This is a post from Alan Salmoni, Alan joined us a few months ago as Head of Research and User Experience. Since then he has been carrying out some really great research, and we think that it’s about time that we shared some of it with you.
The showdown between keyword ranking data for Google Analytics and Google Webmaster Tools
Both Google Analytics (GA) and Google Webmaster Tools (GWT) offer information about a keyword’s rank. We had noticed some discrepancies in these data for the same keywords over the same period of time and wanted to know whether these discrepancies were real and likely to affect SEO work.
We took the top 500 (or up to 500 if there were fewer) keywords for 6 websites and compared the ranking data between GA and GWT for the period 20 June 2012 to the 20 July 2012.
Domain
|
Number of keywords
|
|
1
|
192.168.50.52
|
500
|
2
|
www.dizzyheights.com
|
500
|
3
|
www.petecowen.com
|
500
|
4
|
www.saveamillionshots.com
|
500
|
5
|
www.websiteredevelopment.com
|
432
|
6
|
www.alansalmoni.com
|
203
|
|
Total
|
2635
|
We compared each domain’s keyword rankings individually and across all domains too. Keywords were exact matches i.e., “competitor analysis template” was considered a totally different keyword to “competitor analysis templates”.
Results
We looked at the mean difference in ranks, the standard deviation, and correlated GA and GWT ranks.
Domain
|
Mean rank difference
|
Standard Deviation
|
Correlation
(Pearson’s)
|
|
1
|
192.168.50.52
|
9.95
|
13.13
|
0.99
|
2
|
www.dizzyheights.com
|
103.70
|
160.67
|
0.66
|
3
|
www.petecowen.com
|
121.67
|
41.50
|
0.61
|
4
|
www.saveamillionshots.com
|
3.74
|
7.30
|
0.9957
|
5
|
www.websiteredevelopment.com
|
123.87
|
164.08
|
0.60
|
6
|
www.alansalmoni.com
|
177.94
|
175.86
|
0.56
|
7
|
Overall
|
118.46
|
177.91
|
0.68
|
The correlations are particularly interesting because the domains are clustered into 2 kinds
1. Those with high correlations (0.99 and above) and relatively small mean rank differences (3.74 and 9.95)
2. Those with lower correlations (between 0.56 to 0.66) and much greater rank differences (from 103.70 upwards).
These results indicate that GA and GWT provide very different keyword rankings even for exactly the same keywords over exactly the same date range for the same domain. Our concerns are two-fold:
1) The correlation over all domains was 0.68. This was significant (p < 0.05) but quite low considering all data comes from the same company)
2) The two domains that showed good correlations still showed rank differences big enough to concern SEOs: the mean rank difference of 9.95 for 192.168.50.52 could mean the difference between 1st and 2nd page or 2nd to 11th rank which is a lot in SEO terms.
While Google is explicit about why the rankings are different, we feel it’s of use for SEOs to know by how much. This difference is not the odd rank here and there. We found mean rank differences in the hundreds. The single largest different was 780 and the histogram below shows how large the rank differences can be. We found several instances of a rank difference of over 30 for keywords placed in the top ten by one or other data source.
This implies that SEOs need to be extremely careful when using GA or GWT when working with keyword rankings. While Google is the closest thing to an official rank, it’s important never to assume the two data sources are always congruent, as some sites might be relatively close while others are vastly different.