The four types of web analytics Presentations

I've just been asked to do a presentation at the end of the week that made me realise that there are four types of web analytics presentations.  I thought I'd run through the four types and give my hints and tips (not that I am expert or anything).  I also thought it would be interesting to talk about it, because it will give some insight into the different stages of the job.

1. What is web analytics?

This will be a common one (I would have thought) across the industry.  "What the hell do you do?" seems to be the burning question on everyone's lips.  This is especially common at the beginning of life at your company.  Nobody has been doing analytics before, everyone is kind of wondering why they've brought (bought?) someone in to come and do something that they didn't need before anyway.  This is the one that needs to capture a bit of imagination.  I always think this falls into a couple of groups:

• Brief 5 minutes in front of 50 people - this is almost a pitch to your peers to make use of your time by describing how you can help them do their jobs.  These are difficult - you need to be interesting, you need to get the point across and you need to make sure you don't generate too much work for yourself.  Tips for this sort of presentation - talk about the process being a cycle.  Get your best picture of a circle and talk through the optimisation process - have a look at Brian Whaley's post (it's about development, but the same cycle applies).  This will encourage everyone to realise they aren't at the beginning of the process and you won't have to deal with them all at the same time.
• Longer (10 - 20 minutes) to a smaller group explaining why you are worth investing money in.  These are more difficult.  You have to be able to clearly explain why the work that you are doing is going to make them more money.  They are probably the ones who are going to have to spend the money to accumulate it in the rest of the Business.  Yes, yes, I know it is valuable.  Yes, I know it is fairly obvious.  But this hasn't always been taken for granted.  Here you need to be very clear about the role and define what you are going to do.  Explain how the campaigns that are making money at the moment could make more money.  These are good places to point out spend on marketing (particularly online).  Put graphs in showing the increasing spend.  Research well.  Don't have too many slides and fancy presentations.
• Longer versions, but to a bigger group explaining where it all came from.  I've done this before - it is fairly dull.  Nobody cares that we've only been doing it in total for 10 years.  Nobody really cares about log file analysis from the early 2000s.  Get audience participation. Get them to vote - give them examples of how the system used to be that whoever shouted loudest got the job done, but now you can test your theories with proper figures.  This is a real - lets get interested in numbers and show that really we're too close to the product to make decisions, so we have to be use centric.  Lots of slides.  Lots of pictures.  Pictures of babies I always find work well (everyone goes 'ahhh').

The key to these presentations is that you need to get across that the numbers will help with the decisions that they are going to make.  All you are doing is helping them make sense of those numbers.


2. This is an analysis of your site:


These seem to be the most common ones for me at the moment - which is good.  I like these presentations - usually you are showing them something that is fairly straightforward, but from an angle that they haven't previously thought of.  I always think these presentations should have a bit of a format and it goes a bit like this:

• General traffic figures
• What has worked well
• What isn't working so well
• Recommendations

The first three of these are really good, because your presentation should be very basic.  Get a slide with a graph on it and one or two bullet points.

These graphs should be simple graphs, that they could replicate themselves if they logged into the analytics package.  I'm not suggesting that you won't be believed, but if someone wanted to check again in the future, you should have clearly labelled what is in the graph.

Recommendations are the key to this type of presentation.  The outcome of the analysis should be actionable, so giving your recommendations on what you would do to improve the site and further the business should come.   I like to put my recommendations in my bullet points as I go through and then summarise at the end.


3. This is how you use your analytics tool:


Not really a presentation as such, but sometimes I like to give it as one.  This is a situation where you're having to teach a group of people how to use the web analytics tool.  It's usually not as simple as that - it's fairly easy to use the tool, what you are doing is teaching them how to interpret the results and what to do with it when they've got the data.

I always think that these presentations are completely different.  Why?  Because your teaching as opposed to presenting, mainly.  And this gives you the big difference - they are going to have to go away and do it for themselves.  That means that everything you do has to be very simplistic, step by step co-ordinated and have an example for use as well.  I find that at the end of these presentations people as for the presentation, so you have to clearly explain how you got through each step.

A good point to start in these presentations is a standard set of definitions.  These should always relate to firstly what people see them as (eg a visitor is a person) and also what it really relates to in the real world (its a browser with a cookie on it).  This starting point gives the presentees (is that a real word?) a place to go back to if they are trying to do something.  Whilst you may mock and suggest that everyone knows what a page view and a visit are, they may not know the terms that you have created for your campaign responses, what bounce rate is and what conversions relate to.

Another thing I have found that people say is useful is either a description on how to get to a report, or an image showing it so that the user can find the report that you've suggested they look at really easily:

This report doesn't need to contain recommendations - it should be as concise as possible, but remember that it is going to be passed around afterwards, possibly to people who weren't at the original presentation.  Include your name and phone number so that you can be contacted afterwards.

4. This is your traffic:


This is the most recent type of presentation that I've been asked to do (I keep thinking for the first time, but I'm sure I've done so before).  There is going to be a big gathering of people from a department and they want you to tell everyone how it's going.  Very general, not easy to do.  The keyword for this one is Charisma.

This is where you need to have jokes lined up, interesting facts and something that is slightly different.  A unique selling point for your presentation is probably a key.  I like to use running gags, so I try to use something right at the beginning of these that I can then use throughout the presentation.  Liken your first set of graphs, results, etc to something that you can then use throughout.  Abstract metaphors can sometimes work. 

The other thing that you need to really consider here is the audience.  Who are they?  What do they do?  What can you do to show them what they do is really important and that you are monitoring it.

Another thing to put in these presentations can be some of the challenges that you or the business is going to be facing.  This could get your audience thinking about how they can help you overcome these challenges.  We'll find out later in the week whether I know what I am talking about if it all goes well!!!

Log file analysis is more important than ever

I was going to write this post a couple of weeks ago when I stumbled across Paul Holstein's post on 404 error pages and how to find them.  Then I noticed from the comments on that post that June Dershewitz had also written a post on how to find broken links.  And then, to cap it all off, I have been doing some work on log file analysis for a slightly different reason, so I thought it was just about time that I wrote about it and described what you could do and what you couldn't do.

What are log files?

When your browser requests a page from a website, the server that you are requesting it from can keep a log of your request.  It will log everything that you request from the page: the page itself, the images, the css files, the javascript files, etc.

What do log files look like?

Well effectively they are just long text files with a load of information in them.  You have a line for each request from the server.  They look a bit like the one below (although you obviously can't see all of it).

What information is contained in a log file?

Log files can contain lots of information, including anything that is included in your browsers header or that is on the server.  Here are the important ones:

• The file requested
• The users IP address
• The users useragent (browser and Operating system combination)
• The users referer for that file (the page that the file was loaded from)
• The status of the request (was it successful - 200, was it a failure - 404, or was it redirected to another place - 301, for example)
• Any cookies that the server has given the user

How do log files differ from javascript page tags?

HBX, Google Analytics, SiteCatalyst, etc all run on javascript page tags.  When the user loads the page they load the javascript.  This javascript in turn loads from the suppliers server a file that contains all the information about the user (it is collected from the parameters in the tag on the page).  This means that if you don't run javascript, you won't be included in the data that the third party supplier is giving you.

Similarly, log file analysis suffers from caching issues.  That is, your browser (or your ISPs server) may store a page that you have viewed, so that next time you view it the browser doesn't request the page from the website, but can just request it from its own memory.  This means that you may miss out on lots of page views and visits if you are only looking at what is requested from the server.
 
What are the advantages of log files?

Log files offer one major advantage over normal page tagging solutions - they measure things that don't run javascript.  This means that you get lots of lovely information about robots and spiders, which crawl all our pages on a regular basis.

What can you use log files for?

404 Error pages:
As Paul and June mention in their blog posts, one of the most common uses for log files is to search out 404 errors.  What you do is search your log files for any instances where someone tried to request a page, but failed to get it (status code 404).  This will lead you to be able to see the page that the user was requesting and why they didn't get it.  It will also show you the page the user was on when they clicked on the link to get to that page, so you can check if the link is broken.

Files not being found:
Log files aren't limited to pages, they also contain information on every file that has been requested, so this can be a useful way of finding out if you are linking to any broken images, css files and RSS feeds (again from the 404 errors).  This could be causing the user experience to be slower or indeed compromised.

Search engine Spiders:
Search engine spiders will regularly crawl the site looking for new content and links to other content.  If your pages aren't being indexed in Google or Yahoo! and you want to find out why, you can look in your log files to see if those pages have been crawled by the search engine.  If they have, is there another technical reason (eg nofollow, noindex, etc)?  If they haven't been crawled, why haven't they been crawled (robots.txt, lack of links, lack of sitemap, etc).

Rogue Automated Robots:
If your site is being hit by a DNS attack (where someone hammers your site so that it breaks the servers), you can find a lot of information out about them by looking in the log files.  This could lead you to block them from accessing the site on the basis of their IP address.  It could also help you filter out from your third party systems objects which are skewing your stats.

From log file analysis we have often found out several errors on the site that weren't being picked up by users, but were being picked up by search engine spiders.  We can then alter the site in a way that the user won't notice the difference, but that the search engine will get a cleaner site and won't get confused.  We have also used it to pick up broken links that were never clicked on by users, but were fetched by search engine spiders and were losing us link juice.

There are many ways of analysing your log files.  One popular way is to download a free tool (like AWStats) and then run your log files through the system.  They will then be processed and given back to you in a dashboard.  Note that this can take serious time.  Your log files are BIG.  They contain every file that was downloaded from your server.  That will be a lot of files.  Probably 10+ for each page.  If you servers are load balanced, you may have more than one file.  Sometimes it may just be a bit easier to stick them into text pad and do it yourself by searching for the 404 errors (or writing a little macro that can split them out). 

Microsoft Analytics has had an overhaul

Actually, the headline is a bit of an exaggeration.  Microsoft Analytics hasn't really had an overhaul, but has had a bit of a revamp.  You may remember my previous posts on the subject where I first reviewed Microsoft Adcenter Analytics compared Microsoft Analytics with Google Analytics.  Well now the kind people at Microsoft have decided to update the functionality and features, so I thought that I would run through them for your benefit (although you can view them with a different slant on the Microsoft adCenter Analytics blog).

I suppose the biggest change of all has to be the addition of the dashboards feature.  Something that you probably need to think about first when it comes to analytics (because that is the bit that encourages you to delve deeper down).  And with the new MS interface, it is fully customisable (which is a bit of a step up from Google Analytics).

Well, I say it is fully customisable, it is just about that way.  You have a preset set of dashboards that you can add and move around anyway you want (by clicking on the 'show gadget library' at the top of the page, which is not immediately obvious).  I'm hoping that they introduce a feature similar to HBX where you can add any particular report to a dashboard.

The next area of improvement is the introduction of path reports.  These are very clever as they will allow you to work out how people navigate through your site.  More importantly, they'll allow you to do a couple of things:

• Work out where people are going to from a particular entry page (this is good for not only next page reports, but working out where they go after that - it's all about persuasion architecture)
• It will allow you to work out how people are navigating through your funnel to see if the process is obvious.  In blogs we don't really have funnels, but this isn't the case in commerce sites where there is a check out process
• For a frequently viewed page that isn't high on the entry page list (see the next bit) then you'll be able to see how people are getting there (this should enable you to work out why and whether you want them to or not).

Entry page reports are some of the most useful in the business.  They tell you where your users are arriving at your site.  The report that Microsoft have added would have been a bit more useful if they'd had the bounce rate on this as well (although you can still get that info from the pages view, which breaks your content down into groups based on the url structure).  You can of course, get the same exit page reports.

The main USP of Microsoft Analytics was always that they could conjure up all this data about you from your .net passport.  And this has now been introduced into more of the graphical reports.  Last time I noted that I like the tree map view as a new visualisation technique.  Now they've added the ability to segment by your tree map view by each of these segment (age, sex, occupation).  There does still seem to be a lot users on my site that aren't logged in (see the chart below), but I think having this data does give you that extra little bit.

Overall a few nice improvements that they've made.  I'm looking for them to continue to update this to try and compete with Google.  I do also think that they'd benefit from encouraging people to blog about using it - there are hundreds (if not thousands) of Google Analytics blogs, but few about MS Analytics (bar their own).  Don't ask me how they do this (maybe they need to employ an evangalist like Avanish Kaushik did for Google).