How it works
Each row describes one occurrence: what happened and when, which app and platform, the attributed network, campaign and creative, and any event parameters such as revenue. Delivery takes several forms. Scheduled exports write files or tables to cloud storage or a warehouse every hour or every day. Pull APIs let you request a report on demand. Push mechanisms send each event as it happens. Columnar formats such as Parquet suit large volumes; CSV is easier to open by hand.
Why it matters for an app team
Dashboards answer the questions their builder thought of. With the rows in your own warehouse you can join attribution with backend data such as orders, refunds or subscription status, build your own LTV models, feed BI tools and keep history for as long as you need. It also lets you audit the numbers you are billed on.
In practice
Example with made-up numbers: an app with 60,000 installs and 1,500,000 in-app events a month receives a daily Parquet file with roughly 50,000 event rows. Joining it to the order database shows that purchases from one campaign are refunded twice as often as those from another, something no standard report contained.
Common confusion
Raw does not mean unlimited. SKAdNetwork data is aggregated by Apple and has no user-level rows, and some ad platforms restrict which user-level fields may be exported. Exported rows are personal data in most cases, so access control, retention and deletion requests have to be handled on your side too.