Predicting Your Gender Based On Web Surfing
MADELEINE BRAND, host:
Yesterday on the show, we directed you to our blog, Daydreaming, to take a little test. On our blog, we linked to a program that tries to figure out whether you're a man or a woman based on the websites you go to. Well, it was a spectacular failure, as far as I'm concerned, predicting with 77percent certainty that I am a man.
And it was 100-percent certain that our booker, Jolie, is a man. OK. Well, here's a real man, the creator of the program, an advertising blogger named Mike Nolet, on the line to defend himself. Hi, Mike.
Mr. MIKE NOLET (Chief Technology Officer and Cofounder, AppNexus): Hi. How are you?
BRAND: Good. Thanks. OK. So, how did you come up with this formula?
Mr. NOLET: Well, it - the formula itself is actually some just basic statistics. It's actually quite bad statistics in itself...
(Soundbite of laughter)
Mr. NOLET: Because it was actually meant more to be just kind of a proof of concept, that this is something that you can actually do.
BRAND: Or not.
Mr. NOLET: Well, or not, yeah...
(Soundbite of laugher)
Mr. NOLET: Very true. There's a couple of things missing from it. So, you know, it doesn't actually look at how often you go to websites, which obviously would have a big impact. And also to know how long ago it's been since you went to a site.
BRAND: So, Mike, how does your program work? How does it figure out whether or not I'm a man or a woman?
Mr. NOLET: So, I used a dataset from a company called Quantcast. They do sample studies to determine kind of the distribution of men to women, and also age and income, of various sites around the Internet. So...
BRAND: And how do they know?
Mr. NOLET: Some of them have a - do it through browser plug-ins, others do it through, I think, surveys. So, it'll pop up a question with a little quick interview. Others do it from registration data. So, when you sign up for the New York Times, I think they ask you your birthday and your gender.
BRAND: OK. So, they can say, through all this data, the people who log onto the New York Times, for example, are mostly male.
Mr. NOLET: Yes, in aggregate.
BRAND: OK. So, then, your program does what?
Mr. NOLET: So, imagine if you go - let's say the New York Times is 55 to 45, men to women. And you now go to the New York Times, and I know that fact. Now, I can say, with a certain accuracy, that there's a 55 percent chance of you being a man. Now, let's say you go to another 10 sites, and all of those 10 sites are 55 and then 60 and 75-percent male. So, the more sites you go to that are skewed towards men - and it can be a small skew as well - the more likely it is that you yourself are to be a man. So, let's say you go to 10 sites that are all 55 percent men, and now, you are far more likely to actually be a man.
BRAND: News sites, I suppose, would be more skewed towards men? More men read news sites?
Mr. NOLET: Yeah. Yeah, apparently.
BRAND: And more women read what?
Mr. NOLET: More women - this is the one that surprised me - is actually - if you go to all the banking sites, the Bank of America, CitiCards, they actually tend to be more women.
BRAND: So, women are doing the banking for the house.
Mr. NOLET: Apparently. That's actually one of the interesting things that surprised me when I was doing this.
BRAND: So, now, you run an advertising blog, and what purposes does this serve for advertisers?
Mr. NOLET: Most brands go out and buy online media based on a specific audience. So, you know, if you are running a Coca-Cola campaign, there's certain ads that men respond to far better than women. So, if they can target, you know, the more manly ads, that have more manly colors and different types of figures within the actual creative that they have, if they can determine your gender, they would want to show the manly ads to the men and the womanly ads to the women. Or you know, a specific agency could actually split off and actually show, you know - put to the women - show them they're - trying to think of a good example - their beauty products, and to the men, show them their new Samsung campaign. And that's all just to kind of - to more efficiently buy that media and spend their dollars.
BRAND: Oh, by the way, did you take the test yourself?
Mr. NOLET: I did. It said I was a woman.
(Soundbite of laughter)
Mr. NOLET: Although, if you look at the comments, it did seem to actually work for a fair number of people in there. People just kind of keep coming to the site. So, as soon as the traffic slow down a little bit, I'll do a follow-up and see - go through all this comments and see how accurate I actually was.
BRAND: Which sites were you on that made you a woman?
Mr. NOLET: I do - I actually had just gone to Crate & Barrel and IKEA, because I just moved three weeks ago, and that strongly skewed me the other way.
BRAND: All right. Thanks, Mike.
Mr. NOLET: No problem.
(Soundbite of music)
BRAND: That's Mike Nolet. He blogs for mikeonads.com, and you can check out the gender test there, or you can go to our blog and check it out there, npr.org/daydreaming. Alex, you were there.
ALEX CHADWICK, host:
I just actually posted - I took this test, posted my numbers there, and just see for yourself.
BRAND: OK. See for yourself.
CHADWICK: Although - again, why we're doing this test, if it does not work for you, it's beyond me.
(Soundbite of laughter)
BRAND: Well, one listener checked it out, and apparently MapQuest, very feminine. NPR's Day to Day continues.
NPR transcripts are created on a rush deadline by Verb8tm, Inc., an NPR contractor, and produced using a proprietary transcription process developed with NPR. This text may not be in its final form and may be updated or revised in the future. Accuracy and availability may vary. The authoritative record of NPR’s programming is the audio record.