05 May 2010

The (non)effects of the browser choice screen

Two months have passed since Microsoft released the browser choice screen for European users of Windows and so far it hasn’t made a real difference in the market share. The trends established in the previous months continue largely unchanged, with , and Safari stagnating, steadily on the rise and Internet Explorer slipping to under 60% for the first time in a very long time. I didn’t expect much of an effect from the choice screen to begin with. Since Internet Explorer is associated with Windows, the operating system on their computer, less technical people will probably choose it when faced with that screen. After all, the two products are from the same company, so they must work best together, right?

Like in other cases, browser makers have rushed to report record download numbers originating form the choice screen. This time it was Opera who announced they doubled the download numbers in Europe after the Choice Screen rolled out. I have always been skeptical of such statistics, because they don’t translate very well into actual usage for the browser. To illustrate this with an example from the blogging world, a download is the equivalent of a new visit on a site, while regular usage is like having a loyal reader base, with returning users and subscribers. And we all know first-time visitors don’t convert to subscribers that easily.

The market share numbers for Europe from StatCounter simply don’t show any noticeable change for Opera in March or April. The same holds true for the individual countries mentioned as having impressive growth in download numbers. I have noticed though an interesting phenomenon in Poland, at least in the StatCounter data: Opera has an unusual spike at the end of March, jumping from 12,3% to almost 18% in a matter of days and then returning to the previous values.Browser market share in Poland 2010

At first, it looks like people (Firefox users, to be exact) are testing Opera and then abandoning it for their previous choice. On the other hand, it could be a glitch in the data collection or a selection effect caused by the sample the company is using to determine the market share. Other data collection companies don’t show the same pattern for Opera in that time period.Browser versions stats for Poland 2010

Examining the breakdown by browser versions, the graph shows that, contrary to one would expect, the older version (10.0) is responsible for the spike. It starts to loose share about a week before, which goes hand in hand with the increase for the new release, 10.5. So are the users trying the new version and then returning to the previous one? But Opera can ‘disguise’ itself as another browser pretty easily, so maybe users have tried showing websites they were visiting with Opera, but switched back to the fake Firefox identity because of rendering problems… Both explanations seems stretched given the large spike (almost 50% above the baseline!), so I think the most likely cause is simply a data artifact.

Regardless of this small Polish mystery, it can be safe to say a download doesn't make a regular user. Sometimes I download software I plan to test, but end up deleting the setup file. That constitutes a fake user for the application if they use download numbers as a metric for popularity.

Post a Comment