Hacker News new | ask | show | jobs
by andybak 3506 days ago
I behave online with the understanding that my behaviour is essentially public.

PS - how come all website product recommendations still suck? Data collection is rather ahead of data usage it seems...

3 comments

Oh, yes, the algorithms supposed to target you specifically are often quite bad and annoying. For example on Amazon, when you just bought a coffee machine, your recommendations will show you a bunch of coffee machines. Well, I just made a choice and ordered a coffee machine, I am not going to buy another one for quite some time.
Bad for you != bad for the advertiser, and bad for the advertiser is not necessarily correlated to annoying to you.

While your particular case may need just one coffee machine, there are other scenarios. If you just bought a coffee machine, perhaps it will be defective and you'll need a replacement. Perhaps you didn't know about the one they're showing you and will return the first one to buy this one. Perhaps you are just getting into this coffee stuff and realize that a second one for the office would be nice. Perhaps you're buying a few to compare them.

It could be rational to show you the ad instead of a random individual if the sum of these scenarios is greater than the rate at which random individuals buy coffee machines.

Or, yeah, it could be that their ad-showing algorithm just tells them you have the word "coffee" in your recent browsing history.

> website product recommendations still suck

Targeted advertising isn't about giving you good recommendations. It's about advertisers picking which demographic they want to influence. It may be presented as a "recommendation", but it's still an ad.

It's also heavily weighted by what they want you to buy: over-stock, higher margins, last year's models.
> I behave online with the understanding that my behaviour is essentially public.

Also assume that the connection between any and all pseudonyms used only is public knowledge. Because it is.

Agreed.