I had a weird one yesterday. I went to the New York Times using private browsing and read a single article in the relationship section (something about wedding speeches) and then noticed that the next page I went to 90% of the recommended articles were engagement announcements.

I guess things are hard when you only have a single data-point to base recommendations off.

Very true. The difficult problem in this case is that it's impossible to serve anything else, even though there are various techniques out there that still identify you when using private browsing (such as canvas fingerprinting - https://securehomes.esat.kuleuven.be/~gacar/persistent/index...): even if they're using such techniques under the hood (and that study, apparently from 2014, says 5.5% of top 100k sites were using it), they can't use it for recommendations, for obvious reasons (browser vendors would get eaten alive, private browsing would be overhauled, this technique would no longer work, everyone goes home sad).

I always find it amusing when you sign up for a service, and then for the next little while, half the ads you see are for that service. "Uh... but I've already signed up..."

It's especially hilarious when you see ads for something while you're using it.

So they got the retargeting part right, but forgot to add a second tracking code for those who signed up. "Woops"

I got ads for a mouse (similar model) after it was bought.

I think I learned in my marketing course that we (statistically) tend to look out for ads for things we have bought after the purchase.

The reason is hypothesised to be because our brain is actively seeking confirmation of previous decisions.

That said, your case most definitely was because of indiscriminate retargeting. Also I guess the effect I mentioned is more visible for bigger purchases : )

Ha, they've actually put in a rule to link to an image rather than the listing to stop this from happening.