What a lender asks for in due diligence, and what each number is meant to reveal
CS Sushil Choudhary · Operations · 2026-08-10 · 6 min read (estimated)
A lender or investor running diligence on a loan book asks for four things — collection efficiency, a static pool or vintage analysis, a bucket movement matrix, and a portfolio cut — and each is designed to defeat a different way a headline number can flatter. Being able to produce all four, consistently and quickly, is itself a signal about the book.
When a bank, an NBFC lender or an investor runs diligence on a lending book, the same four requests come up. They look overlapping. They are not — each is designed to defeat a different way a portfolio number can flatter.
Collection efficiency
Collections in a period over the demand raised for that period. Simple, and almost never comparable between two lenders, because the definition varies: whether arrears collected are counted in the numerator, whether prepayments and foreclosures are, whether the denominator is only current-month demand or includes overdue.
If you are asked for it, state the definition with the figure. If you are asking for it, ask for the definition — a 98% number computed one way and a 92% number computed another can describe the same book.
DCB
Demand, collection and balance, as three columns rather than one ratio. It is asked for because it cannot be collapsed: you can see the demand, what came in against it, and what is left, which makes the arithmetic visible instead of asking the reader to trust a percentage.
Static pool
Take every loan disbursed in one month and follow that fixed set as it ages. Because the set never changes, growth cannot flatter it.
This is the one that catches the most. A book growing at 8% a month can show a falling overall NPA percentage while every single cohort performs worse than the one before, simply because the denominator is filling with loans too young to have gone bad yet. Portfolio-level NPA is a ratio of two things moving at different speeds.
Vintage analysis
The same cohorts, arranged for comparison at the same age — each month's disbursement at six months on book, at twelve, at eighteen. It answers the question a static pool raises: is underwriting improving or deteriorating?
What this means for a lending system
All four are historical. They need the position as at past dates, not merely today's outstanding — which means the system has to retain month-end positions rather than reconstruct them. A book that can only report its current state cannot answer any of these questions, and the request usually arrives with a deadline.
This post describes market practice rather than a regulatory requirement. Definitions differ between lenders — collection efficiency especially — and nothing here should be read as the definition. Where you are asked for one of these, send the definition with the figure.
Why four, and why these four?
| Request | What it defeats |
|---|---|
| Collection efficiency | Nothing on its own — but the definition reveals how you count |
| Static pool / vintage | A growing book diluting its own deterioration |
| Bucket movement matrix | A flat NPA percentage hiding heavy flow in both directions |
| Portfolio cut | Concentration hidden inside an aggregate |
The bucket movement matrix is the one lenders under-prepare for. It asks: of the accounts in SMA-1 at the start of the month, how many are now standard, how many stayed, how many worsened? A book with 2% NPA and violent movement between buckets is a different risk from one with 2% NPA and none.
What does being able to produce these quickly signal?
That the underlying data exists as computed facts rather than as a monthly assembly.
A lender who can produce a vintage cut in an afternoon has classification history stored by account and date. One who needs three weeks is reconstructing it, and the reconstruction is being made by the same people who will answer the follow-up questions.
Diligence teams know this, and the time taken is read as information.
Common mistakes
- A collection efficiency number with no definition attached. Not comparable, and it looks like
it is.
- A vintage cut computed from today's classification. Cohorts need the classification as at each
age, not as at now.
- No bucket movement. The most revealing view, and the one most often absent.
- Portfolio cuts that do not reconcile to the total. Instantly corrosive.
- Different numbers in the board pack and the diligence pack. Both were produced honestly; one
used a different as-at date.
Frequently asked questions
What is a static pool analysis?
Accounts grouped by the period they were disbursed in, with performance tracked as the cohort ages — 3, 6, 12 months on book. It shows whether recent lending is behaving like older lending, which a headline portfolio number cannot, because growth dilutes deterioration.
How is collection efficiency defined?
There is no single definition, which is the point of asking. Whether arrears collected count in the numerator, whether prepayments and foreclosures do, and whether the denominator is only current-month demand — each choice moves the number materially. State the definition alongside the figure.
What is a bucket movement matrix?
A table of where accounts in each delinquency bucket at the start of a period ended up at the end of it. It reveals flow that a static percentage hides: a book can hold a constant NPA ratio while a large number of accounts deteriorate and a similar number recover.
How long should it take to produce a diligence pack?
Days, if the classification history is stored by account and date. Weeks means it is being reconstructed, and diligence teams read the time taken as information about the systems.
What do lenders look for beyond the numbers?
Consistency. The same figure in the board pack, the return and the diligence pack, computed as at the same date. A difference is not fatal, but being unable to explain it is.
Related reading: How to manage an NBFC loan portfolio · How to generate RBI returns · SMA-0, 1 and 2 · Classification is a day-end event
Ask for a walk-through of the four views against your own book.
A worked example
A book raises demand of ₹1 crore in a month, collects ₹98 lakh against it, and starts the month with ₹30 lakh of arrears brought forward, of which ₹10 lakh is recovered.
Against current-month demand alone, collection efficiency is 98%. Against demand plus the opening arrears, it is 83%. Both are arithmetically correct, and a diligence pack quoting the first without saying so is not.
That is the whole reason the definition has to travel with the number — and the reason a lender asks for the DCB alongside it rather than taking the percentage on its own.
Read next
- Co-lending: two lenders, one borrower, and two sets of books that must agree
The share, the classification and the customer interface all have to be unambiguous before the first rupee is disbursed. What each side owns.
- What is a loan origination system, and what does an NBFC need one to do?
What a loan origination system is, what each stage owns, and what separates an LOS an NBFC can be audited on from one it cannot.
- NBFC asset classification and provisioning: standard, sub-standard, doubtful, loss — and what you provide at each
When a loan turns standard, sub-standard, doubtful or loss — and how much has to be provided at each stage, with one loan followed through all of them.
From the people who wrote this
Run your lending on Lenviq
Seeing it run on your own book is faster than reading about it — a demo works through your products, your schemes and your classification rules, not a generic tour.
Lenviq is loan origination, servicing and accounting for NBFCs, built by FastLegal Technologies.