4 ms·
Even at big tech companies, a lot of those ML models are fed by nightly jobs, so updated once a day. If the targeting occurs on an outside property, there may b
by Maro 8y ago
Even at big tech companies, a lot of those ML models are fed by nightly jobs, so updated once a day. If the targeting occurs on an outside property, there may be more delay. So if you make a purchase on day X, you should expect to see the ad for another ~48 hours. That's assuming there's logic which actually de-targets you post-purchase.
- oldcynic 8y agoEven Amazon don't get this right and should have unbeatable knowledge of when the sale completes. Their helpful suggestions still only seem to work usefully for books and music.
- Mvandenbergh 8y agoI suspect that's because those are the only things that are easy to model. People buy them: a) based on personal taste b) in volume I just bought a ladder on amazon, what information does that tell them? Since it's the only tool I've bought from them it doesn't tell them much. They now know that I might be willing to buy more tools from them in the future but doesn't help in targeting which kind or how often. Me being a ladder buyer is a very weak signal that I will buy any particular item from them. If on the other hand, I've bought 15 Aubrey Maturin books from them over the last few months, it doesn't take the most sophisticated model to determine that I probably want to buy number 16 next. Or that I might like to read the Horatio Hornblower books.