Unfortunately, the data underlying both use cases is the same.

In order to target adds "at people in a narrow geographical area" I need to know when those people are in that "narrow geographical area". I.e., I need to know their location to an accuracy good enough to say they are in that "narrow geographical area" (or in this case, their phones location, which is a good proxy for their location).

Once I can know "the phones location" to a level of accuracy to make "narrow geographical area" targeted ads profitable, all I have to do is average several overnight readings of 'location' together to get a very good estimate of "where someone lives".

I don't understand what you're trying to say. If you already know where someone is, what's the point of using targeted ads to try to locate them?

The point of OP is that I can't use targeted ads to answer the question "Person X: where are they?".

I'm saying that the underlying data is the same for both, so allowing one (targeted ads based on location) provides the data necessary for "locate where someone lives/works/hangs out/etc.".

With the result that you have to believe that the "targeted ads" provider is both secure and ethical.

> With the result that you have to believe that the "targeted ads" provider is both secure and ethical.

"Secure enough" and "ethical enough". Plenty of organisations know stuff about me that I'd prefer not to be public or sold to the highest bidder. They might leak it to the government or one day be the target of a megahack - but I file that under either "meh" or under "acceptable risk".

But luckily