The author's point is that, with the data available to the platform owners, they could easily deduce the subset where that P value is nearly zero (having already bought the mattress).
Places that would also make sense to advertise mattresses: around universities and dense rental housing, or new developments of single family houses. I required no spy networks to make this list.
It seems that P(returning a bought mattress) is more than 10%, so it probably makes sense to continue advertising to those people.
Also, there are probably people who buy multiple mattresses (eg. parents), that might buy a second mattress.
The point of the article is that platforms know that both your examples aren’t true. The author isn’t a parent and isn’t in the market for buying a second mattress or returning.
I think the issue is that the platforms don’t want to actually target, they just want to pretend and allow for the narrative for why.
I would expect that the search pattern of multiple mattress buyers is very different than single (ie, single searches and buys one and stops searching; multiple searches, buys one, keeps searching). And for returners have identifiable patterns (ie, searches, buys, searches for “how to return”).
Remember that google knows pretty much every single search you run and every page you visit and every checkout you complete successfully (if you’re using chrome).
> The author's point is that, with the data available to the platform owners, they could easily deduce the subset where that P value is nearly zero (having already bought the mattress).
Let's think carefully before we argue that Google and Facebook should expose (indirectly or not) even more of the information they collect to third parties.
Places that would also make sense to advertise mattresses: around universities and dense rental housing, or new developments of single family houses. I required no spy networks to make this list.