: views from the Hill

Wednesday, December 13, 2006

[URL] Boy stuff. Girl stuff. Young stuff. Old stuff. Really!

Ev pointed me to Microsoft adCenter Labs which has this nifty beta tool, Demographics Prediction that alleges to predict "a customer's age, gender, and other demographic information according to his or her online behavior—that is, from search queries and webpage views."

Want to take it for a spin? Pop in a search or a URL and well, here, let me show you.

Say someone heads off to Miss Snark's blog: http://misssnark.blogspot.com/. What does that foray out on the Web tell the predictors at the Labs about the person?

Gender Unknown with following probability:

:0.51 male
:0.49 female
Age: <18 Oriented with following distribution: [distribution set out]

Seems the usual <18 demographic should clock in at 9.8%, but for Miss Snark clocks in at 25.06%, the highest percentage of any of the age groups.

Try another: Preditors & Editors: http://www.anotherealm.com/prededitors/.

:0.44 male
:0.56 female
Age: 18-24. but hard to call, really. the <18 demographic clocks in at 20.34%, the 18-24 at 21.22%, well, you get the idea. The distribution is pretty even.

So what does this all do for you? If you're someone looking to place ads for guy stuff aimed at the 18-24 age bracket, it behooves you to put those ads on Web sites that attract 18-24 year old guys.

Ev has been playing around to see how wildly he can skew the results by searching for things that might be obviously girl stuff or boy stuff. His most skewed was "prom dress" which clocked in with 82% female.

I searched /brangelina/ and /brad pitt/ and /angelina jolie/. Amazed I was to find out that the /brad pitt/ search was a more-likely female search and the /angelina jolie/ search was a more-likely male search.

/ethnic beads/ came up 83% female.

Lars mentions in the comments tail that a search for "Ruby on Rails" is 93% male.


[ref: evhead is tinkering]

No comments: