- April 7, 2010

Unified Digital Measurement™: Not Just Another Pretty (Hybrid) Face

By now, many in the digital space have heard of Comscore’s Media Metrix 360 initiative. This represented a major enhancement to Comscore’s original panel-only audience measurement approach, where we blend server side data with our panel data to create a more holistic, comprehensive and granular view of the digital landscape. What might not be so familiar is the name of the methodology we’ve developed to enable this innovation, called Unified Digital Measurement™ (or, UDM for short.) UDM is a measurement philosophy that will guide the formulation of our products as we advance into the future of digital media. In a nutshell, it is unique and powerful because it forges the server and panel data together at an organic level, with the goal of measuring audiences that reflects the fluid way that consumers use the internet today - across devices, platforms, browsers, machines and locations. It works equally well for large sites as well as it does for smaller more fragmented entities, resulting in a more comprehensive, more precise view of how consumers use the internet in an increasingly complex digital world. That is important to our clients, to the industry, and arguably to the digital economy.

From a business standpoint, Unified Digital Measurement gives clients, for the first time, the opportunity to have internal numbers that are fully reconciled with external audience numbers. Consistency is important. Gone are the discrepancies and the endless debate about which is right or wrong. The usage levels tie. Web site audience reach is fully explained and consistent.

When we announced this back in June 2009 of last year, we referred to our approach as ‘a panel-centric hybrid.’ While we were aware that we weren’t the first to coin the term ‘hybrid’, it seemed like a relatively good descriptor - using server data for what it does best (measure total tonnage), panel data for what it does best (measure consumer engagement and demographics), and blending them together at a very granular level to provide more precise audience measurement. This has been a huge initiative for Comscore, and we’ve learned a lot in the process.

One of my focus areas at Comscore is Global Development. In that capacity, I’ve talked to companies in different countries during the last year who also say they have hybrid approaches to audience measurement. Here’s what I’ve learned: Hybrid is the new black. In some European countries in particular, hybrid approaches to measurement have been in use for over 5 years. They are all somewhat different from one another. But they are also all very different from our ‘panel-centric hybrid’ approach and none can match the quality and accuracy of our approach and the rigor with which we produce our data. So different, in fact, that we think it’s important to separate ours from the pack. That’s why we’re branding it Unified Digital Measurement, or UDM.

How is UDM different, and why does it matter?
One reaction might be, “Great, just what the world needs: another three-letter acronym.” But with the goal of delivering the most precise audience metrics possible, it’s important to understand why these models are NOT interchangeable, and to have nomenclature that reflects that.

Outside of UDM, which is a unique approach, there seem to be two main categories of hybrids - one can be classified as the ‘cut and paste’ hybrid, and the other is the ‘cookie deletion’ hybrid.

  • The ‘cut and paste’ hybrid is where metrics from panel and server data are derived independently, and rendered on the same report. There is no analytical link between the two data sources. This approach can be found in both the US and internationally.
  • The ‘cookie deletion’ hybrid is where cookie deletion estimation factors are derived from either panel data or cookie data, and applied to server data. These ‘panels’ come from a variety of places. Exact sources are typically not revealed, but they include toolbar data, ISP data and other unnamed sources. In some cases, cookie data from server logs is used to make these adjustments.

In the ‘cut and paste’ approach, there is no analytical link between the panel and server data. If the panel is representative of the universe (which is easier said than done) this approach would theoretically provide a reasonable estimate of demographics. But even if that were the case, it does nothing to adjust for all the many other reasons for differences between unique cookies and unique visitors.

As for the cookie deletion hybrid, it’s no secret that both toolbar panels and ISP panels have major issues in the representivity department. That’s a big problem in and of itself. UDM has a big advantage here because of the effort we put into recruiting and maintaining truly representative panels - an important differentiator. . But there’s more. Our research on this issue has revealed that cookie deletion itself only addresses about 1/3 of the difference that needs to be addressed in reconciling unique cookies with unique visitors –the rest is unaccounted for in these models.

“What else is there?” you might ask. Well, there is a lot.

For perspective, we see about 1.5 billion unique cookies in a month in the US as compared to roughly 200 million unique Internet users. UDM addresses a whole cadre of additional dynamics that are important when it comes to reconciling those two very different metrics and accurately estimating unique visitors (people-based) for a given site. Examples include:

  • Multiple users per machine (within households or in public places, like internet cafes)
  • Multiple machines per person (people use different machines within a household)
  • Multiple locations (people access the internet both at home and at work)
  • Multiple browser usage (Chrome and Explorer are open at the same time--that’s two cookies but one person
  • Mobile usage

In addition. UDM includes a rigorous auditing process where panel and server data are used to ‘check and balance’ one another, and to ensure that pages are counted the same way across all sites. Examples:

  • In the server data, we sometimes see multiple beacons fired from the same page, but from the panel data, we can see there was only one page.
  • We sometimes see pages in the panel that are not being counted in the server data
  • Both sources can be affected by non-human traffic, which must be removed

Our approach delivers objective and consistent third party validation of audience metrics, which ultimately plays a critical role in improving the functioning of the digital economy

So when our clients log onto our interface on Friday, they will notice that the nomenclature in our reports indicating when a given site began being reported using our new methodology is changing from an ‘H’ to a ‘U.’ This might seem like a trivial change, but there are weak hybrids and even weaker hybrids, and then there’s UDM™. The ‘U’ stands for something: Unified Digital Measurement, which is uniquely different and superior.

More about

Audience Analytics