[Benjammer]

The key to this is that you need to think of your "online identity" as something that is not directly controlled by you, ever. It's like a collection of related "behaviors" and heterogeneous data points on the internet that can be associated together with some amount of confidence that they are all from the same person, through a myriad of tracking processes.

You actually creating an account (username data point), and logging in and using the service (actions) are actually minimally relevant to the process of ad targeting. All they do is slightly bump up that "confidence" score I mentioned earlier, and even then only for that specific set of behaviors/data for that page/service.

Your user accounts are only transient links between sets of behaviors and targeted ads. The actual account is easily replaced by just making it a generic "entity" not related to accounts and then link any accounts that do end up being created to these "higher level" entities that you already have.

For a simple example: "A web client hit our tracking pixel on X service from the same ip address as a hit on our tracking pixel on Y service, so let's assume these two behaviors are the same person for ad targeting purposes, and we'll just use that address as the unique indicator, regardless of any existing (or not) user accounts on either service. If we get a user signup action that originates from the same address, then we can link in the specific service account/username to our existing tracking entity for this person."