why won't it work out? Well none of your systems are ever going to match up. However if you can make it so that you are only using one system, then you may be in a better world. The difficulty of this is that very few companies make a tool that can do all these things and many companies have finance tools that are too indoctrinated in the company culture to be able to just change. Especially if you sell through many channels.

My advice? Your best value for money is through going through steps 1 - 4 above. Most of them will give you quick, cheap wins. If you're Amazon or eBay you can go for option 5, but for the rest of you it probably isn't worth it.

Why you can't compare figures from different Analytics tools

2010-10-25T19:53:00.000+01:00

I was prompted to write this post by Hitwise (of all people) who told me that they were running a beta test that shows a total number of visits. One of the questions in the test was "Did the reports show the results you expected?" The answer, was of course not. I measure my data based on an analytics tool and it measures a completely different set than Hitwise. Not only are they measuring different things to start with, they measure them in a different way, process them in a different way, exclude things in a different way, extrapolate in a different way, etc. It's the equivalent of comparing eating an apple with a fork with eating an orange with a spoon. And it doesn't matter anyway.

Why doesn't it matter? Well let us go back almost a year to a post I wrote on the difference between accuracy and precision. Remember it? In it I said:

Web analytics is all about trending over time. You can find out how something it doing at the moment and you can get some insight into that - where people are coming from, where they are going to, etc. But really what you want to do is try changing something and seeing the impact that it has one your data

and:

Well for one I won't be worrying about the accuracy of my figures. Do I have 50 or do I have 32? It doesn't matter. When I go from 50 to 51 though, I want to know that this is one additional visit, or whatever in that time period.

I even gave a practical example of this when I looked at some data modelling and statistical testing I had been doing on the site I work on.

These are all well and good, but they don't look at anything other than one tool. There is a reason for that:

Web Analytics tools will never be comparable to each other in total numbers because none of them are accurate.

And I'm including Hitwise in my broad, sweeping statement.

Because web analytics tools are inaccurate in so many different ways to each other, they should never, ever be compared to each other. But it still doesn't matter as long as they are all precise.

I know the media wants to take up its role of reporting which website is better than the other one based on ABCe audits. But it doesn't work. Each website is measured by a different calculation, even if they have the same analytics tool:

They will block different IP addresses (because everyone should block their internal IP addresses so that they can't accidentally over inflate their figures by doing silly things like unit testing against live)
They have different pages. This may seem insignificant, but if the pages are significantly bigger and take longer to load then (assuming you've got your tags at the bottom of the page like you should do) it's possible that some tags won't load.
Different types of user may be more security conscious and more likely to block your cookie.
Different types of website might use different types of cookie, especially if they are sitting on multiple domains.
The processing done by the tool in the back end is different. Processed peas go through a process shock horror! It turns out that your analytics system takes all the data that it collects and processes that into your page views, visits, visitors, etc. Funnily enough, they don't all do it in the same way (if they did, they would all be indistinguishable from each other).
The filtering of robots and spiders is done differently by different systems. HBX, I remember being told, was quite clever and could remove things automatically that were seen as not human simply by their behaviour (rather than any particular technical attribute).
I could go on with more and more differences, but I won't.

Your Average Guardian reader is going to be accessing content in a different way ...

... to your average The Sun reader (I jest, of course)

So does your ABCe audit matter at all then?

Well actually it turns out it does. Firstly you are probably quoting your figures to advertisers, press officers, the guy on the tube next to you call Bernard and all sorts of people. What if you were making them up? Wouldn't those people be really annoyed? Well the ABCe do the sensible thing of making sure that you don't make them up by double checking them. Then they do another sensible thing by making sure you don't over inflate those figures by cheating (by counting everyone twice, making all the visits yourself, etc, etc).

They also provide the likes of you and me a sensible place to look at how they have been performing over time. Of course this is slightly tinged if you take into account that some websites change their measurement system (Protip: the ABCe include a little bit at the end that says the counting method), but you can still perform some measurement over time. Whether you want to or not is a different matter.

Hitwise

Going back to Hitwise - one of the advantages that Hitwise has is that it does allow you to do comparisons from site to site, as you are not comparing one tool's figure with another. You can make changes to your website (or watch as others make changes to theirs) and see the effect on each other.

What you cannot do though, is compare the figures in Hitwise to that of another tool. This is the reason why Hitwise doesn't need to put total visit numbers in.

Interestingly, as an aside at the end, Comscore and Nielsen who do similar things to Hitwise (but with a smaller sample size in the UK) are both trying to come up with a way of integrating their panel data with real analytics data. This is (presumably) a way of getting around the issue that their data isn't accurate. Given the size of their panels it may not be very precise either, but I still don't think that it is worth them doing it. Accuracy is the wrong measure in this industry - we need precision. Nielsen and Comscore would be far better off looking to increase their sample size to improve precision (IMO).

What Adobe should do for SiteCatalyst version 15

2010-10-04T23:12:00.000+01:00

Given that my last post was number 100, I was thinking that this post should be a Web Analytics 101, but then I decided that was too cheesy for even me. So instead, I've come up with a list of things that should definitely be in the next version of the SiteCatalyst tool powered by Omniture. Although, I'm well aware that there is going to be a release of SiteCatalyst version 14.9 coming out soon, which will presumably have some nice feature updates - this is what I think should take the next step in Adobe's next step and will be a bit far removed from improvements on the current system.

Before I go into anything else, I'd just like to point out the Omniture Ideas website. It is awesome. I am going to submit all of these to that website. In the morning, once I have had some sleep.

1. Tagless Events

Let us start with the biggest one that is around. Whatever you want to call them: conversions, events, goals - the things that you want people to do on your website. In Omniture you can set them up as custom events (see the customer conversion section of my post on SiteCatalyst reports) in the tags on your page. I'm not going to lie to you, I think for implementation they are overly complicated. Moreover they are named completely wrong - how on earth can you have 'events' which are effectively conversions, whilst you have 'coversions'? I'm going to attempt to explain it to you here in layman's terms:

A custom event in SiteCatalyst can work in three ways. It can be a:

Counter - every time it occurs it goes up by one.
Currency - it goes up by a certain amount of currency which you can specify in the tag
Numeric - it can go up by a certain number which you can specify in the tag

So first off you need to make sure you set it up in the admin screen and then you also have to code it into the page. So with the counter ones you can just set them to go up by one by one by giving a parameter in the tag that says s.events = "event1". For Currency and Numeric ones you should also add in how much you want them to go up by in a separate variable which is described by the products tag. You can put in here what the product is, how many of them there have been, how much value is associated with each of them. This is amazingly powerful if you have lots of different products on the site that you want to tag through your events (eg an online shop). It's particularly powerful.

I don't like it.

Why not? Let's see how you set up goals in Google Analytics:

See how easy that is? It's one step and I don't need to change anything on the page. The results of Counter Custom events in Omniture work exactly like Goals do in the results part of each of the two tools. Omniture must come up with a way for us to create events on the fly without having to change the code on the page.

2. Custom Traffic variables defaulted to Custom Conversion variables as well

One of the biggest issues that I frequently have is where users of SiteCatalyst are using custom events in custom traffic reports. Why? They give back a gibberish answer.

Ok, Ok, they don't really give back a gibberish answer, but they don't necessarily give back the answer you were expecting. Remember custom traffic variables are counter variables (page views, visits, entries, etc) whereas custom conversion variables are on page tagged campaigns. Therefore having an event convert against a custom conversion variable makes a lot of sense. You can choose in the interface if you want the event to occur against the most recent value, the first noted value or linearly across both (it would be nice if it could be fully attributed to both - the total of each of them doesn't need to add up to the total of all of them). You can choose when you want the campaign to expire (after the visit, after a time length, after converting against one of your events) and you can choose a whole load of other things (see my learnings from the Summit for custom conversions as counters).

However having an event convert against a custom traffic variable doesn't make as much sense. And indeed you can't really track your events against something that are just counted as flat options. Except in the case of page views, where it turns out you can. How does that work? Well instead of attributing it to the first page view (you can do this in your paths reports - if you have it enabled - where you have the choice of measuring against the entry page) or the last page view before you get to the conversion, it gives it a representative value against each of the page views. That kind of makes sense - you might want to be able to attribute a value to all the pages in the journey.

But you might also want to see how good a page is at, say, driving registrations to a site, even though it isn't directly linked to from there. If you want to do this, you need to use your custom conversion reports and set up the same variable as in your custom traffic report. This seems very long winded to me. The data is collected anyway - why can't there be a button in the admin section to just enable it? Or better still, just auto enable it.

Yes, yes I know you can do this with a VISTA rule - but I shouldn't need an Adobe consultant to flick a switch.

3. Default popular plugins

Our site uses a desperately old version of the SiteCatalyst code before plugins became popular. To prove that they have become popular, Omniture's website has come up with a list of them. Some of these seem like a genius has come up with them. Why aren't they implemented as default?

The first one on this list involves taking a query string parameter from the url and putting into into a custom (traffic or conversion) variable. I don't know of any website that doesn't contain the search term that a user typed in as a query string in the url. Omniture can collect the url of the pages (it collects it at least once to populate the pages report to match the page names up to the urls). Why can't it auto-populate an internal search report. Or use an editor in the admin section that allows you to describe which of the custom traffic variable (with a switch for the custom conversion variable) you want each query string to relate to.

even Google has the search paramters in the url

How do we know that they do it anyway? We know because you can set up campaigns to be picked up in the javascript code from the query string. Why can't SiteCatalyst capture all the query strings and then just use the ones described in the admin section.

4. Update javascript file automatically

I wrote a spec a few weeks ago to put a new bit of the site under the latest version of the javascript code. It took a bit of work by me to make sure all the plugins worked correctly and that it would be fine. We've been doing some testing this week to make sure it works properly. It didn't. I went back to my code to come up with the reasons why. In fact I went all the way back to the original code as it was spat out by SiteCatalyst. It turns out in that short period of time, Adobe had come up with a new version of the code and we were working off H.22.

Now there is no way I am going to rewrite the entire thing, so I've told our developers that we're already out of date. We haven't even put it live yet.

When will someone come up with a solution that allows Adobe the right to push our minor changes to the code to everyone else? Could there be a common include in the javascript file that referenced a common file that was the same for everyone? I don't know anything about this, but I can't be the only one to have realised that their code is so out of date that SiteCatalyst is no longer supporting it.

5. Full metrics in Hierarchy reports

This is one of those words that I can never spell and always turn it into a heirarchy. I've come up with a definition for the word: It's a list of people who are in line for the thrown in order.

Anyway, enough of the pleasantries, the hierarchy reports are very, very useful. If you set them up correctly you can drill down level by level of your content. It's a great way to be able to drill down into sections of the site. If you don't use the hierarchy report, I suggest you set it up. You can of course do this with your products that you convert against as well - using classifications.

My big problem with this report is that you can't do anything else with it. It's entirely stand alone. I think that Adobe should improve the functionality of this report to allow all sorts of added extras. The added extras I'm talking about are those additional metrics (like entries, exits, single access, time, etc).

Not only should these added extra metrics be available, but we should also be able to correlate this report against any of the other reports. The reports I'm thinking of that should be defaulted would be the traffic sources, search engines and search terms. In fact, linking it up to the new channel report would be the most awesome thing.

Currently to get around this issue, I'm having to record each of my levels of my hierarchy in a flat format in separate custom traffic variables, each of which need to be correlated with each other and each of which takes up time in processing.

6. Enable visits on correlation reports

For the love of god - do this. It is so annoying that you have to create correlations for reports and then suddenly you are reduced to just page views.

There are ways around this problem. The options are that you can put both values in the same custom traffic variable and separate them with a colon (or some other punctuation), but this causes you problems when you want to know the visits all as a group, rather than associated with another value. It would be so much easier if you could just correlate and keep visits.

Plus, whilst they are at it, they could add the ability to trend the correlated reports.

6 visualisation techniques for post number 100

2010-09-22T19:57:00.001+01:00

Wow, who would have thought it? Almost three years ago when I started this thing I was sure that it was going to be a short little thing that would get no traction whatsoever. But here we are looking down the barrels of post number 100. And what a roller coaster it has been. To celebrate the 100th post I am giving away a blog post about visualisation techniques that will change your life forever (ok for about the three minutes it seems to take people on average to read my blog - no really, I measure it). If you've read all 106,821 words that I've written so far, then you will probably not be too surprised by the rest of this. Most of these are going to be things from Google Analytics, because I'm going to try and show you how many people are looking at this blog.

1. Line graphs

How can something so simple, by so applicable in so many situations. This is one area that Google Analytics does much better than SiteCatalyst, because it is an automatic time sensitive graph that is loaded at the top of every screen. You can then drill down to each of the elements and still have the graph (ok, you'll probably have to export into excel to do any comparisons). SiteCatalyst on the other hand isn't great for totals and it is annoying that you can only graph five elements at a time. However you do have to change view to get to that trended option.

One of the things that I like doing is overlaying a segment on my Google Analytics posts. In this case, you can see how my organic search traffic has been steady (with a nice increase in the last couple of months) despite the erratic total visits.

2. Treemaps

These are my favourite types of visualisation techniques and one of the reasons that I was thrown by Microsoft's Gatineau (until it was renamed and then canned). Fortunately not all is lost, as pointed out by Avinash recently you can do something pretty similar in Google Analytics if you have access to the API and are a bit handy with the developer software.

Personally I noticed that XCelisius 2008 was on a free offer recently and have downloaded it to play with. It's a relatively nice bit of software, even if it is a bit unreliable with Excel 2007. What it does have is the ability to create treemaps, although I was slightly dissapointed that the only way of presenting them was to upload them into a spreadsheet and not to put them live on the web (I'll save it and link to a live version later).

Particularly I was excited by the way that you can map a nice metric (like the one above) into various different boxed and then make it interactive so that you can find out more information about them. Above I have an example of how my site's traffic from the last 3 years splits in terms of where the users come from. That big block in the top right is the direct traffic to the site and most of the big block to the top left is Google. Interestingly the pink the bottom right shows how some of the more niche sites have a better traction with the site than the bigger ones (darker pink is more page views per visit)

In the next example I haven't written any post names on, but these are my blog posts by years that they were written. It's interesting to look at to see how my 2008 posts seem to have attracted more visitors than 2009 (well they've had time to), whereas 2010 is going really strong. But they all completely dwarf those visits looking at the home page (are you looking at the home page now?). Even though this doesn't really give that many actionable insights, one thing it can do is show you which areas you should concentrate on for improvement.

3. Maps

Ok, so you've got a view of your content based on how many people look at it, with different shades for how many pages, why don't we do the same thing for where they are in the world?

This does start to beg the question though, is this good or bad? What does it mean I should do? Is it significant that there are so many more people in the States looking at my site (that's the darkest green). What could I do with this. Well one thing you will have noticed is that I try to avoid doing things that are British English. Uses of the word 'colour' (or 'color') are suspiciously absent and I don't tend to talk about trends of things that are happening in a country too often (although, obviously occasionally).

Also, what is up with Greenland? Seriously you guys, start reading my blog.

4. Motion Charts

My biggest regret of the motion charts in Google Analytics is that I can't post one on the blog. Instead I'm limited to showing you a still of what it would look like:

This is a brilliant way of taking a couple of dimensions and looking to see how it changes over time. Personally you can see from my data that my organic search traffic has increased over time, but pages per visit decreased for that segment. I need to increase my visits to other pages on the site by linking to them or encouraging users to stay around if they come to keyword rich pages. This would obviously make much more sense if I had various campaigns running across this chart and I'd included a goal as an axis too - this would allow you to see if your campaigns were improving over time or not.

5. Dashboards

Dashboards are the bane of my life. No really, they are. Seriously I have written about Omniture Dashboards, HBX dashboards and Hitwise Dashboards. Not to mention a large essay on how much time you should spend in your day on producing dashboards. PS did you like the way that I went from talking about putting more links into the post to actually doing it. You see I am learning!

Why are dashboards such an emotive subject? Well everyone (and I do mean everyone) wants them to do something slightly different. Your dashboard will never satisfy anyone, let alone everyone. So your job as an analyst is to come up with something as simple as possible. This is why the XCelsius product is good (in a way).

Image courtesy of DashboardSpy

One of the reasons that XCelsius appears to be good is that it is small and limited to one page. There is literally only so much that you can fit onto it, but you can put lots of different types of chart on one page. SiteCatalyst's dashboards work in much the same way using the new mechanism, but having the option of putting more on it, encourages people to. Effectively you end up with a big portfolio of reports that you want to see on a regular basis. This dashboard option gives you just the headlines. Want any more? Well then you can ask about specific things that can be set up as an analysis piece, rather than regular reporting.

The other advantages of a dashboard like this is it takes your stakeholders focus away from hundreds of data points and brings it all back to your commentary. That is the value add bit.

6. Wordle

Data isn't just numbers, sometimes it is words and it is the way that you present the use of those words that can detail a lot about them. For example we can use tools like Wordle to create a map of the most commonly used words, but what we really want to do is start grouping these together. In fact, in a perfect world, we would take the entire lot and turn it into some gigantic tag cloud that shows what is going on with your site. I've got an example below of my site from the tagcrowd website (work computers don't like wordle). One of the things that it shows me is that I'm doing something right. The most common words are 'analytics', 'user', 'web'. I didn't realise I wrote so much about cookies though, maybe I'll have to stop doing that in the future.

created at TagCrowd.com

And I still haven't mentioned Google's new weighted search feature. Who'd have thought it.

What, why and how of cookies affecting your privacy

2010-08-31T19:47:00.000+01:00

Privacy on the web is a heated subject. You can tell it is because websites like Slashdot have their own section dedicated to it; at the beginning of every movie that you watch there is a bit that says 'You wouldn't steal a car'; the media erupts into seizures every time Facebook changes their terms and conditions. The Wall Street Journal released a series of articles recently about your privacy online and they started off on the humble cookie. And it turns out that it probably is all the web analytic's industry fault (more on that later).

What is a Cookie?

Let us start at the beginning, like the WSJ did. A cookie is a small text file that is associated with your browser. It can only be read by the website that sets it and also can only be written to by the website that sets it. Typically it is a random string of letters and numbers that are unique to your browser.

Simple? Well maybe so, but the reason it is causing so much controversy is the way that it is used. Initially this cookie was used so that the website you were on could write some settings into it that would allow the website to remember you between visits (or indeed between pages in a session) and allow them to maintain some personalisation (eg a "Hello Alec" message). Well this is where it got complicated, because it turns out that you can then use this information to work out how many people have seen the website, because you can count the number of cookies in your log files.

Assuming that people don't delete them from their browser that is. Or use a different browser. Or a different computer. Or more than one person using the same computer. Ok, they aren't that accurate at all, but at least you can tie a session together. Assuming you don't delete them. Or block them.

This is still all fine, until someone pointed out that looking through the log files was a pain because it contained every single file that was downloaded and every single request from the server and working out what was a page, person, cookie, etc was a huge volume of processing. Plus you had to spend time and effort filtering to make sure that you only counting real people and not robots and spiders. A far better solution would be to have something on each page that only loaded once per page and didn't get loaded by robots or spiders. Like a bit of custom javascript that was sent to its own server.

If you were going to do that, you might as well include in the request from the server some additional information about the page, rather than just the url - so you end up with custom information. And it turns out that you can use this javascript to pick up the movement of the mouse across the page, the things that have been typed in text boxes and all the other stuff. Lets collect it all and we can work out what we want to do with it later.

This would all be fine if it was just information from your website. Well it turns out that some companies didn't have the ability to do it themselves, so they let a third party put a bit of javascript on the page that pointed towards another domain's servers and gave them cookies from that domain. These were third party cookies.

Is this A Problem?

You wouldn't have thought so - the people who allowed the companies to put third party javascript on their pages would keep a close eye on the data and make sure they weren't collecting any private information (or if they were, they were explicit when they did so). And that is the case with most tracking technologies. Even with Facebook - they at least tell you that they are going to be using your information (in an aggregated fashion) to give you custom adverts based on the information you enter.

It turns out that it is the uncontrolled use of the third party javascript and cookies that are the problem. It tuns out it isn't that easy to make money on the web. If you don't sell anything then you have to come up with some new business models. One of the biggest ones is online advertising. However this isn't very easy to do technically because you need to make sure you can get a different banner on each page and that it is only served the right number of times otherwise you'll overcharge your customer, etc. So what companies do is farm out their ad serving to third parties. Who in turn allow other websites to buy ads with interactive creatives which sometimes point towards yet another third party, largely unseen by the original website, almost certainly unseen by the user.

Now in the unlikely event that a social networking site (say facebook) wanted to buy some advertising for a big newspaper website (say WSJ), they'd probably be allowed to put some interactive content into that advert, which would be loaded from their server, which you have a cookie with, because you're logged in on a different tab. Now facebook can serve you a personalised ad on another website without you knowing how it got your information. Not that I would suggest that facebook would do that, of course.

What happens next?

Well if you are a reader of this blog, you may have noticed that the EU are taking a hard line on this and are putting in place legislation to make it more explicit on the cookies you are giving to a user. This is good for the users, isn't it? Well not really, because you will only have to agree once and most of the big advertisers run on big networks. So you could allow them to pass cookies to you on one website and still be allowing them access on another website.

The IAB (the auditors of the advertising industry) are running a campaign on how to improve your privacy and explaining to users what it is that the advertising industry does with their personal information. Although I have to say the first time I saw it was on this blog post by Eric Peterson. And I'm not convinced that it does that much to allay my fears as a consumer (yes I know this is a selective quote, but it was quite near the beginning):

Getting to know you. Sites use demographic data to learn more about you so they can deliver relevant content and ads. If you're a 30-year-old woman, and an advertiser hopes to reach female users aged 18 to 35, the site recognizes that you're “eligible” to receive ads from that particular advertiser.

This leads us nicely on to the WAA. For the IAB this is big business. In 2008 the online advertising industry was estimated to be worth $45bn a year worldwide. This is colossal. The WAA industry is nowhere near that size. Potentially the largest of them all, Omniture, got sold for $1.9bn to Adobe last year. We're basically talking peanuts. But it is the technology that has been built by the web analytics providers that is being used by the advertisers that is causing the trouble. And it appears that we are doing little about it (according to Eric Peterson). At the moment it feels a bit like banning the printing press because it produces pornography.

Really I don't seem to have many answers. The Web Analytics Association seems to be more focussed on the analysts (or at least the consultants who get to use the benefits of their membership) to want to do anything about this. The Analytics providers don't appear to see it as their job to fight the causes of the customers they provide for, although it was interesting reading WebTrends take on the EU cookie issue - more interesting was the silence from Omniture, Unica and even Google Analytics (maybe they have lawyers working behind the scenes). The directive itself seems to have not got much press recently, but is meant to be aimed at improving the privacy of users. Even if it isn't a great way of doing it - surely the better legislation should be websites being more explicit about who they allow to put stuff on their web pages. Any suggestions?

Three tips on how to get non-analysts doing more

2010-08-18T19:28:00.001+01:00

Whilst having a chat with some people recently, it became apparent to me that the level of usage of our Analytics tools varies quite wildly across our organisation. In fact, whilst researching this post I was sure that I had talked about it before, but it appears that I only had on other people's blogs (mainly Eric's Webanalyticsdemystified post on the subject - which appears to be years ago!). So it comes to me again to start talking about how analytics tools have to be set up so that they can be used by many people across an organisation with many different experiences of using analytics and many different experiences of using numbers.

I suppose this is the natural progression of three previous posts detailing what a web analyst is, what a web analyst should spend their time doing and how you should create bespoke reports for non-web analysts. Inevitably there was always going to be a new post on the tools that analysts use (and non-analysts) and how they could be improved. Inevitably I am going to focus on those tools that I know best and how I used them and how they can be used.

Even from doing something as simple as looking at the main navigation of the two tools when you login can give you an awful lot of information about who the tools are aimed at (even if one of them is now named Adobe SiteCatalyst). The Google tool is very much segmented for the power user ('Custom Reporting') and the casual user, Adobe's tools on the other hand intermix the very basic with the far more complex (and in many cases the upsell of products to you as well). I think we can safely say that they're different tools aimed at different audiences, but that's not going to stop me giving out some top tips.

1. Dashboards are your friend

In Omniture there are hundreds of ways of creating dashboards using the old systems or the new dashboard technology. As mentioned already here though, none of the reports that you look at in Omniture aren't standard in the first place. Seriously. There are so many reports out there that the casual user is going to get completely lost and so your job as an Analytics person is to help them find out the things that they need. This is why you need to create a dashboard with those top 5 things in it. None of them reports should be standard. And then you share that dashboard with your team of non-analysts. They can click on each of the reports if they need more detail, but essentially you're just giving them the snippets.

In Google it is the same - as Avinash says in his posts, you need to break away from the norm and do something different to give your users the insight that they need. Those custom reports are actually ways of creating unique dashboards for each of the users. You want to use them to draw in your non-analysts so that you can take them to the next level.

2. Don't train on how to use the tool

I don't know whether I can stress this enough. It is your job as an analyst to train people on how to use the Analytics tool that you have in theory only. What you need to do is train people on how to use the data to help them with their job. This means that you need to give not just training on the tool, but what the numbers mean and how they can use them. In fact, it is almost better to do it the complete opposite way around - teach them what they could do if they know their numbers, then teach them what the numbers actually mean, then teach them how them in the tool. This way you realise that you don't need to teach them everything there is to know about the tool! Saves you half of your job.

It's unlikely you'll ever need this report

In my training sessions that I give, we all have a laptop, but we start by talking about what our job entails, what our website does and how we could make it better for the users. Then once we've extrapolated that into some metrics, then we can look at the best way of finding that information in the tools (and possibly that might mean that you don't need to teach them everything).

Or this

3. Numbers = People

This is the one that everyone somehow always forgets. The trick isn't to think about the numbers and how to improve them. The trick is to relate your numbers to user behaviour and then think about how you can change the behaviour. Too often I have people asking me how they can get more visits to their section, or how to decrease the bounce rate. Really the question they need to be asking themselves is why do people not go to that section, why do people click on the page and leave immediately (yes, some of them really do get exactly what they want from your website on that first page, but that doesn't mean they all do).

This is particularly pertinent when you look at search engine optimisation. You can do all that technical stuff so that Google knows about you and you have all the right links. But really the question you should be asking yourself is if I was a user, which link would I click on and what do I want from it. If the answer is not your link, then Google probably thinks that too and will rank you lower. You need to make your page better for me, as a user, otherwise I won't click on it.

A confused Alec not knowing why he wants to click on your results in the SERPs

Ok - three tips there for you to help the people in your organisation who aren't analytics experts. I'm very big on trying to get people in my organisation to be able to help themselves at analytics and not just be report churners for the sake of it. What is the net result? You tend to get more questions on how to find things, what numbers mean, what people should do with them. Basically lots of mini analysis pieces. This way you get to do the interesting stuff and become a 'ninja' trainer. And then you can make all the comments you want below about 'wax on', 'wax off'.

Two good and three bad stories in Social Media

2010-08-05T19:15:00.000+01:00

Social Media is the biggest thing in the world at the moment. You can hardly look anywhere in the press without being reminded of Twitter, Facebook, Blogs, etc. Well, of course you won't fail to be reminded about blogs, I mean, you're reading one after all. Lots of recent stories of failure (and deceitful success) has made me think that maybe it's time I wrote a bit about it (again). When I started this I started looking at some of things I'd written that I could link to up here (I like doing that), but there appeared to be hundreds of them. So I'll limit it to my top three: The advantages of Social bookmarking; Social Bookmarking is all about brand; and Social bookmarking is about networking. Technically speaking all of these are about social bookmarking but really social media is the same thing - you want your brand to be out there, you want it to be known as the best and you want other people to do it for you.

Success Stories and what they did well

Firstly lets look at a couple of success stories from different sides of the spectrum and things that they do well:

Omniture Care. Quite possibly these are my favourite because I use them so frequently. Who are they? They are a bunch of guys over at Omniture who decided that they should start using Twitter as an addition to their online support. Because, you know, sometimes it's a pain having an online chat service at your fingertips who seem to know most things.

Why does it work so well? The Omniture hashtag on Twitter is a great thing. It is almost like a little community of people who help each other. One person asks a question and it isn't just Omniture Care who now respond, it is the whole community of users. The more the people respond, the more people get involved. It's the ultimate system of getting a load of people with strange Omniture implementations to help each other for no personal benefit whatsoever. I try and help out where I can, but frequently I am amazed by the knowledge of this small community.

Barack Obama. Obama has over 4.5m followers on Twitter. Almost 12m people have 'liked' him on facebook. In a way, this is all of little consequence, because the real work was put in during the election. How did it work so well? ReadWriteWeb will tell you all about the work he put in during the election to get all that traction, but the real key was the people he went for.

Yep, in a nice stubby little paragraph at the end, the reason that it worked so well for Obama was that he went for the people in his demographic. He made sure he was in a space where his voters would be and they could be the ones that helped promote him. I'm sure very little of the work was done by himself - most of it would have been done by ghost writers and so on.

If you want to look at a good comparison, this year's British elections didn't really touch enough on the social media aspect. Why not? Because the demographic of user online in the media that each of the parties went for isn't necessarily the right audience. Obama did a good job of getting young people who don't normally vote to vote for him by using the places that they frequent. The British politicians by contrast appear to have spent the whole time trying petty digs at each other. A don't-vote-for-them strategy if you will.

Failures and what they did wrong

Here are a couple of stories from the recent press about things that went wrong and what they could have done to make it better.

Warner Brothers. This was a bit of a cock up by Warner Brother's and there is a nice post on the IAB site with three lessons that could be learnt. Firstly WB sent a generic letter to Adam Singer asking if he would like to be paid to blog about their television series. It turns out Adam doesn't ever watch television, let alone write about it. Secondly, he has been quite outspoken about writing blogs for money, so he may have not been the best person to ask. What Warner then did was make it worse by denying that they'd done anything wrong.

Firstly if you are going to ask people to blog for you for money - do your research first. The first rule of spam is don't spam and emails like this stink of it. Be personal and be specific about what you want, you'll get better results that way. Secondly, if you have to respond to the criticism (and sometimes silence is the right way) maybe it is just best to admit that you were wrong or that you didn't see any harm in it. Being defensive is just going to make it worse.

It's good that you want to get into the blogosphere, get writers talking about the subjects of your television shows, but paying people to do it isn't really on. I suspect that the writer would get called out by their audience pretty fast, even if they didn't explicitly say so. Work on a different strategy like inviting them to look around the studio, special screenings - all the things that you do with the mainstream press.

Ryanair. Ryan air's story is a bit of an odd one. A mad cap developer went on a foul mouth tirade (I could be a sub for the Sun, couldn't I?) at a blogger who thought they'd exposed a security flaw in their system. It led to Travolution picking up the story and asking Ryanair if they thought that this was responsible behaviour. Needless to say the response from management left a little bit to be desired as they called the original blogger an 'idiot' and a 'lunatic'.

Is this really a failure? I suppose you could argue that it isn't really - Ryanair take great delight in being the bad man of the piece (look at their most recent advertisement campaign), but this surely takes the biscuit. Yes we know they're cheap and no thrills, but this is just a bit personal. If Ryanair had just said "We're really sorry, we've put things in place to stop this happening again" then you wouldn't have the plethora of websites out there trying to come up with ways of avoiding fees. The blogosphere can be cruel when it wants to be.

Digg and Reddit: This brings us nicely on to something that is happening right now. Digg has some users who have been accused of gaming the system. What they have been doing is getting together off Digg to do mass burying of things that they don't like. Of course, I am fairly sure that this has been happening the other way for years - I am sure there were groups of users from various websites who would artificially promote things on Digg using various different user names and a network on email. Reddit, on the other hand, has already been called out by one of their own for doing something like this (even if there was no real evidence).

But for sites that are entirely reliant on their user base - can people get away with this for long. It's fairly obvious when one site keeps getting promoted up the rankings and people tend to notice (a la the Reddit issue). In fact, if you read the comments on two posts you'll see fairly quickly that most of the users don't like this. In fact, Reddit users frequently site this as one of the reasons that they moved from Digg. What with Digg looking to become profitable for the first time this year, the last thing they could do with is a mass uprising from within the ranks.

What do they need to do better. They need to come up with a way of stopping people gaming the system like this and they need to do it fast. For all the algorithms in the world out there, the problem is still that they are fighting against humans who as soon as the computer learns their patterns, will start working in a different way. It's a tough one, so maybe they hope that everyone will start gaming the system in the same way thus negating each others actions. In the meantime, it looks like Reddit is there to pick up the disgruntled.

Marketing Channel Reporting in Omniture

2010-07-26T22:20:00.000+01:00

Recently I was involved in some top secret, hush hush beta testing. But enough of virtually every single Google product that there is out there, I've actually been beta testing some exciting new Omniture products that were released with version 14.8 of their SiteCatalyst tool. Of course most of the chat around has been about how Omniture has now changed to Adobe Marketing Suite, but the casual rebrand and slightly different colour scheme for the home page is a bit more incidental to those who already use the tools.

Of more interest to me was something that ~~Omniture~~ Adobe have introduced that many other tools already have - which is the Marketing Channels report. In Google Analytics you can see this report by going to your Traffic Sources report and selecting to just show the medium, although this normally defaults to Source Medium, which will show you the website and the medium. Note that this isn't as silly as it sounds as you could have one source which is more than one medium (eg users coming from Google as a referrer, organic search, email, etc).

With Google you are limited a bit in what you can have as your mediums. You have the ones that you have set up as campaigns, plus organic, paid and now it appears twitter. Everything else gets bunged into referrer (if it has one) or not set (if it doesn't). This means you need to make sure you set up your campaigns in the right way in the first place, otherwise your classifications here are going to be slightly out (ps it can be quite funny googling links for this blog - "whencanistop campaigns" in Google gave me some links to quit smoking!).

With SiteCatalyst your choices always used to be slightly limited. You had a Referrer type report, however this was very limited - it would tell you how many came from search engines, direct and referring websites, but that was about it. Not only this but your options were limited to just 'instances' which is a rubbish metric which should be retired.

The new channel report does that extra level that you always thought SAINT classifications should be able to in your referring domains report, but could never quite do. What this can now do is come up with some proper conditional logic that allows us to claim all of our traffic in a particular way. Here below is how I've set up our traffic:

It's highly critical that you set this up in the correct way, otherwise you'll have the wrong buckets of data in the wrong place. For example you have to make sure that you have your paid search before your organic search otherwise all your paid search may accidentally end up in the organic search before you've had a chance to include it.

Just out interest - you direct traffic is automatically set up to be only included if it is the first page of the visit. Keep it like this, because if you don't then you might find lots of visits that weren't direct being included as they may not have a referrer in mid visit (some browsers and anti-virus software do this automatically without the user noticing). The other thing that I've done is that I've set these up so that they only count on the first page of the visit for all of the different types. Why? Because I want this to reflect the start of the visit. If someone comes from Twitter, but quits their visit mid way through to search in Google before coming back to the right place in the journey, I want it to be associated with Twitter. We also have situations where the user can leave the site for an alternative journey and then come back, we don't want those clicks to be associated with the second website.

Because these are effectively a custom conversion report, they allow you to do all the things that you can with other custom conversion reports. Notably you need to think about your conversions and your visits for each of your individual channels. If you've set up your revenue reports in this way you can of course work out for each of your channels which individual source is the most costly (although don't forget to count your lifetime value of each of your sources in a backend database, if you can).

The other thing that your marketing channels report does something very clever with its breakdowns. With organic search it is easy because you've got those reports already - your breakdown of your keywords is a standard report. However the ability to breakdown your direct traffic by the page that the user arrives at is something that wasn't previously possible. In fact, you might want to do this for your organic search as well, as you don't really want to have to waste an ASI on Organic search just to find out which are your top landing pages in Google.

The other thing that SiteCatalyst does really well here that I haven't seen anywhere else is that you can not only view your reports based on the most recent channel (in this case the most recent visit), but you can also view your reports based on the first channel that the user came by. This allows you to find out if your paid search is causing users to convert, but that they need more than one visit to do so. Don't forget of course that users can delete their cookies so if your conversion is more likely to be multi visit expect lower conversion rates on your first touch channel.

EU cookie laws verified

2010-07-01T23:23:00.000+01:00

A bit of a stray from the norm this week on this blog, we're going to talk about some legislation that has just been ratified by the EU. Before I start I should point out that I am not a lawyer and so this is just my opinion on the situation. But if you are a regular reader of this blog you will have remembered that in November last year I wrote a post about the EU cookie law that was introduced. Recently, this has been ratified by a working party on the subject to clear up some interpretations.

A potted history of the subject:

In Late 2007, Facebook caused controversy when they used their Beacon advertising to produce detailed targeted adverts based on the information that you had entered in Facebook. The real concern was that your data was being passed to Facebook on their partner sites when you were logged out and reconciled with data from when you were logged on.

In March 2008, Phorm, an advertising agency started to hear a lot of concern about their controversial use of behavioural targeting. Controversial was mainly because of the fact that nobody had ever really done it before with adverts. Whilst it was accepted that Amazon would monitor your web browsing pattern and combine it with what everyone else was doing to provide you with things that you might like to also buy (hence increasing their revenue), it was ok whilst they were just using their own sites data. Phorm on the other hand used data from not just the site you were browsing at the time, but also all other sites with Phorm adverts to find user behaviours to give you targeted adverts. The British Government green lighted this activity as the user had the ability to opt out of the advertising by blocking cookies in their browser.

Late 2009 the EU decided to create a new ruling on cookies. It stated that the user had to give express permission for the website to give the user except in exceptional circumstances "strictly necessary" for "essential services". This was fairly woolly of course and the IAB, the outspoken voice of the online advertising industry, said that users were expressly giving permission for cookies, because they had the option in their browsers to do something about it.

This of course, didn't go down well with the EU who have just verified the rules in the past week. They say that this is not true and the fact that of the major browsers, three of the four of them default to having third party cookies accepted. Users, they claim, don't understand the risks and therefore are not changing these settings. The IAB, of course disagreed with this and were quite vociferous in their rebuttal. The heads of the publishing industry, agreed with the IAB as this is their major source of income (although maybe not for those that are going behind a paywall).

Generally speaking, the word that is being bandied around is 'Shambles'. A more detailed piece on the subject from the last couple of days comes from Struan Robertson at the out-law.

Advertisers and publishers would rather not ask users if they want to be tracked for advertising purposes because users' answers could damage their businesses. But it's hard to avoid asking that question: the Working Party's interpretation of the law is, in purely legal terms, the most compelling interpretation, however flawed and unhelpful the law itself may be.

So the compelling story is that so far this is a bad thing for advertisers and Publishers - it really affects their business models. They are going to have to ask users if they want to be tracked on their advertising and if they don't, then they will lose out on a whole host of information. Click through rates of display advertising have plummeted and stand at about 0.1%. It is generally accepted for display adverts that the main purpose of them is for branding to encourage users to come back to the site at a later date. It is of course more or less impossible to track this if you can't draw information from the advertising back to the sale of a product. This is of course done by linking a tag loaded on an advert with a tag on a sale complete page with the use of a cookie.

Now we have suddenly affected a group of Marketers who aren't necessarily trying to do any behavioural targeting, but they are trying to monitor the effectiveness of their advertising. If they can't do this, they'll move to a more favoured method that means they can monitor this. More of that in a minute, but if the death of the newspapers was started by the web, the death of online news could be started very easily in this new ruling.

Web Analytics

It would be remiss of me if I didn't talk about Web Analytics. Most Web Analytics make use of cookies these days. In case you hadn't noticed there are lots of blogs talking about how to do web analytics and make use of the data collected. Virtually every website in the world has a web analytics tool associated with it. Does this mean that every time you look at a new website for the first time you are going to be prompted to opt in or opt out of using the cookies associated with that website. The ruling seems to be suggesting it that way. This will have a serious impact on your analysis - opt in always means that you are going to be working on a smaller sample that you would otherwise.

What about networks like YouTube? Every time I embed a YouTube video on my site, Google is going to drop some cookies on my computer so that it will know that this video has been viewed and by who. This sounds like not only are you going to have to expressly ask permission every time you look at a website, but potentially at every single bit of embedded content.

Finally there is an opposite end to the advertising - if I pay money for Google to put me on their sponsored search results, I want to be able to tell if causes more people to buy my product. If it doesn't, then I'm not going to spend that money. How does this work? Well it links up the information that is passed in the tag of the landing page that says that you have come from a paid keyword, with the information passed in the tag of the thank you page. How does it do this? Yep - they have the same cookie ID.

What about online optimisation tools like Omniture's Test and Target or Google's Website Optimizer? These tools allow you to do A/B testing on your site, but to do this they need to know which one performs better. Yep, they need that cookie information.

And in all the examples above we haven't really done anything that the user would notice, that would infringe on their personal data rights, but what we have done is optimised our business and increased revenue.

This whole this strikes me as an odd way of working the system. It's a bit like banning glasses because people are drinking too much alcohol. Not only are you having a big effect on things like Web Analytics which are mainly harmless ways of getting information on how to optimise your site, but the behavioural advertising guys are just going to come up with new ways of getting around this issue.

I'm hoping that the UK Government takes a stand and doesn't ratify it. Unfortunately the only ones arguing against are the IAB and the Publishers, who are concerned about their revenue streams. Really we need one of the big analytics companies to make a stand - an Omniture, a Webtrends or a Google.

6 Truths of the Internet and Web Analytics

2010-06-21T23:28:00.001+01:00

Well I was reading an article (essay?) in the Guardian yesterday about the internet and it led me to the conclusion that I don't write enough posts that are number based. So I decided that maybe I should create a list based post. At least then I can sensationalise it on a Twitter with a "Top 6 things to make your Analytics work" headline. And who said headlines were dead in favour of SEO friendly titles? Not me.

Anyway, the inspiration for this post was based on a couple of quotes in that essay linked to up there:

A funny thing happened to us on the way to the future. The internet went from being something exotic to being boring utility, like mains electricity or running water – and we never really noticed.

Do you know why that was?

I'm writing this in 2010, which is 17 years since the web went mainstream.

1. The web happens very quickly

Or maybe that should read, things happen on the web very quickly. Can you remember what you were doing 17 years ago? I can just about. I was 12 years old, I'd just started senior school. In the summer I played football non stop in the park (jumpers for goal posts, etc). My home computer was a BBC micro. I read books and played on a game boy.

The BBC Micro (yes I did play Elite on it)

Today my home computer is a widescreen laptop that overheats when more than one process is running. I read books on the tube (at least that hasn't changed) on the way to work. My gameboy has been replaced by a mobile phone that doubles up as a camera, a video player, a music player, a video recorder, an email client, an internet and, at one point, a spirit level to put up my blinds straight. Plus I've got a dodgy knee injury that stops me playing football in the park with jumpers for goal posts, although I wished it hadn't.

What is the point of this story Alec? Well the point is that in 17 years the whole world wide web has happened. It is a very short time span. We have gone from nothing, to a world where you keep in contact with your friends not through house phones and letters, but by facebook and twitter. Where you write something and within hours people all over the world are reading it. If you don't evolve with the web then you are going to be left behind very quickly.

What do you have to do to stay up to date? You have to change your site frequently, experiment lots and see what works based on the data that you provide.

See a rather brilliant piece by James Kelway on why you should have agile UCD.

2. Don't forget the long term

Whilst you think that changes happen quickly on the web and you should change frequently, experiment and test lots, don't forget the long term goals of your website. Your website does have some long term goals, doesn't it?

Good.

Thanks KiJosh

Whether your long term goals are to be the best in your industry, to gain an audience share or make enough money and customers to sell out to a larger company (or whatever else) then you also need to think about the long term. Long term strategies mean that you can't go chasing quick bucks if you end up with a long term downtrend. This is as true in the media world where publishers go for a short term page view/advert boost for their bottom line now, rather than thinking about how to grow their audiences over time. Maybe Rupert's bold move with The Times is one of these moments. Although the suspicion is that too much of going for short term has affected his long term offering.

The truth of the matter is that whilst there is still an audience who are moving online from offline, it can often seem like it is easy to get a bigger audience (even if you are cannabilising your own audience). However this shouldn't shy away from using competitor analysis to find out if you are doing it better than your rivals. See tips on using Hitwise (maybe I do have list posts) and Avinash's 8 competitor intelligence tools post.

3. The web isn't all of the internet

Ok, I've nicked this one from the Guardian, but they've worded it differently. If you don't believe me, go an have a look. See - they didn't claim they'd stolen if from the Guardian. It is very popular at the moment to start going on about how it is the best thing in the world to embed things on your pages. Whether it be a YouTube video of a cat falling asleep (yes that is cute, isn't it):

Or an embedded twitter stream, or a picture of whatever, or a podcast, or..., or...

It is all very well and good adding these things, but what value do they add. Can you measure the value? Can you measure how used it is?

At a previous company I worked for they introduced a video channel on the site, because video was a new thing at the time. It was used a bit, but most of the users wanted to see related content - be that a video about something similar, other articles about a similar subject, etc. Why does Wikipedia work so well? Because of all those links, inline, to related content. Why do video channels not work? Because only the website owner turns up at the website to see the videos. How do we know this? Because we measured and analysed it.

But on the other hand, ignore these at your peril. If you are providing the content in a variety of different ways, your rivals are going to. get those podcasts in. Get those streaming videos in. Get those links to files attached. You're providing a service, not a website.

4. You have competitors

Without doubt, whichever industry space you are in, you will have competitors. Even I have lots of competitors. If you search Google for Web analytics blogs, you get 44million results. There are two ways you can deal with this information. You can either decide that you are going to embrace this information and treat them as your peers or you can try and crush them as your enemy.

Some loser website's Analytics blog. They'll never amount to anything

Personally I always think that option one is better. If you met an attractive woman who you thought would go perfectly with one of your lifelong male friends (or vice versa) you would introduce the two of them. You'd tell each of them about the other one, the features, the pros, the cons and when you use them. Your new acquaintance is going to value your judgement and keep you in mind next time they think of something similar, to see what you know. Your lifelong friend, on the other hand, is probably going to keep coming back to you because they know that you give good advice. Plus they wouldn't have been your lifelong friend in the first place if they didn't know this.

What's more - in a real world environment of websites, you can measure this. You can tell if people are coming back more frequently or not. You can measure all of this. If you try linking out and find out more people come back, then you are doing a good thing.

5. Your users will get involved

Remember the old adage that 90% of your users will just read your content, 9% will comment and 1% will actually contribute. If you don't embrace that principle, you are going to lose 10% of your audience straight off. Those are the 10% that want to comment or contribute.

Want to know what? This is the 94th post of this blog. There are 57 comments on this blog. There has been one guest author.

What does this mean? If you don't allow your users to contribute, they are going to go places where they can. When they contribute, they are going to persuade some of your 90% to go with them. The user reviews, the comments, the "this is my situation that is relevant", this is what happened to me, this is what I think. They may not all be valuable, but the users who read it are going to be appreciative of it. And the websites are all going to be appreciative.

So if you don't want to put reviews on your website for products? Your rivals will and your users will go there. Don't want to put yourselves on price aggregators? Your rivals will and your users will go there.

What's more - you can measure these things. No really. I just did it up there. You can monitor how many comments you get over time. If you're clever with your analytics tools you can even measure how many people look at the page depending on how many comments you have, etc. Web Analytics, remember, isn't just about what your Analytics tool throws out at you. You need to get all of that customer Insight about your site/brand and bring it all together. That's what being a web analyst is about.

6. Information paralysis is not your friend

There is a lot of information on the web. Not just that, but your website produces a lots of data. I'm going to say it again. There is a lot of data. Your job is to find out which of it is useful. Why? Because your boss probably isn't going to, they are going to rely on the people below them to do that for them. Then they are going to do whatever the one with most convincing argument suggests. You have to be the one with the most convincing argument and the only way you can do this is through providing data.

The Guardian article talks about Orwell and Huxley. Orwell was of course determined that the state would control power by monitoring everything and only releasing the information that was important. Huxley maintained that the way things were going there would be an ever increasing amount of information that it would be impossible to tell what was important. Funnily enough, it turns out that they were both right when it comes to Web Analytics.

Data Modelling and Statistical Significance in Web Analytics

2010-06-07T21:02:00.000+01:00

In the last month or so, I've been doing some data modelling so that we can build up some 'forecasts' for some of our KPIs for the next year. Particularly of interest for you guys will be what I've been doing in forecasting the number of visits and interactions with the site. I've also been involved (although in a rather limited way) of some penetration targets and how we are going to achieve them. But that is not so important here. What I thought I'd do is go through some of my though processes on this one and why it proved to be more tricky than you'd imagine.

First though, lets go back a bit in time to when I wrote about the difference between accuracy and precision (PS, no analytics vendor has yet given me a figure as to how precise their tool is). Let's start looking at some real data so that we'll be able to see what we mean. Below, I have compiled a graph that shows how visits to the site I work on changed month on month throughout the year. I've added in a line that shows the median and a line that shows the mean for this as well.

So here we can see what looks like quite a nice distribution. My rather skewy axis doesn't show you that the mean shows an increase of 14,000 visits a month from organic search, whilst the median shows a 2,500 decrease. The median is skewed slightly in this report becasue there are an above average distribution just below the zero point, whereas the mean is skewed by two very low values (at -300,000) and four very large values (at +300,000).

However there are only 48 data points on this graph, so we end up with quite a large standard deviation of almost 140,000. The standard deviation shows you how far away you are on average from the mean of the distribution.

So has this value been going up at all in the past four years? By the looks of the table above, it looks almost impossible to tell, if anything it looks like it hasn't. So we would use a statistical hypothesis testing technique. To do this, we would take our mean (14,000) and divide it by the standard deviation (130,000) divided by the square root of the number of values (48). Because the number of values is low and standard deviation is high, a hypothesis that the visits (month on month) have remained the same would yield a z-value of about 0.68. Looking this value up in a table it would mean we're getting to about 0.50 (double the value in the table given because it is only looking at half of the normal distribution).

This means that if we had a hypothesis that the value hadn't gone up at all 50% of the time we would be right.

This is not very precise at all. Remember that previous post? Even if we are completely accurate, in a month on month fashion, we can't tell whether we are moving upwards or downwards at all. In fact, 50% of the time we think we're not moving at all! This is really not very precise. If you were going to build a model for you KPIs and you said, I think that we are not going to move up or down, but I can only be 50% sure, I would be laughed out of the office!

So instead we decided to rebase the data. Instead of doing month on month variances, we went for year on year (based on each month):

Here we have a much similar looking pattern of distribution, but on a much smaller scale (the data only goes from 30,000 to 350,000). Because of this, our average is 178,000 and our media is 191,000. Now our standard deviation is only 67,000, however we only have 36 data points as opposed to the 48 above.

Doing the same test as before, we would come out with a figure that was much lower than it was before. Month on month we only had a 50% confidence that we were increasing at all. Year on Year that confidence has gone up to almost 100%, because the standard deviation is lower in this range (even though the volume of data points is lower). This would work out to a z-score of almost 16 (so yes well over 99.99% sure that it is going to change - we're off the radar of the table given).

Normally you would expect to have to increase the volume of data points to be able to get something statistically more likely. Here we have remodelled the data to be able to show that the way to getting good figures is to be more precise. In our case we have lowered the standard deviation of our points to make them more statistically significant.

More traditionally you would use other ways to show that your changes had statistically significant changes:

Sample size. When it comes to research, this can be one of the better ways of improving. Asking a small sample what they thought of your website, changing it and asking another small sample may yield results that aren't statistically significant. Be wary though, increasing your sample size does not proportionally increase your preciseness. Remember that formula up there had a square root sign above the sample size, but this can be very important at the lower end of your scale
Increase your time period. This is usually much easier to do in Analytics. If you don't have enough data for an accurate model, then increasing your date range should give you more figures. This is effectively increasing your sample size by counting for a longer period. Be wary of the pitfalls though - a small change in your site, campaign, real life could all mean that you are not comparing like with like
Improve the precision of the tool. For my example we were looking at hundreds of thousands of visits to a website that was changing on a regular basis to an audience that was changing on a regular basis over a four year time period. If you can, you should break this down into smaller time frames where things haven't changed (as much). This should reduce your likelihood of an outlier, which will much reduce your standard deviation
Look for cyclical patterns. My cyclical patterns are definitely yearly, however you may discover that you are looking at weekly patterns or even monthly patterns that show how people interact - comparing these changes may allow you to see a better pattern emerge and reduce your standard deviations

If you want some additional reading on this Avinash's excellent analytics tips from four years ago (!) may be of some use to you. More recently Search Engine Land has got in on the act and I am sure there are a host of others that you know about that you can share in the comments below.