Enhanced data methodology powers new features and new capabilities

Enhanced data methodology powers new features and new capabilities
Today’s release of the reinvented referral analytics on Compete.com is a milestone for one group of people here in the office:  The data team.  This is the group of white lab coat wearing, PhD having, prime number recognizing people here at Compete that are responsible for taking the massive sample from our multiple clickstream sources and applying the math that turns it from raw data into best-in-class traffic estimates.

These are the folks who come in to work every day and ask themselves how to make the numbers a little better and how to extract a little more value from them for our clients.  In the past months they have interacted with data and measurement practitioners from industry associations including the IAB, ARF and Web Analytics Association, as well as our research colleagues at other Kantar companies.  And as a result of their own work and these conversations they’ve just concluded a project to enhance Compete’s data methodology to allow us to offer new and better services, including today’s release.

How did the methodology change?
One of the key elements of the enhanced methodology is an updated Pageview definition, which more accurately delineates consumer site traffic from artificial site traffic such as distributed content, redirects and/or iframes.  In layman’s terms, this helps us count pageviews for a domain only if they take place on that domain, and not on a widget embedded somewhere else.  This updated definition provides a more representative measure of people’s actual web usage, and therefore the traffic metrics for websites that we measure.

An additional enhancement is an updated Visits definition that breaks out and report visits discreetly within individual internet sessions. This change will also enable us to more accurately reflect total visits to a site, meaning the sum of incoming referrals will equal total visits to a given site. It’s exactly this aspect of the enhanced methodology that makes our latest release, which breaks down visits by referral source and category, possible.

Who does this impact?
In general, most site owners should see very modest changes to their traffic numbers on compete.com. Because this is a change to our core methodology, almost everyone will see some change, but half of sites will vary by less than 20%, and three fourths of all sites will show increases to their monthly unique visitors.

However, approximately 1 in 10 domains will see a decrease of more than 30%.  This group includes some fairly large sites that have a large number of distributed users.  For instance, the traffic to Digg will fall considerably as we weed out the effect of the Digg bar.

Another site that sees a large change under the new methodology is Facebook.  Before this improvement Compete visitors to a third party site that used the "facebook connect" technology sometimes added a pageview to Facebook itself.  We now count those visitors only for the third party site, and the result is a drop in the overall number for Facebook, although the trend of strong growth remains unchanged.

We hope that this gives you some insight into the data work behind the products that you see on compete.com, and that our vertical teams bring to clients. Please feel free to leave a comment if you have questions, or contact us directly at support@compete.com.