Ad Agency Web metrics

Ad agency stats are not comparable to our own internal tracking!

[Hinet]

We're a large corporate in the UK, and use RedSheriff as our web metrics tool. Our marketing guys are throwing questions our way as to why our Ad Agency - who arrange our banner advertising, are saying one figure (which is generally a lot higher) and our own reports are saying something totally different!

At times the difference is up to 45%

Does anyone have any clues as to why this is the case?

Have had a look in the press/web and the industry doesn't seem to have much faith in the Ad Agency metrics? Why is this, some research shows that the methodologies used vary a great deal...if you have any experience with this, would appreciate your views etc.

[lorax]

Hello Hinet - welcome to WebmasterWorld.
There are a number of reasons the numbers will vary. It is not uncommon that they will. The only way to guarantee the numbers you're looking at is to cut through the chaff and view the raw log files yourself.

Any program that is designed to read log files needs to make decisions about the data. One of the easiest examples of this is in what constitutes a session/visit. If someone visits the website today clicks a few links and leaves. That's considered a visit. If they come back again in 5 minutes, what would you consider it - another visit, the same visit? The industry default is a 30 minute absence before declaring it another visit.

And every stats analysis program is inherently faulty. The reason is that many web surfers access the web through dynamic IPs. IP addresses are used to track repeat visitors. If I use my dial-up INet connection today - I'll get an IP of 204.130.60,101. I surf and then disconnect. When I log back in the next time my IP could be 204.130.78.23. How could you know I visited your site? Cookies are the only other method you have for an anonymous tracking of visitors. But even these are not foolproof. For instance - some folks tell their web browser not to accept cookies.

In short, there are an inherent errors in each program. You need to include them in your considerations as you review the data.

All that being said, 45% does seem like a lot.

HTH