I spent an hour or so today playing with the data available via the Google Webmaster Tools “Top Search Queries” report. As you probably know, Google launched an enhancement to the functionality to include impressions and clicks, making a click through rate calculation vaguely possible. The tool was well received by many, albeit rather cautiously. I thought it might be fun to look at the data and draw some charts with Excel Pivot Tables, trying to answer questions like “how far do the click numbers in Webmaster Tools agree with my own figures?”.
Image credit: Hey Paul
A Word to the wise
This post isn’t designed to come to any conclusions on the quality of the data, and I’ll make my usual recommendation now: you should always perform your own analysis, on your own sites to come to your own conclusions. There are some limitations to the numbers I’ve used, the most glaringly obvious being that Google Keyword Tool data is currently exporting values for March 2010. The Google Webmaster data can only export data from mid March to Mid April. That’s probably going to impact the numbers somewhat. The other item to be aware of is this – the ranking position data does not survive the export process, meaning that if you’re not careful you could be comparing impressions data for a keyword on page 2, not page 1. Be aware!
I had quite a lot of fun collecting all the numbers for my data. In summary, I decided it might be wise to be able to narrow my values down to a specific region and compare the accuracy of the values to something more global. So all of the Google Webmaster Tools data used was either “United States” (my largest traffic generating region) or “Global”. I filtered the values down to “Web”. I’m still yet to decide if that was the right move, but that’s the point of this study! Google Analytics Google search engine visits by keyword and region was exported, following this I used both Google Keyword Tools (Beta and External) to get what (US) and global search volume data I could.
The data columns in my master Excel spreadsheet table looks like this:
GWMT US Impressions
GWMT US Clicks
GWMT All Impressions
GWMT All Clicks
Here’s a little screen grab of master data table:
This chart shows a comparison of total visits from the United States, via Google.com to the number of clicks reported by Google Webmaster Tools. The region is set to “US” in Webmaster Tools, selecting “Web” as the preferred source of click data. (Click to enlarge the chart)
The comparison data reveals variances between a minimum of 3% and a maximum of 55% on keyphrases bringing between 1236 and 31 visits over this period. What I found quite interesting was that the error didn’t increase or decrease proportionally to the volume of the query, it just seemed to be random. I’d love to see the variances on a higher volume query set. Some terms such as “html 5 examples” were very close (3%), and the average difference across the top 20 terms was 20%.
Our next chart compares the click data provided by Google and the analytics visits from Google, globally (Click to enlarge)
The range of variance between terms that drove traffic to SEOgadget globally, and the global impression figures is much larger. Variances between -163% and 75% were observed on the top 42 keywords. Certain keywords were vastly different, “how to install ubuntu” presented a -163% difference with Google Webmaster Tools wildly overestimating or underestimating the number of clicks received. Wow. The range of difference in the global data set makes the data particulary unreliable in my opinion.
Next, I took a look at a small range of keywords and compared the impressions data for page 1 rankings to Google’s Keyword Tool and the Beta tool.
The difference is rather clear. Google Webmaster impressions data is significantly lower than the values reported by the keyword tools. I’m actually of the opinion that the impressions data from Webmaster tools is better than the keyword tools data – certainly the CTR% make a lot more sense coming from the Webmaster Tools platform. If you’ve ever tried to calculate click through rates from the keyword tool data, you’ll know what I mean.
If I had more time, there are a few things I’d like to get to the bottom of with this data, in particular working on larger query volumes and more regions to see if certain geo locations offer more convincing figures than others. Remember there’s also a need to recheck these figures against a complete month in the Google Keyword Tool, although I’m not entirely convinced the percentage differences by keyword would be all that different.
Issues aside, it felt that the data fits with the analytics figures when you’re investigating clicks and impressions by region, but be prepared for a variance of anything between 3 and 55%, possibly more.