What's the difference between device-level and aggregate data?

‘Device-level data’ means visibility into events performed by an individual device. This provides the most granular insight into campaign performance, and device-level data can be used for things like retargeting and multi-touch attribution (it also has obvious privacy implications if used unscrupulously).

‘Aggregate data’ is the functional equivalent of taking a spreadsheet of device-level data and running it through a pivot table. In most cases, this is actually what you want to do in order to make sense of the data. However, aggregation is an irreversible process: once you aggregate, you can't get the device-level data back again. And it can be very difficult to re-process the data into any other configuration.

With the justification of promoting user privacy, Apple's new SKAdNetwork system performs this aggregation inside Apple's black box, and then makes only the final results available for use.

Technically, SKAdNetwork postbacks are not 'aggregate data' because each one corresponds to a single install. However, SKAdNetwork only provides data at an aggregate level of detail, even within the individual postbacks.

This means in a world where SKAdNetwork becomes universal for ad measurement on iOS, any attribution workflow that depends on device-level data will become impossible.