Introducing Quantcast and its mystifying demographic data source
Posted by Pat Kitano at 11/15/2006 12:21 AM 
Categories: Technology, New Business Models, Cool Web Tools
Categories: Technology, New Business Models, Cool Web Tools
Joel always finds them (thank you) and I like to add another review layer on top of them. This time he demonstrates Quantcast as a kind of Nielsen analytical tool for audience measurement of small-to-mid size websites. They present site stats with accompanying graphs illustrating demographic and psychographic profiling for the site. Joel marvels at the graph claiming 79% of Redfin viewers make more than $100k income... why are they all so well off? he ponders.
So far, so good... Quantcast has identified an advertising-underserved niche, and they say they are developing ad placement engines. They're involving these smaller websites by getting them to voluntarily "quantify" themselves (adding java script to their sites) in the greater hopes of one day being an attractive ad placement for a future micro-channeling ad buyer.
Quantcast's measurement model is analogous to the higher end Nielsen/NetRatings. Now Neilsen measures audiences the old fashioned way - they claim to have a MegaPanel of a million "trackers" with identified demographic profiles so it's easy to crunch which profile views which site. Quantcast has done away with the Panel! They don't track individual viewers... I can't figure out how they can identify that 79% of Redfin viewers make more than $100k. Are they clicking to Redfin from www.worth.com ? The press is mystified too:
"I have no idea how accurate Quantcast's estimations are, but it is an interesting concept to collect volumes of data without identifying users."
-- Marketing Shift Online Marketing blog - (interesting concept? I'd jump all over this secret sauce Mr. Journalist!)
Quantcast's press releases claim they've developed algorithms that estimate gender, age and other demographic information by aggregating data from similar websites. Yes, but how do they come up with such a refined 79% number? They're very coy about this... Marketing Shift continues:
"The company says it has rocket scientists figuring this stuff out for them, but time will tell."
hmmm...
I bet Nielsen will undertake a survey to compare the Quantcast estimates to actuals...
Technorati tags quantcast Nielsen audience measurement real estate real estate technology real estate marketing venture capital
Very interesting read.
It would seem to be nearly impossible to drill down that data without having personal identifying information - which reminds me of some of the data collection methods used in the last election.
Very big brother.
Reply to this
Hi Doug,
Welcome to the world of blogging! I just noticed your new site... I like how the content stretches out across my full notebook screen and how the blog looks integrated with the website, may I ask which company did your web design? Your writing style parallels my views on the biz... I enjoyed your Cult of Personality and chef articles... I'll watch you on my feed and add you to my blogroll...
Reply to this
Thanks for adding me to your blogroll, Pat - as well as your kind words about my blog!
The design company is called "Me, Myself & I"... LOL - as in "I did it myself".
The reason it looks integrated is because instead of integrating a blog into a static website - I integrated a static website into a blog.
I used Wordpress with a theme called Wucoco for the layout... and customized it from there.
It's not really finished, yet, as I still have static content to write and insert... but it's getting there.
Once again, thanks for your kind words!
Reply to this
They may be doing it through IP address and Census data - but that's pretty clumsy IMO.
Reply to this
Pat, interesting company. I'm always mystified by claims like that. I've seen a bunch of sites recently that claim to be able to measure visitor worth. However beyond GEOip data, I can't figure it out either.
Reply to this
Following up on your and Joel's comment, I can only imagine this. They pull demographic data from the locality (say zip code) of the ip address. Let's say it's a zip code with 40% singles, 10% income over $100,000, 20% Hispanic. These census data determine a "probability" of Hispanicness, or any other data variable. Multiply these data points 100,000 times, weighting them accordingly (if a site gets 40% of its viewers from El Paso, Texas, there will be a Hispanic skew) and a pattern based on the probabilities will emerge.
Reply to this