GA contradicts itself…
One should always bear in mind that correspondong data from different sources won’t always match, but one might hope that data from the same source but arrived at in different ways would still match. Like most analytics tools Google Analytics isn’t perfect, however by any standard what follows appears to be odd.
I’ve seen it before in GA… but in this example a filter was applied in the ‘entry pages’ report using /product/ and trending by month over the past 2 years. Looking at the final month – June 2010 – the volume of entries against these pages seemed a bit low. Shortening the time frame showing only showing June 2010 the data seemed much more in keeping with what would be expected. The point is GA appears to be producing two completely different figures for June 2010 with the only differentiating factor being the time frame of the report.
Concerned about this the report were then segmented using All, New and Returning default segments i.e. not custom built segments and still retaining the /product/ filter in the report. The results looked like this:
June 2010 only
Jan 2008 to June 2010
Although the two screen shots above don’t show June 2010 comparative data it’s still possible to see that in there is something odd going on in the second of the two reports. It should not be possible for All Visits to be significantly less than combined visits from New and Returning visitors.
I can see no reason why it should do this and GA provides no warning in the rport to indicate what it is doing – as it does with sampled data.
The only route to resolve is to shorten the time frame over which the report is run (trial and error unfortunately) and see how far one can get before the data starts to go wonky. You need a good overall knowldge of what to expect from you data in order to spot this.
