PDA

View Full Version : Google Reporting Discrepancies


pompeysie
09-02-2005, 10:36 AM
Hi

I am having a real problem trying to reconcile differences between data within Googles reporting centre and the figures from my web log files.

I am using Googles conversion scripts on a "thankyou" page on my site. This page only appears after a customer has completed purchase on the site. The script is configured to pull out the sale amount and send it back to Google as part of it's conversion reporting mechanism. On the surface all seems ok. Google has reported that in August my ads received:


Clicks = 4083
Sales Counts = 315
Sign-ups Count = 689 (from a different script on our brochure request page)


When I compare this data to my log file data (analysed using ClickTracks) I get the following:


Clicks = 4128
Sales Counts (i.e. from visitors who came from ?source=goog (my AdWords query string) and who visited my post checkout thankyou page) = 216
Sign-ups Count = 698


I think the Clicks and sign-ups count values are pretty much within error. Changing session parameters within clicktracks will vary these numbers to within this error. I realise you also have to take into account server errors etc. so am not too worried about these figures.

The big problem here, and the most crucial figure, is the Sales Counts values, 315 from Google cf. 216 from my log files. Which figure should I put more trust in? I am in the process of modifying my customer database so that it records referrer data but this wont be any good for historical data.

Has anyone expereinced anything similar?

My thoughts at the moment is that the Google data is not correct. Does anyone know what happens if a users hits the back button and returns to the page containing the Google conversion script? Will Google count this as another conversion (i.e. sale)? How about if a user hits refresh?

Any help would be much appreciated. I have been looking at this data for five hours today and am about to scream!

If you need anymore details please ask.

Many thanks

Simon

Chris_D
09-02-2005, 09:30 PM
Hi Simon,

What was the total sales count (i.e. all referrers, not just Adwords?).

Can you break down total sales count by referrer:
Adwords
Overture
Google algorithmic
Banner advert 1
etc

Is the script counting every referrer with Google on the referrer string and passing that back to Google (ie Algorithmic and Adwords?)

AussieWebmaster
09-04-2005, 01:29 AM
Simon,
Google will count the back clicks as it is the pixel on the page that is being called at their end and is counted every time.

Doomsday
09-04-2005, 02:57 AM
Hi Pompeysie,
I wonder if there is any difference in the time zone your and Adword reporting system been followed. I guess adwords gives report based California Pacific Standard Time, if there is any variation, I guess the clicks difference can be contributed to this factor. Ya the Sales Count figure are the cause of concern.

pompeysie
09-05-2005, 08:20 AM
Hi everyone. Thanks very much for your comments on my post. I think we have finally worked out the cause of the discrepancy.

It is simply a factor of the different way's that Google and ClickTracks (in this case) store and report the data.

The Sales Counts from the log files show us the number of unique visitors who made a purchase directly after clicking on one of our Google ads in August.

The Google data shows us the actual number of visitors who made a purchase on our site and who had been referred from Google anytime within the last 30 days - the length of time the Google cookie hangs around. Picture this scenario:

A visitor referred from one of our Google ads makes a purchase as a direct result of clicking on the ad. This is counted as one sales count in Google and one Google sales count in the log file (i.e. source=goog as the referrer). The same visitor returns directly to the site one week later and makes another purchase. This time he has not clicked on a Google ad though, he bookmarked the site after the first visit. This purchase is therefore counted as another sale in Google but not as being referred from Google in the web log and, as such, results in a discrepancy between the two data sources.

So the Google number will invariably be larger than that in the web logs. I am just about to run a report to see if I can track visits and purchases from users with the Google Ad words cookie on their machines. If I can, this number should be much nearer to that in Google.

Hope this makes sense. Thanks guys.

Simon