← Read the pilot measurement guide

SCORECARD / OBSERVED PILOT RESULTS

Did the pilot save work after review?

Compare two observed samples of the same task. Count the human effort to get usable work, including corrections. Check quality separately before deciding what comes next.

Use aggregate counts only. Entries are not saved. This is a descriptive comparison, not a causal study, revenue forecast or approval to scale.

Include every attempted task, including failures. Use comparable task difficulty and the same error definition in both samples. Count tasks with at least one error at the same checkpoint, such as before correction. Count each task once.

Enter total person-minutes across each sample, not averages or machine runtime. Baseline minutes include all review and rework. Split pilot effort into work before review and review/rework, without double-counting.