ZenLearnRead the report
ZenLearn Research · No. 05 · July 2026

Visible at Sanction

The weakness later blamed for the failure was, in most files, already visible in the origination file.

Read the report Request a briefing
19 of 30
Related-party lending — in failures
0 of 16
None of the comparable survivors
50in all
Cases analysed from the public record

The situation

Every large bank measures borrower quality — through scorecards, ratings and early-warning systems. Almost no bank runs an independent, calibrated read of the sanction decision itself: whether verification was followed, whether risk dissents were answered, whether the exposure was framed against the rules that already apply.

This study analyses 50 cases from the public adjudicated record — 34 adjudicated large-credit failures set beside 16 comparable large borrowers that did not fail. In the well-evidenced subset of 30 failures, leverage and group structure were as common in the survivors as in the failures. What separated them was related-party lending, present in 19 of 30 failures and 0 of 16 survivors, and it was legible in the origination file before the outcome.

That is the exposure your board carries when the sanction decision is not independently measured. This report reads the origination file — not the default — and shows that the discriminating weakness was on the public record before disbursement.

Key findings

What the origination file already showed.

Finding 01
Leverage did not separate failures from survivors.

Leverage appeared in 13 of 16 comparable survivors and only 5 of 30 documented failures. Group-exposure appeared in both groups. Neither one is the discriminator.

Finding 02
Related-party lending is the discriminator.

19 of 30 failures carried related-party lending; 0 of 16 comparable survivors did. Governance override — a credit committee sanctioning against its own risk team — carried the same 3-of-30 vs 0-of-16 shape.

Finding 03
It is rarely a single flaw.

26 of 30 failures fired two or more origination signals. Failures are typically a weak borrower framed by a weak exposure, not a lone red flag.

Finding 04
The leading signal shifts by lending segment.

Group-exposure led every infrastructure and real-estate project-finance case; assumption clustering led every trade-finance and working-capital case; related-party routing led the NBFC / financial-institution book. There is no single universal red flag.

Finding 05
The tell is an existing rule going unenforced at origination.

The related-party and group-exposure pattern sits against the RBI's 25% connected-group cap. The assumptions-and-leverage pattern sits against the Master Circular's viability-and-sensitivity standard. Every leading signal maps onto a rule the bank already holds.

The discriminator, in one exhibit

Only related-party lending is absent from every survivor.

Leverage and group-exposure show up in the survivors too, so neither one separates the two groups. Only related-party lending is absent from every survivor — the one signal that divides failure from survival.

Leverage — in failures (of 30)5Leverage — in survivors (of 16)13Group-exposure — in failures8Group-exposure — in survivors6Related-party lending — in failures19Related-party lending — in survivors0

Survivors are a comparable, deliberately-chosen set of 16 large borrowers that carried the same signals and did not fail — not a random sample.

One case, on the public record

Jaypee Infratech, the Yamuna Expressway developer, mortgaged 858 acres of its own land to secure the debt of its holding company Jaypee Associates. The Supreme Court held that transaction to be a preferential transfer under Section 43 of the Insolvency and Bankruptcy Code, reversing about 858 acres of land back into the insolvency estate. The origination file did not treat the pledge as related-party lending; the public record showed it was. This is the discriminator, in one file — a stable-entity asset routed to a distressed sister concern, on the record, before the default.

What's in the report

Ten sections, 43 pages, every case cited to a public document.

01
The origination file shows the weakness before sanction
How the study reads adjudicated records, and what it deliberately does not claim.
02
Why this matters to your bank today
The discriminator sits in a live origination file — not a post-mortem artefact.
03
Related-party lending is what discriminates — leverage is not
The survivor comparison, the frequency picture, and the segment shifts.
04
Three of the four judgment marks are read off the failures
Protocol Discipline, Risk Escalation, Regulatory Awareness — evidenced from the corpus.
05
One file shows the weakness was on the record before a rupee moved
The Jaypee Infratech land-pledge case, end to end, with the Supreme Court hook.
06
Every case and every rejection is named, so the basis is auditable
The screening funnel: 47 candidates screened, 34 analysed, 11 rejected, 2 pre-intake drops.
07
Every analytical decision is reproducible and auditable to source
The 11-signal ontology, 5-tier evidence grades, and the two-reviewer protocol.
08
What the framework expects, and what this cannot claim
The finding against the RBI framework, the international comparators, and the honest limits.
09
Practitioner heuristics show what to look for at sanction and after
Two working lists — pre-sanction signals from the file itself, and post-disbursement signals from money movement — each tied to the RBI Master Circular on Wilful Defaulters, the Large Exposures Framework, Ind AS 24, SEBI LODR and CARO 2020.
10
The next step is to measure the discriminator in your own book
The Independent Judgment Diagnostic — about 20 files, two reviewers, two-week readout.

The next step is in your own book.

We can benchmark about twenty of your recent large sanctions against this discriminator and return a reviewer-disagreement rate, the share firing two or more signals, and a segment-weighted benchmark — with an executive readout in about two weeks.

Read full report
Prepared by
Rohit Kumar · ZenLearn Research
Founder, ZenLearn Research · IIM Mumbai · Former Head of Business, Eko India
ex-CFO: Blackstone IARC, Pristyn Care, GE A&C SE Asia · Engine development, Tata Nano programme
Origination years 2005–2020; adjudication and public-record dates 2017–2025. Coded corpus as of 2 July 2026.
Contact: rohit@zenlearn.ai · zenlearn.ai/judgment
Primary sources

NCLT and NCLAT filings and orders, Supreme Court judgments, SEBI orders, CBI charge-sheets and RBI supervisory publications — every case in the corpus tied to a citable public document. Full reference list in the report.

Related research
Credit-Underwriting Safeguards
The demand-side companion — the safeguards the origination process needs to instrument.
Read
RBI Enforcement Research — FY25
Benchmark of RBI enforcement actions across BFSI in FY25.
Read
© 2026 ZenLearn Research · Judgment Intelligence
HomeJudgmentPrivacyTerms