One of the endless issues we have at Narrowstep are the metrics we provide using our statserver product. Inevitably, customers compare this to stats they obtain using traditional web metric products, and I suspect most, if not all of our competitors use third party systems or metrics.
Statserver accurately measures every stream delivered from our servers - we've found the logs from Windows Media Servers to be highly inaccurate, so we trust our own figures and are an ABCe Associate (in fact their only online tv associate).
But the trouble is, measurement on the web is an art not a science. At my previous company, I architected our own metrics package. However, third part packages such as Web Trends delivered higher stats, so customers would tend to use these (go figure...). It's the size 8 argument (make your size 12 a size 8 and women will be flattered, and therefore will buy your clothes).
This article brings to a head what a can of worms this whole area of online business is. Hopefully it will bring truer accountability to the web industry as a whole.
But the reality is that, if you applied true metrics to traditional TV (cablecos and some satellite providers can actually already do this, but choose not reveal the results), actual figures would be fraction of what they are. A real world study in The Netherlands actually showed that time you were most likely to get a viewer watching an ad was between midnight and 1 am!