Web Analytics is dead

Well if James can declare that Usability is dead, surely I can declare that web analytics is dead.  In fact, this whole point is quite interesting and will actually have some real impact in where we are going and what we are doing.  So lets start at the top shall we - why does James think that Usability is dead and what does that mean for Web analytics?

Content is king

In the old days we used to say that Content was king.  Even Wikipedia has got in on the act.  Search Engine Watch said that it is the golden rule of SEO.  The reason being that if you didn't have any content, there was no way that anyone would ever be able to find your site (certainly not through a search engine) and then you wouldn't get any visitors.  And even if you did get some, if it wasn't original, well written, etc then the users wouldn't stay.

Context is king

James argues in his article that this is no longer as true as it once was and now the key focus should be on context.  Whilst he doesn't necessarily say this in his articles, he is pointing to the fact that many websites have been gasping for visitors when what they are actually looking for is valuable visitors.  Companies that sell things have long realised this and have been making more and more use of their analytics to be able to increase their ROI on their marketing spend.  Don't forget about the good old SEO though - this is very important for companies who sell things - and even unpaid inbound links.

 
 

For me personally though, a good example of context could come when you Google my name.  In this country (as you can see from my unpersonalised results above) you get a series of websites that involve me and I am very proud of that fact (give me a round of applause!!!).  Whilst I am particularly happy that if you type in my name you get to see my blog first (and second) for the simple reason that this is the method I use to promote myself, this isn't so good for other reasons.  What happens if I meet someone whilst in the pub who thinks I am fantastic and wants to add me as a friend on their myspace/facebook/bebo account.  Where is the link for that?  Recruitment consultants - they want that Linkedin profile.  Fellow web analytics professionals - they'll want the first one.  Twitterers (if such a word exists) - they want the sixth one.

So you can see how just on a simple web page the context of the user is vital to be able to tell what they want to do on a particular page.

As James points out in his article, it isn't just the person any more.  It is the person's location.  If you look at things like the Blackberry and the iphone it isn't just the context of the person it is the location of the person as well.  No longer are we segmenting based on users being at home or work, now they can be on the move as well.  Tamar's report into the effect on the internet the recent death of Michael Jackson had (RIP - I had tickets for July 15th) states that Google had a massive spike in traffic from mobile devices.

Suddenly if we are taking into account the context of the users approach to a site and their location we can take in so much more as well.  Look for example at Twitter.  Are all these tweets above about Michael Jackson really that interesting in isolation.  When I look at twitter (@whencanistop, in case you didn't already know) I tend not to look at global trends - especially on my mobile - I am more likely to look at what the people I am following are writing about.  Now we have to take into consideration the network that the user belongs in (not just for twitter, but for facebook, myspace, etc, etc).

So because all of this James wants to take a different approach to usability - one of collaborative approach.  I quite agree with this.  Context is difficult often because it can instantly change.  You either have to design for the most valuable segment (use your research and your analytics to work this out) or you have to create something that is so dynamic that the individual can alter it based on their context.  This is really the route into personalisation.

I've nicked the above picture from today's bbc website (or any day's bbc website for that matter) who had an article on creating graphs.  This is a good point of view though - as we move into this targeted area everything becomes increasingly difficult to measure.

What do we do about it?

James in his comments on his post uses the word 'average' which is something I hate.  So you'd have to do it the way I like to do things (ok, I'll do it this way, others will work on averages).  Firstly an average location is meaningless.  An average number of social network connections is meaningless.  We'll have to come up with new metrics and new ways of measuring them. Can we find out how many friends users have when they pop up on our site from Facebook?  Not really (at the moment), but we can ask them in surveys and so forth.  This shouldn't detract away from us still measuring what the users actually do.

Personalisation is a real problem though for measurement gurus like me.  We have to not only work out what the user has done (or we've done for the user) to their bit of the site before we can look at the data, but then we have to take into account all the crazy possibilities of that action.  Suddenly we are starting to build up individual profiles of users to give them exactly what they want, when they want it, where they want it.

The flip side of this, of course, is all those who don't want to share this level of personal detail.  How do you cope with them?  They're not logged in.  Do you create one massive profile for a person called Everyone Else?  And then how do you measure that and improve it.  I think that those clever people at Omniture, Google, Yahoo!, CoreMetrics, Webtrends, etc are going to start having their databases even bigger to hold this information and it is going to have to be more malleable to pull out reports.

I for one am very excited about the prospect (if not the hard work involved :))

SiteCatalyst Excel client isn't quite as good as report builder

In my previous job, when I first started thinking about moving from HBX to SiteCatalyst, I had a big concern over whether we could use the excel client in the same way as we did Report builder.  Now I use SiteCatalyst for my new company, I wish I'd blogged about Report Builder because I don't have access to it any more and realise how good it was.  So instead I am going to blog about the SiteCatalyst Excel client, because I can show you bits of it and how you can use it to your advantage.

Firstly though, a bit of history.  Report Builder worked so well and was so popular for a number of reasons.  First on this list was that it was impossible to do a number of things in HBX that seemed like they should be standard and secondly, creating filters in HBX could tend to be time consuming because you had to navigate to the correct report, then wait for it to load, set up your filters, wait for it to load, etc, etc.  Far easier would be to set up one report in Report Builder that referenced a cell, then you could change the cell really easily to get a new report.  Or you could even copy and paste it to get two slightly differently filtered lists next to each other.

Particularly useful was the functionality I used for failed search terms on the internal search.  We'd set it up so that we collected not just the search term, but four associated metrics (eg type of search, any other drop down variables, etc).  We would then use this list to look at which terms we weren't throwing up results for.  This was particularly difficult, because really we were only interested in one type of search, but the user interface in HBX didn't allow us to filter these terms for just those using site search.  Report builder did allow you to do this.

The main area I used it for was reporting upwards on work that we'd done.  Specifically the way the tagging worked in HBX meant that every page was named after its hierarchy  followed by its page name.  This way if you tagged all one type of content in the hierarchy, you could easily filter this list in report builder.  You could then create a report showing how a total of all of the entries changed over time to show whether the content was doing better or not.  You could also do your single access to show whether the bounce rate had gone up or down.  Now there was another step.  If you had 20 different types of content, you could paste them into a spreadsheet and get report builder to reference them in a nice copy and paste option.

For SiteCatalyst, they have got around a lot of these issues with the use of correlations.  The one thing I don't like about correlations is the lack of metrics associated with them (you can only have page views).  For a content rich site - where you really need entries, single access and visits - the correlations don't really work that well, so you want to set up reports that you can filter.

This leads us into another point - if you don't want correlations, but you have set up your account to use the custom traffic variables - can these be of use?  Usually recommendations are that you use your custom traffic variables to pick up the hierarchy of your content.  We do that in a really useful way that allows us to set up the correlations that we really need - against referring domains, keywords, etc.

However the downside is that whilst these are good at rolling up visits into a nice deduplicated measure and entries (assuming that you always pass a value in), your single access is nonsense.  In this case single access is just a situation where that was the only variable that was passed through.  So if you look at several different pages with the same custom traffic variable (but different page names), but only one value for that custom variable, you are counted as a bounce.

An easy way around this is to develop an ASI slot for single access visits and compare this with your main account for entries.

This is where the Excel client comes in useful.  Rather than sitting and switching between accounts, it can be useful to compare your two accounts in Omniture using a bit of 'Copy and Paste'.

 

Once you've created your report in sitecatalyst, it is very easy to copy the whole sheet and paste it in a new workbook, before editing the newly copied report to show the single access report.  This is useful particularly if you are going to want to use this regularly, because of the ease of opening up the a report and refreshing all.

I've also found it useful to make sure that you set your date ranges based on excel sheets, so that they are easily changed.  Don't be fooled in the UK if it thinks that you have the wrong date range.  The clever thing about this is when you copy and paste, you can link your cells back to one original, so that you don't have to change lots.  In fact, this is where a whole series of 'vlookups' in your excel become important (ps learn excel like the back of your hand - there are so many functions that allow you to do more or less everything).

The other advantage the excel client has is that it mimics your user interface perfectly (unlike the datawarehouse and discover that lack any new reports you've set up through SAINT).  One disadvantage though is that it mimics your interface's metrics perfectly as well, meaning if you are missing visits in your report in the user interface, you'll be missing them in the excel client.

One area that report builder particularly outweighs the excel client is that you can't run any of the correlations in the excel client.  Really I used report builder to pick up places that I couldn't create graphs over time and would have had to have downloaded report by report by report to get a series of months.  With SiteCatalyst you can do all that in the user interface except when you do traffic correlations.  It would be nice if you could run correlations in the excel client, so that you could create a report over time to see if something is going up or down.  I suppose you can't really have everything.

Personally I think maybe Omniture missed the point of report builder.  Where it was probably created to allow people to create automated management reports, I certainly didn't use it for that - I had a whole library of reports that could just be updated.  I find this hard to do in the Omniture Excel client, which has to be mainly used as creating a set of high level management reports.  Check out the Omniture blog post on the excel client for more hints and tips.

SAINT classification - a walk through

Calm down all those Roger Moore (or Val Kilmer if you're a bit younger) fans, I'm not going to talk about him at all in here.  Sorry.  Instead I am going to talk about something that I am doing at the moment at work (surprise surprise) and funnily enough this is something that Adam Greco has talked about before on the Omniture blog, so I'm going to try and do this in a different style, otherwise I'll be saying the same thing has he said and I'll be adding no value at all.

Essentially SAINT is a type of classification that you can do in SiteCatalyst that allows you to classify one of your variables into some different stream, if you can't code it in on the page.  In our case, we have a series of content that are written by people who work in various different departments and in various different teams in those departments and we don't capture in our CMS what they are.  However they are keen to know as a department and team within the department how their content is doing as a whole.  This is where the classifications work.  We have a list that is coordinated for one group of content that describes this one to many relationship and what we want to do is put this into SiteCatalyst without having to code it into the pages.

With HBX in previous companies we did this by populating the custom elements of the tags from the database, but sometimes it wouldn't work and sometimes it wasn't right, so you'd have to go back and get your dev team to sort out the tagging.  This could be expensive, time consuming and in the mean time you'd lack any data on the stats.  Well with SiteCatalyst one of the benefits is that you can run the historic data as well and if you make a mistake, you can correct it and it will replace all the data.  How many times have you sat there and someone has said "Oh, I didn't realise that this fell outside the standard and wouldn't be included!"

The process starts rather simply.  You have to name your fields.  In the Report Suites section of your user management, choose the report suite that you want to add the classifications to and then go to edit settings (as in the image below), choose 'Traffic' and 'Traffic Classifications'.

This will bring you up with a new screen, where you can add in your new reports that you want to create:

 

Here in this case, I've got my 'Guide' variable which I have mapped across to the different departments and am now going to add in one classification of 'Department', another of 'Agency' and another of 'Service'.  Remember you can add classifications to classifications, but the upkeep may be more (or less) difficult.  In this case, maybe it would have been better to create the first classification of 'Service' and then a classification of 'Service' could have been 'Agency'.  Given the Agency that a Service sits in is unlikely to change that frequently (if at all, but you never know with the British Government), then this may have made more sense.  In this case, I've actually added in as three separate columns because in some cases there is no real agency or service and these fields are going to be left blank in favour of just a department.

Now, as you can see below, I've added in a few extra fields into my left hand navigation when looking at data:

Unfortunately there is nothing in these data reports at the moment, because I haven't told SiteCatalyst which Guide relates to which department, agency and service.  So I need to upload this into SiteCatalyst.  But there is one thing I need to do first.  I need to check the format that we are picking up the guide name in.  This will determine how I am going to upload the data.  In this case, there is a bit of a difference from just the name of a guide and whether I've truncated it to a certain number of characters, etc.  I'm saying this deliberately here, because I know I have and I do.  It means a simple list of guide names with their associated Department and agencies isn't going to work.  I need to use excel to concatenate the name of the theme into the name of the guide.  I also need to check through the work done elsewhere to find any discrepancies.  Unfortunately this is a one to one relationship, so any particular piece of content that has two creators is going to have to name one as the main owner.

This brings us on nicely to the point where we have to upload the data onto SiteCatalyst.  Fortunately SiteCatalyst gives us a nice option of doing this in a very easy way by giving us the template that we can download.  Again in your Admin options, you have the option of SAINT Classifications, which this time is the area you want to be looking in:

The template you download will have very little in it.  It will tell you the title of your columns and a few lines at the top that don't really do anything else.  This is where you need to put your classifications in.  So your first column (the key) will have the values that appear in SiteCatalyst already (in this case it is the name of our guides) and in the second, third and fourth (and however many other classifications you have) you can put in the other values:

 

One thing to bear in mind, is that you need to make sure that you don't save this file in another format.  Make sure it sticks as .tab and doesn't pick up any excel formatting, otherwise you'll end up not being able to upload it properly.  

Importing it is easy.  Just select the import option from the same menu you created your download from, choose the report name, report suite and filename and then upload.  Don't worry too much about overwriting data on conflicts - this will only be a problem if you are changing the data from previous imports.  At the end of this you'll get a report telling you how successful it has been (I just got this one below):

Now it is time to go and look at your lovely data!!!

One thing I'd say is that you have automatically created correlations between your different new reports, but not to your original report.  One thing you could do would be to create another column in the table that linked these into the original report for instant correlations at no extra cost.  I think this is something that I am going to try for next time.

What affect will the online publishing crisis have on web analytics?

Keeping my eye out in the press as I do, I haven't failed to notice that the online publishing world has been going through a bit of a crisis recently.  Rupert Murdoch recently claimed that he might move his newspapers to a subscription model.  Has it really come to this?  Will we have to pay to read news online as we used to in the old print world?  It wasn't that long ago when I was writing about how the business models of publishers work. Even Time magazine put the death knell on 10 news publications in the States.  So if the newspapers are suffering, why are they suffering and what does that mean for people who work in web analytics in that industry.  More importantly what can they do about it?

To sum up the problem - in the old offline world, people paid money to advertise in newspapers and magazines, the cost of printing and distributing was then covered by the cost of the publications plus or minus a bit extra depending on how much they could get away with charging for it.  This was a double edged sword - lower the cost of the publications to increase your distribution, the more you could charge for the adverts and the more you got from selling the thing.  Want to make more money instantly?  Put a couple more pages in so you can sell more ads; bump up the costs of the publication to make more in subscriptions; charge your advertisers more.

In the online world that hasn't really worked so well.  Advertisers don't pay per person online, they pay for a very definitive number of adverts, so if you have more people, you can't really charge more.  This led to the sudden plea from publishers for more page impressions (cf ad impressions).  More page impressions meant more available advertising space, but that is only true until the point where advertisers are willing to pay for those adverts.  If you have spare advertising inventory, more page views doesn't mean more money.

As publishers realised that this was happening, they started to wonder if they could charge their advertisers more by getting more people seeing their message.  Well this is ok too, but you're still paying for a set number of ads, not a set number of people (imagine if someone brought out an advertising model saying they would guarantee a number of users seeing the page rather than page impressions - would you do it?), so advertisers were reluctant again.

This was all fine whilst the print revenue was gradually increasing and the online revenue was going up rapidly, but in 2008 the increase in online spend was only 2% in the last quarter compared to the previous year, whilst offline advertising shot down with the recession.  See the Wikinvest page for New York Times:

The drop in revenue in 2008 was primarily attributed to a 13.1% decrease in advertising revenue- print advertising declined 16.7% and online advertising advertising [sic] increased 8.7% during the year.

So that is what is happening in 2008, with 2009 getting worse.  This means that publishers are having to cut back on resources and investments in an attempt to cut costs and keep profits up (or debts down).  One of the areas that people often see as being culpable for job cuts are things not are not production based (I won't get into the whole debate about whether sub editors are or not).  Web analytics is not production based.  And whilst the analytics vendors are going around telling you (and your boss) that it is easy, then you are likely to be the one who gets the chop.

So what can you do to avoid it?  You need to be adding value.  You need to be showing that without you, your company would make less money.  If you read Eric Peterson's blog post on beating the recession, he makes five recommendations:

Tip #1: Focus on Increasing Profits, Not Minimizing Spend
Tip #2: Don’t Be a Report Monkey
Tip #3: Start Watching the Job Boards
Tip #4: Think About Your Skill Set
Tip #5: Network, Network, Network

Whilst numbers 3 and 5 are specific for making yourself recession proof, numbers 1, 2 and 4 are areas that you can think about in your current job. 

Focusing on increasing profit is difficult in this industry.  As we've just seen above, the profits aren't really going up that much any more.  What it does mean though is that your reporting shouldn't just be focusing on page impressions and visitors.  You should be looking at how your business model is set up and reporting against that.  Get your company into a focus that there are many metrics and the most important one is on the bottom line with a pound sign (or dollar).  So you need to work out what is going to increase that.

More importantly I think is that you shouldn't be a report monkey.  Any reports you are asked to send out should not be reports.  They should be an analysis.  They should be telling your customers what the data means to them and what they should do about it.  In the past I have found that the commentary that I have put into reports and lead some people to not even bother opening the report.  If they are at the point where they go off and take your recommendations without even bothering to look at the numbers, you are doing well.  Without you, they'll have to sit and do their own analysis which may lead them in the wrong direction.

Thinking about your skill set is something that is great for your personal improvement.  It is also great for your company's point of view, because it means that you should be able to relate your web analytics reports and analysis back to the more realistic methods for your recommendations.  Want to tell people how to improve their blog?  Try writing one.  Want to tell people to improve their SEO for improved conversions?  Why not look into ways of doing this - try some link building yourself, talk to your search team on technical improvements, etc.  Does the social media work not create much engaged traffic that wants to view many pages and complete lots of money generating actions?  Look into how you could do it yourself.  Sign up to all the social media things.

Remember that if you work at Google you are asked to spend 20% of your time working on other projects that will help Google in the long run.  If you spent your 20% of time learning new things in the publishing world that would help you write your recommendations that can only be good for you and your company.

Personally of course, I left the online news industry at the end of last year for a Government content website.  We don't have quite the same issues.

More Web Analytics for Content Editors

Well I'd been writing the previous post for ages and decided that it was time to cut it in half and write one half that dealt with the metrics that content editors use and the second half was going to talk about how to use those metrics to work out the best way of creating new content.  This is essentially the crux of the matter is more or less what I wrote at the bottom of the post.  We want to be able to use the analytics we have on our existing content to work out how we should write or edit our content.  Also we need to work out how we are going to use that information to market our content to get the most visits and the best conversion, whatever that may be (and remember that you don't have to measure conversion in a conventional manner if you are not selling anything per se).  I think this is the process that the content editors should take when they have a subject to write about:


1. Find content on a similar subject

This may sound a little odd at first, but if you are writing a new bit of content, then you want to look back at your old bits of content.  No really, that is a great place to start.  Particularly you want to find things that were around the same sort of subject.  This is where having a good taxonomy will come in useful because you'll be able to search through your archives in a better way.

Remember that taxonomy doesn't mean navigation structure.  Your taxonomy should be a series of words that describe the content (like the ones on the left hand side of this blog), whereas you navigation structure should be grouping similar bits of content into a navigational structure that the user will understand.

Using the taxonomy approach it means you should be able to cut across many ways to find similar content.  Particularly you'll be looking for where the subject matter is similar or the same and/or where the topic has been similar or the same.  Eg if you are writing about the measuring SEO with Omniture tools, then you should look at your other SEO content and your other Omniture content.

2. Find the content in your web analytics software

This might be that if you've set up your content in good groupings in your web analytics tool, you can do step one with step two.  But I'm going to assume you haven't.  So the best way is to work out how the pages are named in your solution.  Remember that although in Google Analytics they are named after the url of the page, in many tools they can be completely bespoke.

 

Now really what you want here is a number of metrics.  You want to see visits, Entries, Single Access (or bounce rate).  In Google you can't do that, but remember in Site Catalyst you can choose your metrics in your pages report, so I'd advise sticking all three of them in there.

This gives you a good overview of how your content was accessed and what people did with it.  Particularly of importance is the entries v visits.  If you have many more visits to entries, then that implies that previously the users found the content from elsewhere on your site.  If virtually all your visits entered the site at that page, it shows that they were on other websites when they found your article.


3a. More people found your page through your site than from other sites


This means that you need to look at the sort of pages that they got traffic from.  This is really easy in virtually all web analytics tools.  You don't need to go looking at all the paths, you just need to look at the previous page report for this page:

In Google Analytics you need to choose your page and then click on the 'Analyse' drop down to choose Navigation summary.  It'll show you how many visits came from external sources again and how many came from internal sources.  As you can see in my example, most of my visits coming internally came from one particular other bit of content.  When trying to drive traffic to a page, you need to explore the possible avenues of getting people to the page by getting those links in other bits of content. 

Can you get more links on this old page?  Just putting a new one at the end as a 'related link' might help. 

Why did the piece of content you are looking at work so well with the other bit of content?  Can you find some similar content for your new piece of content?

In SiteCatalyst of course finding out these paths is even easier.  You can click on your page name, then on paths and then previous page.

3b. More people found your page through other sites than from your site


I'm fairly confident that this is the most likely situation for most of your content.  Given now the increased use of Google, many users will just be searching for stuff and your site will pop up.  Alternatively you'll have got traffic from a news aggregator or even better some nice chap will link to you from his blog.  In this case, you really want to know what those sites are.  This is where you can use the fact that other sites have linked to you in the past to help build links for your new bit of content.  Of course, it does no harm for your research for your new bit of content if you can find related analysis on the web.

And of course if you put a link to their content in your content, it isn't going to encourage them to link back to you at all.  No siree.  Nope, not even if you send them an email and tell them.  I wouldn't advise doing that.  Or if it is a blog, don't put a comment up and tell them that you found it interesting and you've just written something else with a link.

I don't think I can stress enough that we aren't in a 'build it and they will come' world.  You've got to market your website.  If that means going out begging, stealing and borrowing links where ever you can, that is perfectly fine.  The more people who read your content, the more links you get, the better Google thinks you are an pushes you up the rankings.

If you do discover that the majority of the traffic last time was from google - look at the search terms:

 

 

Find out what sort of things that you can be writing about.  These key terms will help you pick up long tail traffic.  Especially if you link back to the content with those keywords in anchor text.  Particularly see how this fits into your keyword footprint.

4. Look at what links people click on in the content


A long time ago I wrote about how active viewer was awesome.  Well HBX may well be near on defunct, but that doesn't mean that there aren't new toys to play with.  In Omniture you have the equally brilliant clickmap.  I have to say, it'd be nice if Omniture could update this at least so that they could give it the right look and feel to fit in with v14, but it does its job still:

I have to say I like this tool better than active viewer.  Probably because Active viewer suffered from one of the issues that you ended up having the same anchor text repeatedly on a page and then it would group them all as one.  With clickmap, they describe the location of the link as well as the url, and you get individual figures.  Admittedly if you try and view it in any other format than on screen it becomes near on useless for dynamic sites and sites without a good url naming structure.

Why is this important?  Well, really this is all about 'persuasion architecture' and 'scent'.  Your pages are very much part of a journey that a user takes.  This journey doesn't begin when someone enters your site at the home page and exits having bought your product.  The journey begins at the very first tentative step of turning on the computer.  The user doesn't want their journey interrupted, but if they feel that it is there is a very easy option in the back button.

So that means that you are looking at building a page that gives the user the right set of links on the page for the place that they came from.  By analysing the links on the previous content, you can work out for that article, what type of things people were interested in.  This is key - you are trying to get your users to stay longer on the site.  They'll get a better impression of your brand and are more likely to convert (buy, sign up, etc) if they are on the site.  If the content has a high bounce rate, the users haven't seen enough to want to convert and give you money.

That back button is very powerful.  Be wary of it.  Make sure you know how you are going to market your content before you write it so that you can get the best links in it.  If you have personas this can be built in here as well.  Writing content on the web isn't just about writing good, readable content.  It's about tailoring it for a very niche audience and providing a very valuable added extra.