Sampling loan files without quiet bias

How stratified pulls keep a financial audit of loan origination applications honest when product mixes shift mid-period.

Charts used while designing a loan file sample

A pure random sample looks fair until you discover eighty percent of draws came from one branch that imaged files early. Stratification by product, booking month, and channel keeps the audit conversation grounded in the book you actually run.

When a credit policy changes mid-period, split the sample across both checklist versions. Findings that ignore the version switch create false “control failures” that are really change-management noise.

Bring the booking extract with application IDs, product codes, and booking dates. Without that spine, reviewers invent convenience samples — and convenience samples flatter whoever prepared the archive room.