Google Webmaster Tools SERP Clickthrough Data Unreliable

Last month, Google significantly revamped the “Top Search Queries” section of Webmaster Tools which put some very interesting data at the fingertips of webmasters and search marketers. It shows the keywords and keyphrases for which a site has been returned in Google search, the number of impressions (the number of times a SERP with a result for your site is displayed), the number of clicks through to your site, and the percentage clickthrough rate (impressions divided by clicks). This data can also be segmented by time period.

The data was actually of limited use and easy to misinterpret, because it was difficult to see how the site ranked for each keyphrase (and ranking drastically effects both Impressions and CTR). It appears that Google listened to feedback regarding this, as on Monday they added an additional column which shows the average position of the site for a keyphrase over the selected period. As before, this can further be broken down to see performance at different positions in search if a page’s ranking has changed over the selected time period:

google webmaster tools data

However it appears there are some serious bugs with this average rank data, particularly for smaller sites. When ordering keyphrases by position, there are many clearly erroneous keyphrases included reporting a high (first page) rank for the site.

One of our new SEO clients gave permission for us to use data for their site to create this blog article.

A significant proportion of the keyphrases reporting a position of “1.0”, “2.0” or “3.0” are erroneous.

The search for “email list cleaner” does not seem even remotely topically related to our client’s site and we found no pages for our client’s site returned for this search:

Similarly with “commercial bathroom cleaning” – this is not a service advertised by our client and there are no pages from the site returned for this search.

Finally, the supposed position one for “nude house cleaners” is particularly unusual as the site does not contain the word “nude” in content (or in backlinks for that matter)

It seems that Google clearly have a few gremlins in the works at the moment. This is not the first time we have found GWT data to be unreliable but previously this has only really been apparent via comparison with analytics data.

The data for average position appears to be somewhat more reliable for sites seeing a lot of traffic, though in these cases useful data can easily be obscured by literally hundreds of obscure 1.0, 2.0 and 3.0 rank keyphrases with less than ten impressions. It would certainly be of use to be able to filter this data before exporting to a spreadsheet.

Share this Post: Facebook Twitter Pinterest Google Plus StumbleUpon Reddit RSS Email

4 Comments

  1. Not surprised to see this, some of the results I’ve seen look a bit dubious too (though none quite as bad as this).

    However I’m slightly mystified as to why people are paying so much notice to it anyway as there is one very obvious fact which makes the data difficult to interpret – the positions quoted by Google are for any inclusion, not merely for organic. They could be Product, Local, Videos or Image, so without doing a lot of additional checking you don’t really have a clear idea of what’s going on.

    The data is of some use, and can suggest areas to explore further, but it’s limited and can easily be misleading. And it’s certainly not a replacement for organic serps results.

  2. As with all of Google’s free tools, you need to take these stats with a massive pinch (oh, ok, bucketload) of salt.

    These do seem like very strange errors though.

  3. There’s actually a reasonable explanation for most of this. The searches can be conducted with any of the various options enabled, which means results can easily produce fairly unique SERPs. The “nude” result seems to have cropped up when safe search options are on.

    So, Google ignores the “unsafe” word. The site in question could be in first position for this search under those circumstances.

    It’s also worthwhile ensuring you restrict options like “the web” when analysing the data, as otherwise there may be image searches etc. which can present seemingly impossible results, but in reality, it’s almost always a by-product of the number of search options and customisations available to users.

    Remember that the data has to come from somewhere 😉

  4. As good as the new interface/tool seems, I am not yet sure that they display the correct figures in regards to geo-location.
    But it is a good start and I believe they could improve it and be able to provide the exact figures for the users according to their search locations/country.

Leave a Comment

nineteen − fifteen =