Dataset: AI Usage
Entity: AI Usage Measurement
Field ID: aggregation_level
Type: Select list
Description: What the measurement aggregates over. Possible values are:
USERthe measurement covers one actor.REPOSITORYthe measurement covers one repository.ORGANIZATIONthe measurement covers the whole account.
Note: Aggregation levels are not interchangeable and must not be aggregated together. See Definition: AI Usage Record Structure.
Source: Calculated
Transformation logic: Assigned from the provider report a record comes from.
App Mapping
GitHub Copilot |
|
Anthropic Claude (coming soon) |
|
Cursor (coming soon) |
|
OpenAI Platform (coming soon) |
|
OpenAI Codex (coming soon) |
|
Reporting Use Cases
The Aggregation level field is one of the two fields that must appear in almost every AI usage filter. It says whose activity a record counts.
Always filter on one value: Mixing levels adds a day's per-user counts to that same day's organization total. A report that does this roughly doubles every figure.
Per-developer reporting: Filter to
USER. This is the level that carries an actor, and the one from which active developer counts are derived.Repository impact: Filter to
REPOSITORY. Only GitHub Copilot publishes this level, and it carries pull request measures only - no tokens, no actor, no model.Organization totals: Filter to
ORGANIZATION. These are published by the provider rather than summed from user records, so they will not exactly match the sum of theUSERrecords for the same day.Beware overlapping organizations: Copilot attributes organization usage by membership, so a person in several organizations is counted in each. Do not sum organization records across several Copilot integrations.
