Hacker Newsnew | past | comments | ask | show | jobs | submitlogin

You sell cars.

I listen to Donovan, The Beatles, The Stone Roses, Mozart, Au, REM, Dj Shadow, Handsome Boy Modelling School, Chopin, Ludovico Einaudi and Daft Punk (to name but a few).

I've liked pages for five star greek hotels, the burger place on the corner, the guy who sells awesome falafel's on Hoxton market. You know I saw this video of a talking dog that I thought was totally awesome.

Just this month, I read an article on the Greek debt crisis in the economist, an article on David Beckham in the Daily mail website, I shared a wikipedia page about Rommel, I liked an article about continuous integration with node.js and Jenkins.

Now sell me a car.

I actually like Volkswagens and Volvos, but you won't find that anywhere in my Facebook data.



I actually like Volkswagens and Volvos, but you won't find that anywhere in my Facebook data.

I'd bet there's actually a pretty high correlation between many of the things you just listed and a preference for VWs and Volvos. It doesn't really take that many data points to start filling in the blanks.

But I think this kind of targeting is going to be more useful in selling you a product you don't yet know about but are likely to want than it is a car that you are almost certainly already aware of.


Yeah, the preferences listed may not indicate a preference for Volvo over Audi, but they're certainly consistent with someone I'd target for near-luxury European cars rather than giant Chevy pickup trucks, Korean mini-vans or tiny econo-boxes.

Daft Punk? You're 30-ish. Mozart and Chopin? College graduate. Like a 5-star Greek hotel (but not that many other luxury items)? Shit, I've narrowed down your disposable income into a fairly narrow bucket. I can also tell that you don't have any children. I would target the shit out of you for Volvo, Volkswagen, Acura, Infiniti, Mercedes, BMW and Audi.

And that's just my own guesstimates. Backed up by proper data I'm sure I could do way better.


>"I would target the shit out of you for Volvo, Volkswagen, Acura, Infiniti, Mercedes, BMW and Audi."

And why not Lincoln, Cadillac, Lexus, Saab and Jaguar? In fact, Volkswagen isn't a luxury brand (no more than Honda, Nissan, Toyota), and hardly fits among the brands you mentioned above. Yet it's a preference of the OP. So, I think you've proven his point; your targeting hardly tells us anything.


Sure, maybe those too. Except perhaps Saab, which last I heard was broke, and Jaguar, which I'd probably target at a slightly older crowd.

Lincoln and Cadillac maybe not, too -- perhaps Americans who holiday in Europe are much more likely to buy European cars. If I had more data I could draw up some correlations and tell you exactly how much more likely they are to buy European cars and whether it's still worth waving a shiny new MKZ in front of our Volvo fan.

There's always exceptions to any of these correlations between stuff we like and other stuff we might like. In fact we're all exceptions; we all like things which are anticorrelated with other things we like. But that's not a big problem -- your targeting algorithm doesn't have to be perfect, it just has to be significantly better than random scattershot... and ideally, better than Google.

In an ideal world, I'd only serve Acura ads to people who:

a) Are buying a new car soon

b) Are certain to buy an Acura if they see an ad for one, and

c) Are certain to not buy an Acura if they don't see an ad for one.

but failing that I'll settle for targeting people who are in vaguely the right age group and income bracket.

Actually, I think car sales are a bad example; if I wanted to sell cars I'd use google ads, because people buying a new car already know they're buying a new car, and are probably out there doing research on the subject of new cars. Whoever posted the milkshake example had the right idea for facebook ads.

Google ads can sell you stuff that you're looking for, Facebook ads (if implemented properly) can sell you stuff you don't yet know that you want.


Yesterday I created a paste bin with the brand of footwear I prefer. Go!


Are you sure? Advertisers have been comfortable with finding correlations between unrelated consumer behaviors for decades. I can almost guarantee you that, on average, someone who likes five star Greek hotels is more likely than the average American car buyer to consider a Volvo.

I think a lot of Facebook's value depends on just how useful that personalized information is, and how difficult it will be for a competitor like Google to extract in other ways. My guess is that it will turn out to be pretty useful as advertising campaigns (and brands themselves) become more and more segmented, but also not so hard for traditional cookie-tracking methods to get 80%+ of the same useful information.


Car salesmen aren't the only people that want to sell you things though. A music website, a holiday company, a food company or a bookshop now have loads of info that could help target ads at you.

Not that I'm defending Facebook advertising (there's other reasons that I think it's largely pointless, at least from a click-through perspective - the primary one being that if I'm going on a social media site, I'm usually going to be social, not to buy a new car or book a holiday), but the fact that one particular market can't necessarily target one particular user doesn't seem to be a great argument against it.


You're looking at it from the wrong point of view. The car makers want to sell cars, they don't care who to. Someone who lists their favourite car will see car adverts, whereas you should see adverts for concerts in your region, watching economics documentaries on Netflix, and holidays to major falafal-eating countries.

Well, that is, if Facebook does it properly. And you don't block them.


I would have liked your comment better if you hadn't given away what kinds of cars you like and made them guess.


See my reply to PlanetGuy.




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: