How to measure collections from a loan tape
Cash a cohort has paid back, divided by the principal it was lent, read by months since origination. Why collection efficiency answers a different question, and the three tape mistakes a fund checks for.
To measure collections from a loan tape, divide all the cash a cohort of loans has paid back by the principal it was lent, month by month since origination. The answer is a multiple of original principal. At 1.0 times the cohort has returned exactly what it received, and the excess above 1.0 pays for funding, losses and costs.
The question a fund asks is per dollar lent, and only a fixed denominator answers it
A fund reading a lender's tape for the first time usually looks at three things: delinquency, volume and collections. The third one is a single question. For each dollar this lender put out, how much has come back, and how fast? A monthly collection rate does not answer it, because that rate divides by whatever was due that month, and what was due changes with growth, write-offs and restructurings.
The per-dollar measure divides by a number that never moves: the original principal of a cohort, meaning all loans originated in the same month or quarter. It is the same choice that defines a static pool, applied to cash instead of losses. The arithmetic is old and public. A 2017 paper by an analyst at the consumer lender Axcess Financial computes it loan by loan, «dividing total payment amount collected by the loan amount financed», then cumulates it by month on book for each vintage. Encore Capital has shown investors its «cumulative actual collection multiples by vintage year», defined as «Total collections / Purchase price», in material filed with the SEC since at least 2012. There the denominator is the price paid for the debt. For a lender it is the principal disbursed.
the line a cohort crosses once it has paid back what it was lent
Cumulative cash collected divided by original principal, read by months since origination.
Collection efficiency is a different number, and a write-off can improve it
The ratio most often reported as «collections» is collection efficiency. ICRA's securitisation methodology of June 2025 defines the cumulative version as the «ratio of cumulative collections to cumulative billings, including opening overdues», and states its weakness in the next paragraph: «A company that adopts a more stringent write-off policy generally shows a better collection efficiency ratio [...] because a delinquent contract once written off is not included in the billing amount, going forward.»
The ratio also passes 100%. In ICRA's rationale on two Shriram Finance vehicle loan pools, payout month May 2025, cumulative collection efficiency was 101.2% and 104.3% while loss plus 90-day delinquency stood at 1.6% and 1.0% of the initial pool. Both numbers are correct and neither says how much came back per unit lent. Collections per dollar lent cannot be flattered by a write-off, because a written-off loan stays in the denominator at the principal it was given.
Illustrative cohorts. Cumulative cash collected as a multiple of original principal, by months since origination. Originated in millions of the lender's own currency.
| Cohort | Originated | Month 6 | Month 12 | Month 18 | Month 24 |
|---|---|---|---|---|---|
| 2023 | 310 | 0.38x | 0.74x | 1.04x | 1.21x |
| 2024 | 420 | 0.36x | 0.70x | 0.97x | not yet |
| 2025, renewals counted as collections | 560 | 0.37x | 0.71x | not yet | not yet |
| 2025, borrower cash only | 560 | 0.31x | 0.62x | not yet | not yet |
At month twelve the 2023 and 2024 cohorts have returned 0.74x and 0.70x. The 2025 cohort shows 0.71x if a new loan that repays an old one is counted as a collection, and 0.62x if only cash from borrowers is counted. The second number is the one a fund rebuilds from the payment history, and it is the one it will price on.
The table describes no real lender. Its point is the comparison it allows: cohorts are read at the same age, so month twelve of 2025 sits next to month twelve of 2024, and a newer cohort running below an older one is visible months before its losses are written off.
A collection rate tells you how this month went. Collections per dollar lent tell you whether the loans are paying for themselves.
Three mistakes in the tape change the answer, and a fund checks for all three
The first is counting a renewal as a collection. When a borrower takes a new loan and its proceeds repay the old one, the old loan shows a full payoff, but the cash came from the lender. Counted as collected, it lifts the old cohort and hides a borrower who never repaid from income. Flag refinancing proceeds in the payment history and keep them out of the numerator.
The second is dividing by today's balance instead of original principal. The balance shrinks as good loans repay and as bad ones are written off, so the ratio rises for reasons that have nothing to do with collection. The third is reading by calendar month instead of by age. A cohort originated in January and one originated in October are at different points of their life in the same December, and comparing them compares maturity, not performance.
A lender builds it from two files it already has, and states what it counted
The calculation needs the loan tape for each loan's identifier, origination date and original principal, and the payment history for every receipt with its date, amount and type: principal, interest, fees, recoveries after default, and refinancing proceeds flagged separately. Group loans by origination month, sum receipts by month since origination, cumulate, and divide by the cohort's original principal.
Two lines of definition make the result usable by someone else. Say whether the numerator is all cash or principal only, because a principal-only multiple that never reaches 1.0x measures loss, while an all-cash multiple above 1.0x measures return. And say whether fees withheld at disbursement reduce the denominator. A reader who knows both can rebuild the number from the raw files, which is the first thing a fund does with any figure in a data room.
Questions this raises
- How do you measure collections from a loan tape?
- Group loans into cohorts by origination month, add up all cash each cohort has paid back by every month since origination, and divide by the principal it was originally lent. The result is a multiple of original principal, read at the same age across cohorts.
- What is the difference between collections per dollar lent and collection efficiency?
- Collection efficiency divides collections by amounts billed, and ICRA notes that a stricter write-off policy improves it because written-off contracts leave the billing. Collections per dollar lent divide by original principal, which a write-off cannot change, so they show how much came back per unit lent.
- Should renewals and refinancings count as collections?
- No. When a new loan repays an old one, the cash came from the lender, not the borrower. Counting it lifts the old cohort and hides a borrower who never repaid from income, so refinancing proceeds are flagged and kept out of the numerator.
- What does a collections multiple above 1.0x mean?
- At 1.0x a cohort has returned exactly the principal it was lent. Above 1.0x, the excess has to cover the lender's cost of funds, credit losses and operating costs. An all-cash multiple measures return, while a principal-only multiple measures how much of the principal was lost.
- What data does a lender need to calculate it?
- Two files: the loan tape, for each loan's identifier, origination date and original principal, and the payment history, for each receipt's date, amount and type, with principal, interest, fees, recoveries and refinancing proceeds kept apart.
Effect is not a bank, a lender, a broker-dealer, an investment adviser or a credit rating agency. Nothing here is investment, legal or tax advice. The cohort table is illustrative and describes no real lender. Every definition carries the version of the document it was read from.
Where every figure on this page comes from5 sources
- 01ICRA, Rating Methodology, Securitisation Transactions, June 2025, section on collection efficiency, read 6 October 2026.
- 02ICRA, Shriram Finance Limited, rating rationale for PTCs issued under two vehicle loan securitisations, payout month May 2025.
- 03Encore Capital Group, investor presentation filed with the SEC as exhibit 99.1, 2012: cumulative actual collection multiples by vintage year; and Q3 2023 investor presentation: purchase price multiple definition.
- 04Sridhar Sankranti, Axcess Financial, Approaches in Loan and Vintage Level ROI Forecasting, A Consumer Lending Industry Perspective, WUSS 2017.
- 05Effect, What a static pool is, 28 September 2026: why the denominator is the cohort's original principal.