A mid-size consumer lender can hold tens of thousands of active accounts. Reviewing every file is neither useful nor affordable. The craft is choosing a sample that can still support a board-ready conclusion.
Stratify before you count
Split the book by product, origination quarter, and risk grade. A flat random sample often under-represents the thin vintage where waivers clustered. Stratification forces those cells into view.
Separate origination from collections
Income and KYC evidence belong to the origination sample. Contact attempts, forbearance, and write-off triggers belong to a collections sample drawn from delinquent accounts. Mixing the two in one worksheet confuses error rates.
Document judgment, not only checkboxes
When a reviewer waived an income proof, record whether the waiver followed a written authority or a habit. Habit-based waivers are the findings investors care about; documented exceptions within policy are usually commentary.
Keep the archive honest
If digital files are incomplete, say so in the engagement letter and the final memo. Inflating sample sizes against missing archives creates false precision. We would rather report a constrained sample with clear limits than invent coverage.
Sizing still depends on your risk appetite and any funder checklist you must satisfy. Bring those constraints to the brief; they shape the fee as much as the headcount of active loans.