The model can be tested against consented summary records without collecting raw intervals, addresses, postcodes or public exact contributor counts.
Eligibility permits private human review only. It cannot calibrate, promote or publish a model automatically.
Accepted only in bands
Flow basis, system-size band, consumption band, observation-duration band, consent version and source version.
A withdrawal removes the record from cohort processing.
Raw intervals, exact contributor counts and location identifiers are rejected at intake.
A qualifying cohort can advance to private human review only.
Public reporting uses a minimum contributor threshold and coarse published bands.
The protocol is public so the evidence boundary is inspectable. Participation, where opened, remains authenticated and consent-led.
Read the methodology