Almost no NBFC misstates its NPAs deliberately. They misstate them because a rule was read once, years ago, implemented in the loan system by someone who has since left, and never tested again.
The problem is that provisioning errors compound. A classification rule applied wrongly does not produce one wrong account. It produces every account of that type, for as many years as the rule has been running — and when a supervisor finds it, the correction arrives in a single quarter.
1. Classification is a day-end position, not a month-end one
The requirement is that an account is flagged overdue as part of day-end processing on the due date, and the days-past-due count runs continuously from there. Classification as NPA happens on completion of 90 days past due.
Where systems are configured to stamp at month-end, or where the ageing is recomputed from a fresh start each month, accounts that touched 90 days mid-month never show up. The book looks cleaner than it is, consistently, in the same direction.
How to test it in an afternoon. Take fifty accounts that were overdue at any point in the last year. Rebuild the days-past-due from the transaction log yourself. If your system's figure and your own figure disagree on even a handful, the configuration is wrong for the whole book.
2. The upgradation rule is stricter than most people apply
This is the single most common finding, and it comes from an old habit. Many lenders upgrade an NPA the moment the borrower clears enough to bring the account under 90 days overdue.
That is not the position. A loan account classified as NPA may be upgraded to standard only when the entire arrears of interest and principal are paid. Part payment that merely reduces the overdue ageing does not do it.
The difference is not academic. Under the wrong reading, an account can move back and forth between standard and NPA repeatedly, resetting its provisioning each time, and never be recognised as the chronically stressed exposure it is. Under the correct reading, it stays NPA — with provisioning that increases as it ages — until it is genuinely cleared.
3. Restructuring that nobody called restructuring
When a borrower cannot pay and you extend the tenor, lower the instalment, grant a moratorium, or convert overdue interest into a fresh facility, that is a restructuring. It requires recognition, and it carries its own classification and provisioning consequences.
In practice these arrangements are frequently documented as ordinary amendments — a revised repayment schedule, a supplementary agreement — and never enter the restructuring register. The borrower's file shows the accommodation plainly. The classification does not.
This is the finding that most damages management credibility, because the records make it look deliberate even when it was not.
4. Evergreening through the top-up loan
A borrower is struggling. A fresh loan is sanctioned, ostensibly for a new purpose, and the proceeds service the old one. Nothing is overdue, so nothing is NPA.
This is visible in the data and supervisors look for it specifically. The markers:
- Disbursal and repayment on the same or adjacent dates
- New sanction amount close to the arrears on the existing facility
- A pattern of top-ups clustered at quarter end
- Repeat restructuring of the same exposure under different names
Where genuine additional finance is being given to a viable borrower, say so, document the viability, and classify accordingly. The transaction is defensible. Disguising it is not.
5. Provisioning on standard assets, quietly forgotten
Standard asset provisioning is not optional and is not covered by your expected credit loss model for regulatory purposes. Where you report under Ind AS, you are required to compare your ECL provisioning against the prudential floor computed under the RBI norms — and if the prudential figure is higher, the shortfall is appropriated to an impairment reserve, not left as a note.
NBFCs that moved to Ind AS and assumed ECL replaced the IRACP norms are a recurring category of finding.
6. Security value that was never re-tested
Provisioning on secured exposures depends on the realisable value of security. Valuations that are five years old, on machinery that has depreciated or property in a market that has moved, do not support the number.
Two questions decide this: when was it last valued, and by whom. A valuation obtained by the borrower, for the borrower, is not independent evidence of your recovery position.
What a shortfall actually does to you
Suppose an inspection concludes your provisioning is short by a figure equal to 12% of your net worth. Three things happen at once, and only the first is obvious:
- Capital falls. The provision is taken, profit reduces, net owned fund and CRAR move down together.
- Your covenants reprice. Lenders with net worth and CRAR covenants have a technical breach to consider, and the conversation about your next line of credit changes.
- Your reported numbers lose their standing. Every subsequent submission is read against the fact that your last one was wrong — which is why the second inspection is usually harder than the first.
A test worth running before anyone asks
Independently of the team that prepares it, take a sample across product types and rebuild the classification from source data under the Directions as written. Not under the practice you inherited.
Three outcomes are possible. Your number matches, and you have evidence you can show a supervisor. Your number is conservative, and you may be over-providing — which costs you real money. Or your number is short, and you have found it while you still have several quarters to absorb it, instead of only one.
All three are worth finding out. Only one of them is worth discovering from someone else.