A couple of weeks ago I posted about the accuracy of the Google keyword tool. July’s figures are now available in the Keyword Tool, so I decided to run the same set of ‘100% impression share’ keywords through to see if there was any change in the accuracy demonstrated last month.

Once they have been rounded to one of Google’s established reporting integers (see my previous post), we get the following results: 

                                                                                         

Keyword	Actual   Impressions July 08	KW   Tool 'Approx Search Volume: July'	Difference   from actual	June   Difference
Generic   Keyword A	60500	18100	234%	125%
Generic   Keyword B	40500	33100	22%	22%
Generic   Brand A	22200	14800	50%	83%
Generic   Brand B	9900	6600	50%	0%
Generic   Brand C	9900	9900	0%	0%
Niche   Brand A	2400	1900	26%	50%
Niche   Brand B	1900	1300	46%	46%
Niche   Brand C	1000	880	14%	39%

So, three of the differences are the same (although ‘Generic Brand C’, which was correct last month, actually increased both its actual impressions and its KWTool result, changing figures but remaining correct) and three of the results are closer to the true value than they were last month. Two of the results are worse – one of them was actually correct last month but has increased impressions whilst the reported value decreased, and one of them was the most inaccurate anyway, and the KWTool value has just moved further in the wrong direction.

Because of the even larger difference in Generic KeyWord A’s figures, overall numerical variation has actually increased, despite a greater number of results being more accurate.

The ‘average search volume’ remained the same in all but one case - Generic Brand C, which increased to match the July, and ‘real’ figure. It’ll be interesting to see the accuracy of the ‘averages’ after a longer test period.

If anyone has any thoughts on this, or has drawn any conclusuions with the help of this data, I’d be very interested to hear them.