Google search queries data inaccurate

From the data loss in Google Webmaster Tools early June 2010 to a remarkable click growth in late June 2010 it is evident that Google has some quite notable accuracy issues.


Despite the vast amount of traffic data Google has been collecting since day 1 (or maybe because of) Google is apparently unable to produce reliable reporting in GWT. And when impressions suddenly take a sharp peak upwards in a normally dead period one it becomes evident that there is more than a ghost in the machine. Perhaps this is just the Caffeine effect kicking in, then it's a good thing I guess if it in percentage continues to generate the corresponding amount of clicks. Time and repeat validation will tell. Take the GWT numbers with a huge pinch of salt!

Lose some

Early June 2010 a sudden loss of data occured in the reporting of search queries in Google Webmaster Tools (GWT). Search results hit a hard incorrect zero in GWT (given the evidence in web analytics tool). Where that data went and how it was found no one knows, but it was solved.

Gain some

Amazing click increase during summer declineSo when the opposite happened in late June 2010 it was no surprise. During summer traffic to this web site drops every year as does the number of clicks in SERP's.

As a result any kind of growth during the period is remarkable and renders attention. Using the functional feature of starred queries makes pinpointing the 2 search phrases that are driving the inflation of the search queries quite easy.

Just 2 words behind clicks going ballisticThe rather steep increase of search phrase impressions makes the 2 words appear very close to flatliners when looking at their previous results in June 2010.

Since it is impossible to validate the GWT impression numbers, a combination of pure gut feeling and common knowledge of summer decline points towards a simple the conclusion. That is that these numbers are not as reliable as one could wish.

293 measured by proper WA vs 210 by GWTHowever the number of clicks from the SERP's can and should be validated against other sources. It's an easy process to use web analytics tool as well as use logfiles.

Anomalies are easy to spot in the search queries reports, as an example look at the number of displayed impressions of the starred report. If 2 selected search phrases are chosen, the report states that 1.000 impressions are displayed. But 590 + 480 = 1070. Odd math due to "Caffeine" rush?

According to GWT there where a total of 210 clicks that occured during the full month of June 2010, however the web analytics tool reports 293 (see image above). This indicates that GWT in this specific case only detects ~72% of the clicks, and with that kind of gap the numbers are merely estimates.

Google Webmaster Tools clicks vs. real clicks

Webmaster tool click comparisonThe combined number of clicks for the 2 search phrases struck by GWT inflation are 44, and breaking down the click count on a per word basis reveals even more differences.

The word "webbanalys" was clicked on 37 times (GWT: 22) and "webbstatistik" 26 times (GWT: 22). And the 22 + 22 = 46 result as per the displayed count in GWT only cements the conclusion that GWT can't count.

Even the average position in SERP's is unreliable, but that's another story.

External link: to more information:

Google new search index: Caffeine