Metric contract and aggregation rules

Define each metric once, the same way everywhere, and add results up honestly. Roll up decisions, not scores.

For every core metric, record

Definition, numerator, denominator, and unit

Population, workflow, and reporting period

Baseline or comparator

Data source and method

Owner and review cadence

Quality threshold and decision trigger

Subgroup breakdowns and privacy protections

Limitations and potential double counting

Aggregation rules

1Sum only nonoverlapping quantities with compatible definitions.

2Recalculate rates from underlying counts; do not average percentages indiscriminately.

3Deduplicate people when reporting institution-wide active users.

4Do not count one shared workflow improvement as separate individual, team, and unit benefits.

5Preserve subgroup and use-case results where campus averages would conceal differences.

6Keep unlike mission outcomes separate; avoid a universal AI value score.