decadentist My threat model is to be able to use apps while sharing the least amount of information.
I think many people share that goal, but I'm not sure that matches the security/privacy-professional use of "threat model", which I think is: which information one wishes to keep confidential/intact, from whom (at what cost). For example, using a VPN shifts who learns about one's browsing patterns: instead of one's ISP knowing, one's VPN service does. But using a VPN does not change the amount of information being shared.
Some apps are designed to harvest data. Providing those apps with data will result in harvesting. OsmAnd lacks many features that Google Maps has, but OsmAnd was designed to harvest much less information (which can be set to zero).
decadentist Maybe my strategy is flawed - google could create pseudo accounts without an account, which would be made up of my IP [address], hardware ID, user agent and location.
I think there is a lot of history suggesting that organizations intent on tracking people do invest effort on tracking people -- as just one example, font fingerprinting. Thus providing data-harvesting apps with data while trying to reduce sharing data with the owners of those apps faces a fundamental conflict.