You can also map a home out depending on signal strength. That gives you approximate size of home which gives you approximate income.
It can also correlate it with geolocation data. Google, for eg, sniffs all broadcasted SSIDs with their StreetView cars. If you can pick up on a SSID (or any of the MAC addresses of the other devices), you can buy the data set that includes it which further pinpoints demographics given the neighborhood AMI.
You can also build behavioral profiles patterns based on things like, for eg, if a baby monitor model is present or a robot vacuum, if certain devices only connect at certain times, etc.
I think the general rule for adtech is that profile guesstimates just need to be around 70%+ fidelity to determine if a sale can be made.
Lastly, you can also just sell the data on the gray market. The more datapoints, the higher the price. Most consumer product companies do that since we have little-to-no data privacy laws and the people who seem the most aware of it also are generally very apathetic and disinterested in advocating for them.
Also the social-graph aspect.
For example, your aged mother's phone will get pinged within X meters of an urgent-care facility, or she'll do some web-search about "hip pain", and then all the adult children start getting ads to about elderly-parent-care.
Or perhaps the pervy-panopticon decides some phone-on-wifi events look like adultery, and both spouses start getting ads for divorce lawyers. (Bonus if certain specialized "adult" toys are detected on Wifi or Bluetooth...)
Yeah, it definitely makes sense to me if what adtech is often doing is just backing out from specific data to a general profile like income, location, etc, so that that level of targeting can work.
Its just sick.