in Data & Insights

Comscore Implies There’€™s No Need to Pay for Comscore Data?

Last Thursday, Comscore published a blog post entitled “Who’s Right?”. It’s a provocative read, though sadly not because it helps our industry face up to some of its most important challenges. We’ll let you draw your own conclusions (we really recommend you read the post), but we’d be remiss if we didn’t offer some commentary to better frame the issues raised by Comscore.

Before we do that, in case you don’t have time to read their post or this one, here’s a key conclusion reached by Comscore (and bear in mind, the Quantcast service is free whereas Comscore charges tens of thousands of dollars per year):

Quantcast numbers are within 3% of those sold by Comscore. Source: Comscore.

Strange that they would be so eager to highlight this in the current economic climate – if you’re negotiating a new license you might want to let your procurement people know!

This post’s objective is not to say we’re as good as Comscore – they’ve already done a rather nice job of that – but rather to emphasize that our collaborative approach is about having the whole industry participate in a way that improves transparency and accountability.

Our Quantified Publisher program enables us to collect real data from millions of distinct web destinations. We believe the industry approves of and supports this approach and it’s why leading agencies (Avenue A/Razorfish, Mediaedge:cia, and Mediavest) and publishers (NBC, CBS, Fox, Viacom, Business Week, Time Inc., Martha Stewart and over 75,000 others) choose to participate in, and use our service.

We’re excited about the progress we’ve made over the past two years. It’s a function of our progressive approach to engaging all industry participants and involving them directly in the development of solutions. We’re eager to facilitate a new wave of innovation, so please – keep challenging us with questions and asking for enhancements. It’s the only way we’ll all win.

More generally, Comscore’s post is yet another attempt at rationalizing panel-only measurement. We think panels are great too, but we see them as a starting point, especially in a digital landscape where media consumption can be measured directly (and in fact, already is). That doesn’t mean panels will become completely obsolete, but you simply can’t ignore the power of direct measurement and hybrid approaches in audience modeling – 11th century King Canute surely didn’t really believe he could stop the advancing tide from wetting his feet and robes (for more, see ‘Ruler of the Waves‘). To be fair, we don’t think Comscore really believes panels are the end point for digital media measurement – and if they did, why would they be testing a pixel based solution from

Comscore’s line is one of maintaining the status quo, perhaps because they are so heavily invested in it. Ours is challenging it and helping the advertising industry improve the accountability and actionability of audience data. You’ll notice that not once did Comscore cite an outside publisher or agency that had praise for their data, or approach to solving industry challenges. We believe that industry recognition and support is a critical measure of innovation and we’re thrilled to have people talking publicly about our solution.

The post is a compilation of slight-of-hand, unsubstantiated accusation and misdirection. It’s not aimed solely at us, Google is also mentioned, but perhaps most surprising is that it characterizes publishers as being naive by maintaining that their log file data directly represents people. Let’s be clear: major publishers don’t believe their server counts represent unique people. They, along with everyone else in the industry, understand the issues related to cookie deletion, non-acceptance, multiple machine use, and bots and spider filtering.

Cookie deletion is a problem, but it’s not an insurmountable one, and earlier this year we launched the industry’s first cookie to people translation model. Our white paper on the subject clearly explains how we address all of the issues that Comscore correctly identifies. In fact, Quantcast is the only measurement service today publishing both cookie counts and people counts for the millions of Quantified Publisher destinations we measure.

For a company that references opacity of others’ approaches, it’s puzzling that there is so little detail on the methodology behind the core comparison cited in the post. Being right, and substantiating a case that people can believe in or at least debate, requires transparency.

Unfortunately, Comscore’s analysis breaks down when held to this standard. The post states that “the average audience reported by Quantcast is 3 percent lower than the average comScore data for the same sites.”

But as a top-line statistic this masks important underlying differences in how the data might be collected. While many properties reported by Quantcast are based solely on our own panel estimates (non Quantified publishers), we have never claimed that panel-based estimates are the best way of gauging traffic. That’s why we offer the free Quantified Publisher program, which enables anyone to have their census-level traffic data incorporated into a much more robust audience modeling service. A comparison of Comscore’s panel estimates to a set of Quantcast’s directly measured Quantified Publisher people data (cookie adjusted) would be much more revealing.

There are many other important factors that potentially skew Comscore’s comparison: how were the 10,000 sites chosen, is US only data being compared, and is audience data domain level or does it roll up multiple sites to common owners?

One additional factor is the time period of analysis. Comscore cites monthly data, Quantcast measures data on a rolling 30 day period. Most sites have lots of ‘passers by‘ – meaning every day there are new unique users who extend the audience size. Many months contain 31 days, approximately 3% more time than our 30 day period. You do the math.