Housing Loan Underwriting: The Signals Before Sanction
The signals that predict many future NPAs are already present before sanction. The challenge is finding them consistently.
The situation
India's housing lenders operate under a specific, well-drafted set of sanction-stage rules — a sanctioned building plan before a construction loan, disbursal linked strictly to construction progress, an independent valuer, verified title. The public record shows those rules being missed at the file level often enough, and recently enough, that the regulator had to remind lenders of one of them again in November 2025.
None of the tools already watching a housing loan file are built to catch this. A forged title or an uninspected valuation passes every bureau check and leaves the capital risk-weight calculation unchanged, because neither one looks at whether the underlying file was itself verified — bureau scoring measures a borrower's repayment history, and capital rules measure loan size.
Court and tribunal records document 112 cases where a sanction-stage check was missed, spanning nearly every major bank and housing finance company. Three failure types — a forged title document, an uninspected or inflated valuation, and a disbursal made without checking construction stage — are present in 64% of them. That is the exposure a board carries when the sanction decision is not independently measured, and it is narrow enough to act on.
Key findings
What the origination file already showed.
A housing loan file can pass every bureau check and every capital rule while carrying a forged title, an uninspected valuation, or unreleased construction funds — because those checks measure the borrower and the loan size, not whether the file's own documents comply with the sanction condition.
Across 112 court-documented failures spanning nearly every major bank and HFC, a forged title, an inflated or unvisited valuation, and a disbursal made without a construction-stage check together appear in 64% of the set — a pattern narrow enough to instrument.
In FY25, advances fraud totalled ₹33,148 crore — 92.1% of all bank fraud across every category. No equivalent figure exists for housing finance companies: neither the RBI nor the NHB publishes a fraud line for HFCs, in any format, in any year.
Upper-Layer HFC asset quality — the tier under the regulator's closest supervision — is worsening while every other lender tier improves, and there is no published fraud figure to explain what is driving it.
The three checkable steps — title verification, valuation inspection, construction-stage disbursal control — are not new requirements. The Reserve Bank published them in 2013; the NHB re-issued them in November 2025. The gap is not policy. It is consistency, and it is measurable.
Title, valuation and disbursal checks are the ones that fail most often.
Each bar is the share of the 112 court-documented cases in which that sanction-stage check failed. The three highest-leverage families — an inflated or unvisited valuation, a forged title, and a disbursal made without a construction-stage check — are the recurring spine of the failures.
Shares sum to more than 100% because a single file often failed more than one sanction-stage check.
In the largest single sanction-stage failure in this report's 112-case set, a housing finance company's own enterprise software was used to generate fictitious retail home-loan accounts at a non-existent branch. No field visit was made, no property was verified, no construction stage was checked. The accounts cleared internal processing because the checks that should have caught them — title verification, valuation inspection, and disbursal-stage confirmation — were not run. A Supreme Court judgment later stated plainly that “the underwriting procedures for loan sanctioning and disbursal were not followed.” The loss was not discovered through portfolio monitoring; it surfaced through a forensic audit, after the fictitious accounts had already grown to a scale the routine MIS had no mechanism to flag.
What's in the report
7 sections, every case cited to a public document.
The next step is in your own book.
We can run about twenty of your recent housing-loan files against these signal families — entirely on exported credit files, with no core-banking access and no customer data leaving the bank — and return a signal count, a policy-deviation map and board-ready evidence in about two weeks.
Supreme Court and High Court judgments, DRT and NCLT orders, RBI and NHB circulars and supervisory publications, and rated-pool performance data — every case and figure in the corpus tied to a citable public document. Full reference list in the report.