ROE compressed ~550 bps without a recession — but Q4 FY26 snapped back and the multiple still pays for the old regime.

• Three years of compression, no shock. ROE slipped from 23.5% in FY23 to ~18% in FY26; financing margin lost ~600 bps from its 39% peak. No COVID, no IL&FS, no demonetisation — spread compression and a mix shift toward lower-yielding mortgages and gold.
• The Q4 bounce is one print, not yet a regime. Q3 FY26 financing margin compressed to 28% on MSME stress, then snapped back to 35% in Q4 with PAT of $590M (+22% YoY). Management tightened the FY27 credit-cost guide to 145–160 bps. The cohort regression says each 200 bps of ROE compression strips ~1x off P/B.
• Two quarters decide the multiple. Today's 5.21x book and 31x P/E embed a recovery to 20%+ ROE. Q1 plus Q2 FY27 financing margin above 34% and credit cost inside the guided band, with no fresh 'exceptional' charge, are the conditions that would defend the multiple; otherwise the cohort math points toward the de-rating Macquarie's ~$8.96 (₹860) underperform target frames (Apr-2026 reiteration).

Book still compounds at 16% a year, but the four lines that mattered most are all bending the wrong way at the same time.

PAT growth decelerated from a decade-long ~30% CAGR to ~15% in both FY25 and FY26. Financing margin fell because cost of funds reset higher and the book shifted toward lower-yielding mortgages and gold; ROE compressed because equity grew faster than profit over the past three years. The 5.21x multiple holds only if FY27 delivers the guided 22–24% AUM, 19–20% ROE, and 145–160 bps credit cost — anything softer and the premium loses its anchor.

Three 'exceptional' provisions in seven quarters — credibility move or insiders bulletproofing a book they can see migrating?

• The mirror trade. Q3 FY26's $157M 'accelerated' Stage 1/2 ECL provision (LGD floor) was charged through P&L within 0.7% of a $158M BHFL stake-sale gain parked in consolidated reserves. No regulator or auditor compulsion. Stage 3 provisioning coverage lifted from 54% to 60% YoY.
• 'Exceptional' is the new recurring. Three of the last seven quarters carried 'exceptional' or 'additional' ECL charges — Q4 FY25 $42M, Q3 FY26 $157M, Q4 FY26 $15M.
• Capital signals are aligned. Dividend payout has roughly doubled from ~10% in FY15 to ~20% in FY26. The long-term ROE corridor was walked back from 21–23% to 19–21% in the FY25 strategy update; the 3% payments-GMV-share aspiration was cut to 1%.

Five stacked advantages no single NBFC peer can replicate — attacked simultaneously on the most material leg.

• The bundle. AAA funding, deposit licence, 119M customers (13× Cholamandalam, ~5× LTF), 1.01% GNPA against a 3–4% peer band, and the most diversified retail book in the cohort — four sector shocks survived without a single GAAP-loss year. Defence rests on simultaneity, not any one inimitable advantage.
• The bank-side breach. BFL's 7.41% cost of funds beats most NBFC peers by 80–130 bps — but HDFC Bank and ICICI Bank borrow at ~6.5%, a ~90 bps gap that has been widening as rate cuts favour deposit-funded banks. UPI plus Aadhaar e-KYC closed the digital-onboarding speed gap. The highest-yielding consumer segments are now underwritten by banks at structurally lower capital cost.
• The silent compounder. Opex/NTI fell from 58% in FY08 to 32.8% in FY26 (comparable basis) — a ~25-point step-down. Management has staked FY27 credibility on FinAI: 442 bots live, ~$215M originated by bots, target 800 autonomous agents. Each 100 bps of further compression adds ~30 bps of ROA — enough to defend ROE even if NIM never recovers.