Everything you learned about valuing banks, and about the excess-return machinery, transfers. Value the equity directly, with justified P/B = (ROE − g)/(r − g). Treat value as book plus the present value of excess returns, and reverse the market multiple into the claim it implies. NBFCs need three calibrations of their own, and then comes the worked pair.
What changes when you value an NBFC instead of a bank?
Calibration 1: the discount rate carries the funding fragility. An NBFC's cost of equity should be higher than a comparable bank's. Its earnings stream is levered to a liability side that can reprice or vanish, and the ROE line doesn't show that risk until it does. In practice, building it bottom-up the way you built cost of equity in the valuation phase, you'll land at ~13% for a fortress franchise and 14 to 16% for wholesale-funded or mono-line lenders, as of mid-2026 conditions. Verify your inputs against current G-sec yields and your ERP method.
Advertisement
Calibration 2: growth is capital-hungry, so g and payout are chained. The identity is the same as for banks: sustainable g = ROE × retention. A 20%-ROE NBFC growing AUM 25% must either dilute or lever up. There's no third option. So when a fast grower trades on a story of "25% growth, no dilution, stable gearing," the three claims can't all be true. Work out which one management intends to break.
Calibration 3: book is a hypothesis, so adjust it before ratioing it. ABV = net worth − net Stage 3, the conservative convention you used for banks. For wholesale books, also subtract a haircut on the lumpy exposures you can see. Compare lenders on P/ABV and leave plain P/B aside, because lenders differ in how candidly they provision.
Now the worked pair: a franchise against a wholesale lender. Both are synthetic and illustrative, calibrated to recognizable archetypes.
Worked example: Falcon Finance, a franchise lender
Falcon Finance is the Bajaj-style franchise: ROA ~4.2%, gearing ~4.5×, ROE ~24%, AUM compounding ~20%, retention ~83% (so g ≈ 20%), a granular 100%-retail book, a liability stack of deposits/NCDs with CP <5%, LCR 130%. Falcon trades at 4.5× book. Is that sane?
Run the two-stage excess-return model at r = 13%, book₀ = 100. In Stage 1 (years 1 to 15), excess return = (24% − 13%) × book. It starts at 11 and compounds at 20%. The PV of that stream = 11 × [1 − (1.20/1.13)¹⁵]/(0.13 − 0.20) = 11 × 20.9 ≈ 230. For the terminal value, book₁₅ = 100 × 1.20¹⁵ ≈ 1,541. Fade to ROE 17% and g 5%, and terminal excess value = (17% − 13%) × 1,541/(0.13 − 0.05) ≈ 770, with PV ≈ 770/1.13¹⁵ ≈ 123. Add it up: 100 + 230 + 123 = 453 ≈ 4.5× book.
So the multiple isn't absurd. It's a precise claim: fifteen more years of 24% ROE with 20% compounding, then a durable 4-point spread forever. When you buy it, that's what you're underwriting. "Great company" won't cover it. You need two decades of moat. This is the reverse-reading habit at work. A franchise multiple is a statement about duration, and your job is to name the year the model assumes competition arrives, then ask whether the moat really holds that long. For an NBFC that moat is the liability franchise plus underwriting data plus distribution. Notice too what the model says about drawdowns. Cut stage-1 growth to 15% and ROE to 20% and the same arithmetic gives ~2.7× book. You've just taken a ~40% derating on assumptions that would still describe an excellent company. Franchise multiples are fragile to small story revisions, so size positions accordingly. Phase 9 will formalize this.
Worked example: Penna Capital, a wholesale lender
Penna Capital is the wholesale lender: reported ROE 16%, gearing 7.5×, 60% of the book in developer/structured credit, CP + sub-1-year NCDs = 40% of borrowings, trading at 0.9× book. The naive read takes justified P/B at ROE 16%, g 5%, r 14% = (16 − 5)/(14 − 5) = 1.22×. "It should trade at 1.2×, it trades at 0.9×, therefore 35% upside."
A professional runs two tests first. Start with book truth. Net worth is 4,000 against 18,000 of developer loans. Assume a modest stress, with 20% of the developer book defaulting at 50% LGD. The loss = 1,800, which is 45% of net worth, so ABV = 2,200 and the stock trades at 3,600/2,200 = 1.6× stressed book. Then test funding survival. Some 40% of borrowings reprice or mature within a year. A 2018-style freeze forces the death sequence from the ALM section, and a lender that has to shrink 30% to survive won't earn 16% ROE while doing it. It'll earn roughly nothing for two years.
Now the 0.9× reads correctly. What looks like a discount is the market quoting you a probability, the odds that the book is real and the funding survives. Sometimes that probability is mispriced, and wholesale lenders at 0.5× book have been great buys. The survivors of 2018-20 rerated 2 to 3×. But the entry ticket is the ALM note and the exposure file, whatever the P/B screen says. Cheap-on-stated-book is a hypothesis, never a fact.
Should you value an NBFC on P/E?
Where P/E fits. The market quotes premier retail NBFCs on P/E, at 25 to 35× for the franchises as of mid-2026, because their earnings behave like a consumer compounder's. Use P/E to communicate, and never as the anchor. Always translate back (P/E = P/B ÷ ROE) and check the implied book multiple against the excess-return arithmetic above. A 30× P/E on a 20% ROE is 6× book. Now ask the duration question with the tool that can answer it.