After his breathless article last year, proclaiming Google Analytics to be something like a cross between the second coming and Barack Obama, Brandt Dainow seems to have soured on the big G, proclaiming this week that GA contains ‘disturbing inaccuracies’:
Google Analytics is different from other products in that it has been intentionally designed by Google to be inaccurate over and above the normal inaccuracies that are inevitable. These inaccuracies are so glaring that most people are getting a very false picture of what is happening in their sites.
Dainow’s main beef with GA is two-fold:
- It treats single-page visit as valid visits (i.e. it doesn’t remove them from visit counts or other related measures)
- It includes single-page visits in average visit duration calculations
He also remarks that Google did in fact change the way that GA calculated average visit duration last year, but then changed the calculation back in the face of user pressure:
Google intentionally rolled Google Analytics back so that it produced an incorrect average duration…It’s been that way ever since — Google is intentionally and knowingly providing inaccurate numbers because a few people preferred neatness to truth.
Brandt then proposes two alternative measures – ‘retained visits’ (the count of visits with more than one page impression) and ‘true average duration’ (the average duration of retained visits). These metrics are not without some merit – it’s useful to know how many visits contained more than one page view, and the average duration of these visits. But Brandt goes on to assert that these two metrics should replace the standard measurements of visits and average duration in GA and (presumably) other tools. This suggestion is ridiculous, for the following reasons:
- Contrary to Brandt’s assertions, there are a host of scenarios where a single-page visit is a perfectly valid visit, including, for example, this blog, for crying out loud, which has a high proportion of single-page visits because readers either just read the homepage and leave, or click through to an article from their RSS reader. So chucking all these kinds of visits out is crazy.
- Whilst the inaccuracy of including single-page visits in average visit duration calculations is known to be a problem, removing these visits from the calculation doesn’t yield a magically ‘accurate’ number, it just yields one that is inaccurate in a different way. You still have no idea how long people looked at the final page of their visit for, and with a two-page visit this can introduce a huge potential inaccuracy.
- Such standard metrics as exist in the web analytics industry are the result of long and arduous wrangling. There are no sacred cows, but you need a really good reason to exchange a simple and easy-to-understand metric for one which is more complex and offers no discernible benefit.
Whilst I can understand Brandt’s motivations for posting these ideas (which, I imagine, lie somewhere on a spectrum between a genuine desire to spark debate and a desire to generate a lot of traffic to his blog, in which regard I am obliging him), his remarks do irk me a bit (can you tell?), principally because he commits the unpardonable sin of absolutism when talking about web analytics, bandying about words like “truth” and “wrong” when really he is just presenting his own preferences.
When, as an industry, we can’t even agree what constitutes a visit, it’s pretty rich to start decrying one tool or another as ‘inaccurate’ simply because it takes an approach to data that you don’t believe in. And besides, as Brandt surely knows, Google Analytics now has the capability (via its custom segmentation) to calculate the metrics he seeks.
Finally, as every half-experienced web practitioner (of whom Brandt seems to have a low opinion also) knows, the key to success in web analytics is to pick your metrics, stick to them, and measure them continuously as you make changes to your site and your marketing, to see what is working. If you’re looking to increase engagement, and have decided that visit duration is a good measure of this (a debatable point, as it happens), then it doesn’t matter whether you include single-page visits in your duration calculation – if your visit durations are going up, you’re happy. And if your visit durations suddenly jump because your web analytics vendor has changed the way they calculate the metric, this could in fact cause more pain than benefit, perhaps causing you to go to said vendor and say, “Oi! Change it back to how it was!”.
So feel free to read the article, but be warned: it’s not very accurate.