Define the bounded sample
Choose a clearly segmented group of commercial files and document the referral reason, account ownership, client authority, permitted information use, contact restrictions, return conditions, reporting cadence, and sensitive-data handling. Do not mix unrelated consumer and commercial processes or leave legal and enforcement roles undefined.
Measure the quality of the operating record
Track accepted-file rate, missing-information rate, reason for information requests, time to first documented decision, and the categories that do not fit the defined lane. These measures reveal whether the referral design and file preparation are adequate without predicting money collected.
Measure decisions, not promises
Track whether files move to evidence requests, information refresh, defined commercial review, qualified referral, hold, return, or closure. Escalation mix and closure reasons are useful process indicators, while outcome claims require substantiation and should never be generalized from a small pilot.
Use findings to refine boundaries
A credible partner model can refine segmentation, document requirements, reporting, communication standards, and referral criteria after a pilot. It should not imply that a different agency, stronger tone, or additional pressure creates a guaranteed recovery path.