Hacker News new | ask | show | jobs
by majormajor 1109 days ago
You and your wife's phones are almost certainly tied together in multiple databases that adtech companies (especially the quieter tiers below Google/FB themselves) maintain and sell. And then you get re-targeted for ads you or her see, or ads based on things you or her see (retargeting was a really big thing back when I was in the industry, since behavioral prediction wasn't quite living up to the hype - so just re-show stuff you already think is relevant! I'm sure you've seen stuff like "here's thirty other lightbulbs" after you buy one lightbulb on Amazon in the past decade.)

Here's how it worked ~12 years ago: At some point you almost certainly have used the same home internet connection. A cookie associated with some account(s) of hers gets associated with that IP. Same for some account(s) of yours. Find IPs that over small time frames only have a handful of distinct users to filter out coffeeshops and such. Bam, likely family or roommate connection. Some sticky "supercookie" stuff or similar gets set so this can be both resurrected even when cookies expire (through like Etag cache headers or whatever) and then there's a lot of behind-the-scenes ad network offline-cookie-syncing.

One of the companies in that web partners offers ads on FB, another runs ads on Google, she searches and click something on your wife's phone (at this point regardless of network), it gets tied to you too.

It could be more innocuous but tracking same-household, cross-device for retargeting... all that is really old hat at this point.

I intentionally moved my career pretty far from ad-tech in the past decade, but from what I've heard in adjacent convos and such is that the newer developments include some of the newer platforms like smart TVs into all that old cross-device fun, too. Both as data source and ad display destination. A lot of TVs phone home about what you watch, even - like https://clinch.co/clinch-partners-with-samsung-ads/ There's a million ad-tech companies out there, they aren't all just writing real-time-bidding algorithms for traditional display ads.

2 comments

This is really interesting. I would buy a short book describing and explaining this privacy-related stuff the way you write about it.

But every time I read something like this, I am puzzled how they can have so much information and yet be so dumb. One example (but I could give you many): I like cycling and I subscribe to some cycling-related YouTube channels (Lanterne Rouge, Dylan Johnson - normal stuff, informative, bits of light humor now and then). YouTube keeps recommending me one cycling channel I absolutely hate - it is as if it was designed to piss me off. Sensationalist, negative, clickbaity titles, made up doping allegations every other post, throwing up emoticons in titles, creepy thumbnails with syringes and distorted faces. I never watch those videos, I clicked "do not recommend this channel" multiple times, now I just leave YouTube immediately every time this appears. And somehow, with all the data they have on me and all that sophisticated AI, they still have not figured it out. How? Any explanation?

I never saw much at all in terms of real, taste-level, user-based personalization beyond "keep hammering them with this one thing they clicked once" retargeting.

Everything is is by aggregate. So they can put you in a bucket with a bunch of other people with a similar footprint, and if 75% of people in that bucket convert when being shown that one you hate, you get to see it too.

Low-level optimization beyond that was rare at the places I worked.

"Sneaky but obvious" was way more common. And the browsers and platform vendors kept making it harder to do certain things, so it was a constant rat race to maintain the same level of browser tracking.

This happens to me too. I end up blocking the channels which works, no other signalling does the job.
Having worked with ad networks, publishers, and advertisers, it’s incredibly clear that ad targeting is barely this intelligent.

But the serendipity it can create is truly wild.

There's nothing particularly "intelligent" in any of the stuff I described - "see common device/profile/IP/whatever, make link."

Tracking is easy, intelligent personalization is hard. But tracking alone takes a lot of constant investment with tech development and arms race and all.