Glossary
What is raw data?
Raw data is granular, unprocessed event-level records — individual installs, clicks and in-app events — before any aggregation or transformation has been applied.
Why access to it matters
Aggregated reports answer the questions the vendor anticipated. Raw data answers the ones you have.
Every aggregation discards information, and which information was discarded was decided by someone who did not know what you would want to ask. A platform reporting installs by campaign and day cannot answer a question about the interaction between creative, time of day and device — the data to answer it existed and was thrown away before you saw it.
Raw data also lets you reconcile independently. Where a platform's aggregate disagrees with your own system of record, only event-level records let you find out why.
What it costs
- Storage and pipeline. Volume is large and needs somewhere to live — usually a data warehouse fed by ETL.
- Skills. Raw data is useless without someone who can query it competently.
- Governance. Event-level records are typically personal data, with retention limits and access controls that follow.
- Vendor fees, since raw export is frequently a paid tier rather than a default.
What is no longer available raw
A growing share of measurement is aggregate by construction rather than by vendor choice. SKAdNetwork postbacks carry no user-level identifier at all, and privacy thresholds withhold detail below minimum cohort sizes.
So raw data access is still worth negotiating for the sources that have it, and it is no longer a complete picture on any platform. The gap is filled by causal measurement rather than by more granular records — which is a genuine change in how analytics teams have to work.
Common questions
Do I need raw data access?
If you have the capability to use it, yes — it is the difference between answering your own questions and answering the vendor's. Without a warehouse and someone to query it, the export is a cost with no return.
How long should raw data be retained?
Long enough for the longest analysis you actually run, and no longer. Retention limits are a regulatory requirement in most jurisdictions, not a storage preference.
Where we write about raw data
More in analytics and data
Where the numbers are stored, shaped and read.