Comment by wcfields

This data is used in aggregate to decypher household penetration and refine known data about demographics and household income.

Here's an example of the sorta-end-game: Currently they can know how many kind of "devices" are in your household and using probabilistic statistics give a decent n value of how many "devices" are in a certain zip code. Using that, your advertising can become more efficient by only buying ads in zip codes that contain certain "devices".

I say Zip code because that's what I've worked on in the past at the most granular level for Marketing Mix Modeling or MMM. You can easily venn diagram your first party data with 3rd party brokers, and you can cleanroom the whole thing to get a decent venn-diagram of the overlap.