The most common case of "Sharing data with 3rd parties" is when the first party uses their targeting platform (usually DFP/GAM) to serve a campaign to an advertiser looking for a specific audience. For example, NYT might sell a campaign targeted towards young people, and therefore would allocate inventory on their tech articles to that campaign. The advertiser or their DMP/DSP now serves content from _their first party domain_, allowing them to cookie the user. This gives them an opportunity to say "NYT said this user was in their 20s", which they can store in that cookie's profile, and later they can see that cookie when they're bidding for that user in a programmatic, non-direct setting. They can target the same user again, later, for cheaper than NYT direct pricing.

Its talked about as "selling data", but its important to understand how that works, because its inherent to the value proposition of digital advertising. Its not "we have X data, do you want to buy it for $Y?". Its not about exporting CSVs or connecting data warehouses. Its "We have X data, do you want to buy inventory for that audience?". So the only way that a publisher can really sell ads without "selling data" is to do it on a non-targeted basis, so either through programmatic, non-direct, or through non-targeted direct campaigns. In isolation, a publisher who is selling non-targeted direct campaigns is not going to do well. Its just not a competitive offering to advertisers, unless the site's entire audience falls in a certain demographic--in which case, advertisers will just mark those users as having that characteristic, so in the end its still "selling data".

One way to look at the value of selling first party data is to look at the gap between programmatic and direct sold inventory. Direct campaigns in the US typically go for $10-50 CPMs (highly dependent on the audience, context, and units). Programmatic averages closer to $1-2. So having first party data _and scale_ can get you a 10-20X over not having data (and scale).