The most important question in web analytics is “Why?”.

The answer to every “Why?” in web analytics is also “Why?”.

This is one of the (many) great challenges in the field, but it is also what keeps the practice of web analytics fresh and interesting.  There is no end point.  There is no perfect site.  There is only “Why?”.

Having said that, there is sometimes a “Because of…” – but that usually means “Probably due to…”.

“Why?” is the most important question we can ask when a customer says “Please can we have a thousand actions [aka goals] in our account?”.  It may seem cruel not to simply acquiesce, it may seem like a blocker to progress, but there is a reason.

Web Analytics is not about the tool (though not every vendor may concede this).  It is about collecting, slicing, manipulating and attempting to understand data, and then using your findings to improve your website, your customers’ experience, your revenue and so on. The tool you use should depend on your needs, your resources and your capabilities.  If you’re tracking your blog, you don’t need an enterprise-level tool.  If you’re tracking a large online store, why would you use a “hit counter”?

While it is tempting to track everything and sort it all out later, the reality is that a large amount of the data you will collect is statistically insignificant.  Adding to this by tracking, say, every interaction with every video on a page filled with videos “because it may be useful information at some point” will surely begin to drown you in data.  And sifting through that lot will be like emptying a lake with a fork.  A fork made of paper.

Planning your implementation, understanding your reporting needs and identifying what is actionable and what is not will save you time and tears in the long run.  Of course, all the planning in the world will not prevent the collection of trivial data; it will not prevent the collection of data on non-converting window shoppers and accidental tourists and day-trippers on your site – but it will lessen it.  For everything else, there’s filtering and segmentation – so you can use your “Question Generation Machine” (as I like to call any analytics tool) to ask “Why?” and extend it by asking “From Where?”, “Who?”, “When?” etc.  By analysing your site, you can begin to reduce the instances of the simple “Why?” and start to increase the number of actionable answers and insightful questions.

As a simple example, say I want to know Why my on-site targeting isn’t working.  I’m trying to run a promotion with a certain demographic in mind and it doesn’t seem to be effective.  I decide to look at my engaged visitors for that sector – and first things first, I must decide how I define “engaged”.  While there are various definitions out there, you must also take into account your particular business and website.  For example, for metrics like Bounce Rate (a visit consisting of only one page view), the oft-accepted view is that a high bounce rate is bad – however, your focus at this time may be on a one-page microsite for a particular campaign, and in this instance a one-page visit is completely irrelevant.  So, before you worry about what your metrics are telling you, ask yourself what the metrics will mean in the right context.

Now, back to my “engaged” visitors…

I will deem an “engaged” visitor, for the purposes of this example, to be one who has returned to the site more than once and always extends their visit beyond a single page view.  They must also have triggered a particular action.  I can design a segment with these criteria (I’m using Yahoo! Web Analytics):

I can then apply that segment to my reports, compare the segment to the full dataset, customise the report, filter it further and save it for ongoing reference.

So now I have a clearer idea of who’s currently engaged with my site and I might consider being a little more even-handed with my targeting efforts if I discover that this filtered segment is quite a small portion of the full engaged segment.  It can be a grave mistake to give all statistics equal weight or to fail to break down your full dataset.

This will undoubtedly lead you down the wrong path and ultimately present you with the question “Why didn’t I analyse my analysis?”.