Quantcast Comments on Facebook Traffic Dip

As we discussed with Om Malik yesterday, media measurement panels -- including Nielsen's, comScore's, Hitwise's and our own -- show a strong household bias.

As seen in yesterday's debate over the perceived September dip in Facebook's traffic, this can result in a dramatic undercounting of student activity once Fall hits and college students head to the dorms. 

Why?  Panels can only reflect the internet activity of panel members, which are typically households -- not workplaces, and not colleges or universities.

In August, students are still home with their parents -- plugging away on their favorite social networks and media sites.  Some of these homes are in media measurement panels.

Come September, these students go back to school and their internet activity disappears from the panel. 

Panel estimates therefore show a decline in traffic to sites like Facebook despite the fact that activity (for many student-centric sites) actually increases.

This is the problem with panel-based audience measurement. Direct measurement is simply the only way to get an accurate read of web site traffic. 

While panels provide directional guidance and insight into trends, they lack the accuracy and precision that the industry requires in order to create a viable data currency that can attract brand advertisers' dollars.

That why we created the Quantified Publisher program to provide publishers large and small with the ability to accurately measure their audiences -- free of charge.

We start with a large panel of two million internet users. We then calibrate that panel information with directly measured traffic data to provide the accurate third-party metrics and easy-to-read profiles you see labeled with this Quantified Publisher icon.

Have questions?  Add a comment to share them here or email us at contact AT quantcast DOT COM

Share on Facebook

Comments

[Ed Baker]

Comscore has a university panel. Explain that.

[Krista Thomas]

Hi Ed,

We are aware that comScore, like many other services, attempts to incorporate as many audience dimensions as possible into the estimation process.

However, no panel -- no matter how carefully constructed -- can hope to capture the full range of nuances and biases of the world's varied web site audiences (and this includes our panel).

Panels are generally good at providing broad visibility and insight into general trends and patterns, but they are weak at precisely estimating absolute numbers and can be easily confounded by unknown biases in a website's audience.

More importantly, we don’t need to rely on panels alone to get this data. The Internet enables direct measurement which can be combined with panel based research to offer a much greater level of resolution and accuracy.

In the case of Facebook, we hypothesized that the apparent dip was due to the limitations of panel based estimation processes. This would appear to fit well with Facebook's public statements relating to the matter.

See the updated GigaOm posting here: http://gigaom.com/2007/10/11/facebook-dip-seasonal/

As Om Malik notes, a Facebook spokeswoman got in touch with him from London and explained…

"It’s important to note that we focus on active monthly users rather than registered users. Active monthly users in the U.S. grew from 17.7 million in August to 19.4 million in September. This active monthly user base has been growing at a rate of about 3 percent per week since January 2007 which means we have been doubling about every six months. Globally, we ended September with 44 million active monthly users as compared to nearly 9 million active monthly users for the same month in 2006. Our page views have also been increasing and went from 54 billion page views for August to 57 billion page views in September."

[Craig Earon]

I have to agree that panel measurements are not accurate. I often check on website statistics from other marketing research companies that use panels and the results vary widely for the same website. We selected Quantcast because we found that Quantcast's direct measurement feature could easily be compared to our own website logs and the match is excellent. Whenever I see a panel measurement used elsewhere, I take that information with a grain of salt as an educated (but questionable) result.

I firmly believe that third party statistics such as those Quantcast provides produces an accurate way to compare two different websites from a marketing perspective. I know that we have received calls from advertisers that looked at our website statistics and called because the information met their customer demographics requirements. They find that the stats we provide match public information available on Quantcast. Not all websites do that, and fudge their numbers a bit if they feel they can get away with it to make a sale.