How are mobile ad mediation services able to provide impression-level revenue values?

As of mid-2019, it's becoming common for mediators to offer impression-level (also marketed as "user-level") ad revenue data to publishers via SDK callbacks, data pipelining to analytics services, or both. This offers a big advantage over the previous standard, which was to pipeline daily revenue data to analytics providers, segmented only by country. This daily revenue data would be essentially averaged over users from various campaigns based on retention and other questionably relevant signals, creating a "dumb average".

By offering impression-level data, mediators enable publishers to account for variations in user ad LTV by attribution source. This allows a fine-grained understanding of how much users from different sources contribute to ad LTV. This is important because users from different sources will be worth different amounts to ad networks, and knowing this variation on a detailed level can help publishers increase the accuracy of their ROAS metrics. See this article if you're not sold on the importance of this granularity.

Amongst mediators who offer this data as a service, there are generatlly two variations of each impression revenue event:

• Estimated Revenue: Inferred revenue based on the eCPM floor of the ad unit that filled the impression. This is the most common case as of mid-2019.
• Exact Revenue: Verified revenue based on real-time bidding from DSPs or in-app header bidding networks. In these cases, the exact CPM of each impression is known.

The mediator should differentiate between the two types of revenue, although most times the publisher will just use them in aggregate, as even a rough estimate on an impression level is still useful.