For example Visa has an exclusive deal with oracle. So only oracle can buy audiences with visa data, and visa has super strict requirements and only builds them in house. If you say “I want users who purchased x product” the size must be 5mm users minimum (I think) and visa models it up using lookalikes/etc to 20mm+ users (maybe slightly off on sizes). Then it’s like $4 cpm to use at a dsp. Brands/agencies etc have to go through oracle to get visa data.
Here's one example - http://www.dev.kbmg.com/services-solutions/data/data-solutio...
I've seen this, and many other similar data sets. You can easily look up yourself, your friends, etc.
The example I linked, Amerilink data set, I have a copy of (via my company), sitting on our internal corp NAS drive, and it is 100% raw data with PII. Can query on name, mailing address, DOB, etc. We have an unrestricted license that we pay a good chunk for annually.
1,000+ attributes on every US consumer.
I attended a presentation some years back from TMobiles marketing team.
According to the speaker, when you visit their homepage they pull up complete financial records on you, including how much your home mortgage is, if any, and use that to customize what products you see promoted first.
Matching a cookie to individual names is not instantaneous and would be very expensive for someone at t mobile’s scale to do every time a session started.
Some b2b companies do something similar where they can match your ip to a list of known corporate ips to find where you work in real-time, but that’s pretty expensive.
https://www.vice.com/en/article/m7vzjb/location-data-abortio...
> SafeGraph calls the location data product “Patterns.” In total, the data cost just over $160. Not all Planned Parenthood locations offer abortion services. But Motherboard verified that some facilities included in the purchased dataset do.
> SafeGraph’s data is aggregated, meaning it isn’t explicitly specifying where a certain device moved to. Instead, it focuses on the movements of groups of devices. But researchers have repeatedly warned about the possibilities of unmasking individuals[0] contained in allegedly anonymized datasets.
And related to [0]: https://www.nytimes.com/interactive/2018/12/10/business/loca...