YouTube’s Demographic

ad.jpgI’m really tempted to have my buying agency call YouTube to ask about price and demos, then blog about it. It’s not appropriate and I won’t do it, but I’m still tempted. What’s the profile? I think I read it skews male and has a higher average age than you’d think. But what about median income? Urban/rural split? Are we buyers of high-end electronics or Nascar fans?

Even without inside information, the ads sometimes reveal the demo. Anytime you see “shoot the monkey” ads and this one (targeting people who have bad credit) you get a sense of the target.

The problem with YouTube is that it doesn’t know me accept by age. Do you know how quickly the YouTubers would give out personal information (as an option for some marginally premium service)? Do you know how much of a premium YouTube could sell the ads once they targeted by demographic?

I should charge for this stuff.

4 Replies to “YouTube’s Demographic”

  1. YouTube knows me by age and general location, but only because I’m too stupid to lie.

    Judging by the profiles I’ve looked at, however, there are an awful lot of 99-year-olds who still look remarkably like teenagers and twenty-somethings. I mean, they wouldn’t lie about their birthdates, would they?

    You are right – they could sell far more appropriate ads with more demographic information, but I know I’m not revealing my income (or other personal identifyers) to them any time soon.

  2. We see some of the same things at Microsoft (MSN and Windows Live in particular) that you describe here – users aren’t always 100% honest about their age or gender when they sign up to use a free service. Things are a bit different when users create profiles their friends or peers see (e.g. Facebook, Windows Live Spaces or MySpace). This information tends to be quite accurate. [For context, I run a team at Microsoft that designs ways for advertisers to reach specific audiences across Microsoft channels, including techniques such as demographic, geographic and behavioral targeting.]

    You get even better accuracy when you have a billing relationship with your users (such as AOL dial-up or paid services like personals), or when user information is necessary to provide the service (e.g. a zip code or city to get a weather report.

    Another technique becoming more popular is attribute prediction. To do this, you start with profiles you know to be accurate. You take the behaviors exhibited by these users as a baseline and infer relationships between that behavior and the users’ attributes. Then you build models that predict other users’ attributes based on their behavior. You’d be surprised how accurately you can predict someone’s age and gender based on the keywords they search for, the music they listen to, the videos they watch or the content they read.

    If you look at Google, for example, they don’t ask for any demographic information when you sign up for a Google Account (via Gmail for example), but all their services are tied to that ID. Using that ID, Google probably has the technical ability (I have no idea if they’re actually doing it) to tie a user’s behavior and information together across all their products. Where they do capture user data, they could easily use that as the baseline for building predictive models.

    Amazon is perhaps best known for the modeling they do to provide product recommendations based on the behavior of their user base and the things you search for, look at and buy.

    Failing any of those techniques, you can always fall back on data from companies like comScore or Nielsen. These companies have panels of users who agree to provide all kinds of personal information and allow their browsing/search behavior to be monitored. Using all this data, comScore and Nielsen publish characteristics of users on all kinds of Web sites. Most advertisers buy this data but still target their ads to improve their efficiency/ROI.

    Regarding the “shoot the monkey” ads – you may not want to make inferences about an audience based on ads like “shoot the monkey” or “toe fungus.” The companies advertising these products usually buy inventory from big ad networks. The ad networks buy inventory across “reach” Web sites (like a YouTube, MySpace, or any Web email service) where there are so many users, it’s likely they’ll hit their target market. Those buys may end up with impressions shown to people outside the target audience, but often times, the CPMs are so low that the campaign efficiency is still pretty good. Some ad networks optimize the display of those ads to minimize wasted impressions as well.


    I write about goings-on in the online advertising space on my blog ( Feel free to check it out and share your thoughts.

Comments are closed.