Skip to content

Instantly share code, notes, and snippets.

@Zarkonnen
Created June 20, 2012 19:50
Show Gist options
  • Save Zarkonnen/2961818 to your computer and use it in GitHub Desktop.
Save Zarkonnen/2961818 to your computer and use it in GitHub Desktop.
Magic Targeted Advertising
"Why do some people say that if they have to have advertising, they'd prefer it to be targeted?" @katzenfabrik asked me.
Good question. The idea of targeted advertising, or rather, the idea of the detailed file on people's habits such targeting requires, creeps us both the hell out. And we both aggressively ignore ads. I'd certainly agree with George Monbiot when he calls advertising a poison that distorts our view of the world and relies on destroying self-confidence to get you to buy something.
Still, the preference for targeted advertising is pretty commonly voiced. I think this preference requires two things: a lack of worry about the details of your identity being public, and a belief in something I'm going to call the magic advertising targeter.
Targeted advertising works by putting people into buckets based on characteristics that distinguish their buying behaviour. Women buy handbags. Men buy beer. Geeks buy computer games. These rules don't have to be absolute, they just have to be good enough to improve the proportion of people who buy something when shown the ad. Since anything about a person can potentially be a useful differentiator, advertising networks want to get all the information about a person they can get. (Which is what Facebook's for.) Some of those characteristics can be pretty intimate, though: your sexual orientation. Your marital status. Your political affiliation. A male, straight, married US republican probably has rather different buying habits from an Italian lesbian who's just come out of a long-term relationship.
Now if your personal characteristics are such to afford respect and good treatment from others, there's little fear in letting other people know about them. But you may really not want everyone to know you're gay, or disabled, or into kinky sex, or radical politics. Certainly you have a right to not everyone knowing. But let's say that you - correctly or not - think that widespread knowledge of your intimate personal characteristics doesn't endanger you, or hurt your career prospects, or make you a target of ridicule. In that case, one objection to targeted advertising evaporates.
Point two: it must actually advertise things to you that you might want to buy. Despite the vast amounts of data being collected, targeted advertising often still gets it wrong. I might see car ads because I'm a man - but I don't drive. Or ads for an upcoming PC game - but I run a Mac. Or, you know, sexy singles in my area - when I'm quite happily in a relationship. This subtle wrongness can be positively dispiriting, such as being flooded with ads for dating sites when Facebook detects you've broken up.
This is where the magic advertising targeter comes in. Surely, the problem with this mis-targeting is an insufficiently fine grid of sub-sub-sub-categories. If only the system were aware of my inability to drive it would pitch bicycles or driving lessons. It would tell me only about Mac games, or about virtual machine software that lets me play PC ones. It would suggest romantic weekend breaks. It would understand that this last breakup has left you quite disenchanted with the whole idea of romance and in dire need of some fresh experiences, and will pitch city breaks, extreme sports, and ideas for new hobbies at you. Only after a well-calculated interval will it tentatively suggest that a dating website might just be the way to find someone new.
Of course, this system would still fail. I already have a bike. I don't want to learn to drive. I can't stand most A-list computer games. The weekend breaks would all be unrealistically priced. The city breaks and extreme sports are of no interest to someone whose idea of a good time is scrabble down at the pub with their friends.
So let's pour on more magic! Now the MAT is all-knowing and can predict consumer choices with perfect accuracy. It would tell me only about turn-based strategy game of reasonable depth and polish that run fine on my aging hardware. It would introduce me to amazing obscure books and pursuits that would enrich my life. And I would be guaranteed to buy every single one of those things, and it would advertise every single thing in the world which I would buy. Single impressions in this systems would cost dozens of dollars, as each impression is guaranteed to make a sale.
It's not possible to make this system, but given enough data and clever inference we might get close enough that every day, out of the dozens or hundreds of ads shown, there would be one thing to my liking.
Unfortunately, this system is also unnecessary. In practice, advertising mostly works by creating a need, not fulfilling it. A car ad isn't there to inform you of this new make so that you can go "Oh, it has a built-in MP3 player. I must purchase it!" It's there to, by endless repetition, create a feeling in you that you like this car and want to own it.
Moreover, advertising is only interesting when aimed at large demographics. Why go to all the effort of making a campaign for selling weekend breaks to 28 year old software developers with cranky beliefs about the commercialisation of romance, when you could be targeting a much larger and more lucrative market?
Advertisers aren't interested in targeting discriminating consumers. A sale is a sale, and it's best to make it where it's easiest. By virtue of even thinking about the nature of advertising you have already exiled yourself to a small and barren demographic no marketer would ever bother targeting.
In conclusion, advertising, targeted or not, will always try to sell you things you don't want. That's what it's for.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment