Recovery & Write-offs
What happens after a loan is fully provided: write it off, then chase it via SARFAESI, IBC, or an ARC sale.
Why you care
Your NPA cluster covers the slide in. This is the exit. Once a loan is fully provided, the bank writes it off — takes it off the books. Write-off is not waiver: the borrower still owes the money, and the bank keeps chasing it via SARFAESI, IBC, or selling it to an ARC.
Run the numbers
A ₹100 cr loan, fully provided, is written off → GNPA drops ₹100 cr with no real improvement (illustrative). Two years later a SARFAESI sale recovers ₹35 cr — already provided for, so that ₹35 cr is pure write-back to profit.
Where this goes
Write-offs cut gross NPA — which is why a falling GNPA doesn't always mean a healthier book. Recoveries on written-off loans flow back as income, adding to net profit. A loan is only written off cleanly once provisioning fully covers it.
Why you care
Recovery and write-offs are the exit from the NPA lifecycle. The rest of the NPA cluster tracks a loan sliding into trouble — the 90-day line, the classification ladder, the provisions. This node is what happens at the bottom of that ladder, once a loan is fully provided and the bank stops pretending it will be repaid normally.
The single most misunderstood word here is write-off. A write-off is an accounting move: the bank removes the loan from its balance sheet because it has already set aside 100% provision against it, so keeping it on the books adds nothing. It is not a waiver. The borrower still legally owes every rupee, and the bank keeps chasing recovery. This distinction matters enormously in public debate — headlines about banks "writing off ₹X lakh crore of corporate loans" are routinely misread as banks forgiving those loans, which is not what happened.
The second thing to internalise: a falling gross NPA is not always good news. A bank can slash its GNPA in a quarter simply by writing off a pile of fully-provided loans — no borrower repaid anything, the underlying book is no healthier, but the ratio drops. Reading asset quality means checking why GNPA fell: genuine recovery and upgrades, or just housekeeping write-offs.
Run the numbers
Follow one ₹100 crore loan to the end (illustrative):
- It goes bad, ages down the classification ladder, and over time the bank provides the full ₹100 cr against it. Net carrying value: zero.
- The bank writes it off: GNPA falls by ₹100 cr overnight. No cash came in; the loan just left the books. The P&L feels nothing, because the pain was already taken through provisions.
- Two years later, the bank enforces its security under SARFAESI, sells the collateral, and recovers ₹35 cr.
That ₹35 cr is the interesting part. The loan was already fully provided, so every rupee recovered is "found money" — it flows back as income ("recovery from written-off accounts" is a real line in bank P&Ls) and lifts net profit. This is why a bank sitting on a big pile of written-off, well-secured loans has a hidden call option: recoveries in future years are almost pure profit.
The recovery toolkit, roughly in order of speed:
- SARFAESI (2002): for secured loans, the bank can seize and sell collateral without going to court.
- IBC / NCLT (2016): drag the defaulting company into insolvency; either a resolution plan or liquidation.
- ARC sale: sell the bad loan to an Asset Reconstruction Company at a discount, taking cash now and handing the recovery chase to a specialist.
- One-time settlement (OTS): negotiate a partial repayment to close the account.
Where this goes
Write-offs cut gross NPA — the reason a shrinking GNPA always deserves a second look. Recoveries on those written-off loans come back as income and lift net profit, the mirror image of the provisioning hit that put them there. And the whole move only works cleanly when provision coverage is high enough that the write-off costs the P&L nothing. That's why well-provisioned banks can clean up their books without a fresh hit.