Lies, Damned Lies and Web Analytics

Measurable results have been the key reason for the incredible growth of web marketing over the last decade.  That is what has separated web advertising from traditional media. Yet I find that the first task I have to do for almost every new client is to put in place a proper measurement system.

Site owners know full well when their site isn't working.  The sales graph is flat and the lead pipeline empty. Finding the reasons behind the failure is much more difficult.  The most frequent reaction is that the site does not appear in relevant search results.  So the cry goes up: "we need SEO, Search Engine Optimisation".  But, for most clients, that is likely to prove only one  part of the story.

Web analytics programmes are there to help you understand what is happening on your site.  They tell you who is visiting, where they are coming from, and what they are doing on the site......or do they?  Why is it that 2 analytics programmes can give very different data about the same site?  Even if you look at the same supplier; why does Google Analytics data not always give the same result as Google AdWords?

There are 2 fundamental problem areas: 1) the software design and 2) the limitations of the internet.  These are then further complicated by the differing needs of individual users of the data and the complicated patterns of use of the normal surfer.

Programmers naturally design their software to give what they believe to be the most relevant answers. So one might define a visit to a site as time to at least allow the page to load fully in the browser. His partner might work on the basis that one click on a link should be counted a visit even if the surfer hit another link immediately after.  There is a basic difference in the information that can be collected by log files on the server and what is seen by an ASP service recognising special code on the page. The former will see the traffic of search engine spiders, the latter will be more effective in recognising repeat visitors.

Analytics programs have a basic problem in identifying the number of visitors as against the number of visits.  The marketer will want to know how many people have been attracted to the site. But if 2 people use the same computer there is no way of telling who the user was.  And, if on a dial up connection, there may be no way of identifying it even as the same computer.

If the marketer wants data about the people visiting, the webmaster may be more concerned about the activities on the site and the amount of data the server is able to process quickly. So the latter wants to look at total hits and bandwidth which are irrelevant to marketing.

Then the human surfer rarely follows a totally logical pattern.  Looking at the web statistics, you may think you see approximately the number of visitors and where they come from; but you are seeing only part of the total pattern.  Your visitor may arrive as the result of a personal recommendation and yet still use a search engine to find you.  Do you count him as a referral or a search engine success? 

So use your data with a healthy disrespect.  Changes and trends should be reliable if the data is collected in the same way but the absolute figures are suspect.  Check your information against other data sources, particularly the contact you have with any enquirers.  Use the information to build a hypothesis that you can test by changing the site so as to measure the difference and gradually improve the performance.