You make an excellent point, that probably wasn’t very clear in the post – the analysis wasn’t intended to replace traffic analysis tools or logs for own website. That really would be overengineering the solution completely.

However, to get a feel for which of the providers of web analytics were saying something useful, and which were saying something useful that was different from all the others, you need to apply a mutlivariate regression approach. In a multivariate approach, each variable only contributes to the total regression what it can add in terms of new information. That’s quite an interesting result in itself.

But where the tool may have some practical value, is in understanding a competing or complementary site. I hadn’t thought about this while I was doing the work, but it can provide a better, bigger, more transparent window into the success of another sight other than just looking at a single metric. Of course, another excellent source for this is the SEOmoz pagestrength tool.

Thanks for the post and the details results. I’ve been through most of this at uni, but it’s great to see it applied in practice. This analysis is all fine and well and interesting, but you don’t surely expect people to collect all the data from the various free sources of web ratings to plug into a model to estimate their web-traffic? It seems like a lot of hard work when they could just use google analytics (http://www.google.com/analytics) to get accurate traffic data?