alirezarezvani/claude-skills / finance/skills/stock-analysis
stock-analysis
UnverifiedAnalyze, research or value a listed stock or company (US, global, NSE or BSE) with verified data, ratios, scores and valuation. Use whenever asked if a stock is a buy, what it is worth, or how a company looks.
references/sectors/banks.md
View on GitHub# Banks and deposit-taking lenders — sector playbook
Use this when: the company holds a banking licence and funds itself with public deposits — Indian scheduled commercial banks (private, PSU, small finance, payments), US money-centre and regional banks, European and other global banks, and bank holding companies whose profit is dominated by a lending subsidiary.
A bank's balance sheet is its product, not its plumbing. Leverage is the business model, deposits are raw material, and reported profit is a function of provisioning judgement that management controls. That combination breaks almost every ratio in the generic checklist and inverts a few of them, so the sector needs a replacement metric set rather than an adjusted one. Two further facts govern everything below: the loss is written in the boom and recognised two to three years later, and banks do not die of slow insolvency — they die of a liquidity run in days.
If the entity does **not** take deposits (NBFC, HFC, fintech lender), use `references/sectors/nbfc.md` instead — the asset-quality toolkit is similar but the funding, liquidity and valuation logic differ materially.
## Contents
- [Why the generic ratio set fails here](#why-the-generic-ratio-set-fails-here)
- [The metrics that actually matter](#the-metrics-that-actually-matter)
- [How to value companies in this sector](#how-to-value-companies-in-this-sector)
- [Peer set construction](#peer-set-construction)
- [Sector-specific red flags](#sector-specific-red-flags)
- [Cycle and structural context](#cycle-and-structural-context)
- [India vs global notes](#india-vs-global-notes)
- [Checklist](#checklist)
## Why the generic ratio set fails here
Do not compute these for a bank. If a screener hands them to you, state that they are inapplicable rather than reporting them with a caveat — a reported number gets used.
**OPM / gross margin / EBITDA margin — undefined.** There is no COGS and no single "revenue" line. The Indian P&L (Banking Regulation Act Form B) shows "Total Income" = interest earned + other income; a margin computed on that gross number moves purely with the rate cycle and tells you nothing about the franchise. Interest expense *is* the cost of goods sold. Depreciation is trivial (typically 2–4% of opex), so EBITDA is an economically meaningless construct.
**EV, EV/EBITDA, net debt, Debt/EBITDA — undefined and never used.** Enterprise value assumes debt is financing you can strip out. For a bank, deposits and borrowings are the raw material, and cash / balances with RBI (CRR) and SLR securities are operating assets, not idle cash. Netting "net debt" subtracts the very franchise you are trying to value. Equity-side multiples only.
**D/E and interest coverage — mis-signed.** Healthy banks run 8–15x assets/equity *by design*. D/E of 10x is normal, not distress; a bank at D/E of 2 is under-earning its capital. The binding leverage constraint is regulatory (CET1, CAR, Basel III leverage ratio), not a covenant. Interest coverage is nonsensical when interest expense is an operating input.
**Free cash flow and FCF yield — noise, and inverted.** Operating cash flow is dominated by deposit inflows and loan disbursements, capex is immaterial. A fast-growing, high-quality bank shows deeply negative "FCF" because loans grew; a shrinking, de-risking bank shows strongly positive "FCF". An FCFF/WACC DCF cannot be built at all, because operating and financing cash flows cannot be separated.
**ROCE / capital employed / asset turnover — meaningless.** "Capital employed" would include every deposit. Asset turnover (income/assets) is just the asset yield, already captured better and more cleanly by NIM.
**Current ratio, quick ratio, working capital, inventory and receivable days, cash conversion cycle — undefined.** Bank balance sheets are not classified current/non-current (India: Form A under the BR Act, exempt from Schedule III classification). Liquidity is measured by LCR, NSFR and ALM buckets instead.
**P/E used standalone — dangerously procyclical.** Bank earnings peak when provisions are at a cyclical trough, so P/E looks cheapest exactly at the top of the credit cycle — Indian PSU banks in FY18–19, US regional banks in 2006–07 and again in 2022. Reported "E" is also the most management-discretionary number in the market: provisioning, NPA recognition, write-off timing and ECL/IFRS 9 staging assumptions all sit inside it.
**Standard three-step DuPont — must be re-specified.** The bank version decomposes ROA into NIM + fee income − opex − credit cost − tax, all as a % of average assets, then ROE = ROA × equity multiplier. Generic DuPont produces a "profit margin" on gross income that is not comparable across banks with different funding mixes.
**Growth as an unqualified positive — inverted.** In most industries revenue growth is good news. In lending, above-system loan growth is the single best leading indicator of the next credit cycle's losses. Treat sustained hypergrowth as a risk factor to be explained, not an achievement to be scored.
## The metrics that actually matter
Ranges are **indicative only**. They vary by market, by sub-sector, by rate cycle and by period, and any of them can be wrong for a specific bank in a specific year. Comparison against a tight peer set and against the bank's own 5–10 year history overrides every absolute band below.
| Metric | Definition / how to compute | Indicative healthy range | Why it matters |
|---|---|---|---|
| **Net Interest Margin (NIM)** | (Interest earned − interest expended) / average interest-earning assets, annualised. US banks usually report tax-equivalent (FTE); some Indian banks use average *total* assets — check the base before comparing. | India large private 3.5–4.5%; India PSU 2.6–3.2%; small finance banks 6–9%; US large/regional 2.8–3.6%; Europe 1.2–2.0%. Stability matters more than level — a swing above 30–40 bps in a year needs an explanation. | The core spread engine, the closest thing to a gross margin. Tells you whether profit comes from a genuinely cheap deposit franchise (durable) or from lending down the credit curve at high yields (rented, repaid later as credit cost). NIM rising on unsecured/MFI mix shift is a different quality of earnings from NIM rising on CASA. In India, zero-yield CRR and SLR holdings structurally cap NIM vs DM peers. |
| **CASA ratio / non-interest-bearing mix and deposit beta** | Current + savings balances as % of total deposits (India). DM equivalents: share of non-interest-bearing demand deposits, and *deposit beta* — the fraction of a policy-rate move passed into deposit cost. Track with cost of funds and the share of bulk deposits and CDs. | India: >40% strong, 30–40% average, <30% weak (SFBs and newer banks 20–35%). US: non-interest-bearing >25% of deposits is a strong franchise; cumulative deposit beta below 40–45% through a hiking cycle is good. | The deposit franchise, not the loan book, is what a durable bank multiple is paid for. CASA is sticky, cheap and slow to reprice, so a high-CASA bank keeps NIM when rates rise and can undercut on price when they fall. Falling CASA replaced by bulk deposits or CDs is an early, quantitative sign the franchise is being funded by paying up; margin compression follows with a lag. This is the number that separated survivors from casualties in the 2023 US regional bank episode. |
| **Return on Assets (ROA)** | Net profit / average total assets. Decompose on a % of average assets basis: NIM + fee & other income − opex − credit cost − tax. | India private 1.6–2.2% top quartile, 1.0–1.5% acceptable, <0.8% weak; India PSU 0.8–1.2% is now considered good. US >1.2% strong, ~1.0% average. Europe 0.5–0.8%. | The only clean cross-bank profitability comparison, because it is leverage-neutral. Two banks can both post 16% ROE — one from 1.6% ROA at 10x leverage (well capitalised), the other from 0.9% ROA at 18x (levered to the regulatory ceiling, one credit shock from a dilutive raise). ROA also sets the self-funded growth ceiling: ROA × leverage × retention = loan growth achievable without issuing equity. |
| **ROE / Return on Tangible Equity (ROTE)** | Net profit / average shareholders' equity; ROTE strips goodwill and intangibles from the denominator. Always read against cost of equity. | 15–18%+ excellent, 12–15% good; below COE (roughly 12–14% India, 9–11% US/EU) means value is destroyed at the margin. | ROE vs COE is the single input that determines the justified P/B, so it drives valuation directly rather than being a standalone score. Read with the capital ratio: ROE lifted by running CET1 down toward the minimum is borrowed, not earned. ROTE is the relevant form for banks grown by acquisition, where reported book is inflated by goodwill. |
| **Gross and Net NPA (Stage 3 / NPL in DM)** | Gross non-performing advances / gross advances; NPAs net of provisions / net advances. India uses RBI IRAC norms — 90 DPD, hard rules. DM uses IFRS 9 Stage 3 or US GAAP non-accrual, judgement-based, plus Stage 2 as the watchlist. | India: GNPA <2.5% and NNPA <0.7% strong; GNPA >5% or NNPA >1.5% needs a full workout thesis. DM: NPL <1.5–2%; Stage 2 <8–10% of the book and, critically, not rising. | The stock of recognised damage — and the *least* informative asset-quality metric, because it is the most manipulable. Technical write-offs shrink the numerator; rapid loan growth inflates the denominator; so GNPA% can fall while the absolute problem grows. Never read without slippages, credit cost and write-offs alongside. In India, always reconcile to the RBI divergence disclosure. |
| **Slippage ratio (fresh NPA formation)** | Fresh slippages into NPA in the period / opening standard advances, annualised. DM analogues: gross NPL inflows and net charge-offs (NCO) as % of average loans. | India: <1.5% healthy, 1.5–2.5% watch, >3% stress. US NCO: 0.3–0.6% for a normal-cycle retail/commercial bank; card-heavy books run 3–5% by design. | The flow, not the stock — and flows turn first. GNPA can be flattered by recoveries, upgrades and write-offs; slippages cannot. Flat GNPA with rising slippages means the bank is running to stand still and is a quarter from a negative surprise. Earliest hard number saying the last 2–3 underwriting vintages were too loose. |
| **Credit cost** | Loan-loss provisions (plus write-offs, for the honest version) / average advances, annualised. Compare to the bank's own through-cycle average, never to zero. | Through-cycle 0.6–1.2% for a diversified Indian bank; 0.3–0.5% for a mortgage-heavy secured book; 4–6% for cards/microfinance. Sustained sub-0.4% on a diversified book is a cyclical trough, not an achievement. | The swing factor in bank earnings and the main lever for managing reported profit. The correct analytical move is to **normalise**: re-run EPS at through-cycle credit cost *before* applying any multiple. Most bank value traps are the market capitalising trough credit cost as if structural. Also check credit cost net of releases from prior contingent/floating provisions — releases are non-recurring earnings. |
| **Provision Coverage Ratio (PCR)** | Provisions held against NPAs / gross NPAs. Insist on the figure **excluding** technical and prudential write-offs; the including-write-offs number Indian banks headline is far more flattering. | India: >70% ex-write-offs is the RBI-anchored comfort level, >75% strong, <50% means unrecognised losses sit inside book value. DM: Stage 3 coverage 40–60% typical (lower, because collateral is marked more explicitly). | The buffer between reported book value and real book value. A bank letting PCR drift down while GNPA is flat is releasing reserves into the P&L — earnings quality decays while EPS grows. Look for floating/contingent provisions above the regulatory requirement: a genuine hidden reserve and a marker of conservative management. |
| **Cost-to-income (efficiency ratio)** | Operating expenses / net total income (NII + other income). US "efficiency ratio" is computed essentially the same way. | Best-in-class 35–42%; 42–50% good; 50–60% for banks in build-out (SFBs, new private banks, digital scale-ups); >65% structurally sub-scale. Three-year direction matters more than the level. | The operating-leverage metric that replaces OPM — does branch and technology spend convert into revenue? Watch the denominator: a bond-trading windfall or a large one-off recovery produces an artificially good ratio that quietly reverses. Recompute excluding treasury gains. |
| **CET1 and CAR/CRAR** | Common Equity Tier 1 and total regulatory capital as % of risk-weighted assets under Basel III. Read with RWA density (RWA / total assets) and the Basel III leverage ratio. | India (RBI): minimum CAR 11.5% including CCB, CET1 minimum 8%; comfortable is CET1 >13% and CAR >16%. DM: CET1 >12–13% with a clear buffer above the MDA trigger; US large banks manage to SCB-adjusted requirements +50–100 bps. | This, not D/E, is the leverage constraint that actually binds. Capital determines how much a bank can grow before it must issue equity — and equity issued below book permanently destroys per-share value. Falling RWA density with a growing book can be a genuine mix shift to mortgages, or IRB model optimisation. Check how much "capital" is deferred tax assets and revaluation reserves rather than retained earnings. |
| **Credit-deposit ratio; deposit growth vs credit growth** | Advances / deposits. Track the YoY loan-growth minus deposit-growth gap, and funding from CDs, refinance (NABARD/SIDBI/NHB in India) and wholesale borrowing. | India: 75–85% comfortable; >90% is a funding-stress signal (the system CD ratio near ~80% in 2024 is what triggered RBI push-back on unsecured growth). US: <90–100% generally, though securitisation and FHLB usage change the model. | The fastest read on whether growth is funded by franchise deposits or borrowed money. A bank persistently growing loans 5–8 points faster than deposits is about to pay up for deposits (NIM compression) or raise capital. In India it feeds directly into marginal cost of funds and hence lending-rate competitiveness. |
| **LCR, NSFR, ALM buckets, HTM unrealised losses** | LCR = HQLA / 30-day net stressed outflows. NSFR = available vs required stable funding. Alongside: the ALM statement's 1–30 day and 1-year buckets, and the mark-to-market hole in held-to-maturity securities. | Both >100% regulatory minimum; comfortable operating LCR 120–140%. Unrealised HTM losses small relative to CET1 — above 15–20% of CET1 deserves serious scrutiny. | Banks fail from a run, not from slow insolvency. Silicon Valley Bank was profitable and passed its headline capital ratio while carrying an HTM loss exceeding tangible equity, funded by uninsured deposits. For Indian banks, watch uninsured deposit share (DICGC covers ₹5 lakh per depositor) and top-20 depositor concentration. |
| **Loan mix and concentration** | Advances split by product (mortgage, vehicle, personal/cards, microfinance, MSME, corporate) and by sector; top-20 borrower exposure as % of net worth; share in the riskiest 2–3 sectors; geographic concentration. | Unsecured retail (PL + cards + MFI) above ~20–25% of the book warrants a higher normalised credit cost. Top-20 borrowers well below 100% of net worth for a diversified bank. CRE above ~250–300% of capital is the classic US regional flag. | Two banks with identical NIM and identical current GNPA can have completely different loss distributions. Mix tells you what credit cost normalises to and how fast the book can deteriorate: unsecured retail goes bad in 2–4 quarters, corporate in 2–3 years, microfinance can go from pristine to 10% GNPA in two quarters on a local political or weather shock. Concentration converts an idiosyncratic default into a capital event. |
| **Operational productivity KPIs** | Business (deposits + advances) per employee and per branch (standard Indian annual-report disclosures); deposits per branch and branch vintage mix; digital transaction share and active mobile/net-banking users; new-to-bank additions and acquisition cost; cross-sell ratio (products per customer); monthly collection efficiency for retail/MFI books. | Business per employee ₹15–25 crore strong for Indian private banks, ₹10–18 crore typical for PSUs. Deposits per branch rising in real terms. Collection efficiency >98% on a secured retail book; below 96% is an early warning. Digital share of transactions >90% for a modern retail bank. | These are the leading indicators sitting behind the financial ratios. Deposits per branch tells you whether branch expansion is earning its cost before it shows in cost-to-income. Collection efficiency turns 1–2 quarters before slippages, which turn 2–3 quarters before GNPA — the earliest asset-quality signal available. Cross-sell and fee per customer distinguish a primary-banking relationship from a rate-shopping deposit book. |
| **Fee / non-interest income quality** | Split other income into core fees (transaction banking, card interchange, third-party distribution of insurance and MFs, forex, processing) vs treasury/trading gains, recoveries from written-off accounts, PSLC income and stake sales. Express core fees as % of average assets. | Core fee income 1.0–1.5% of average assets is strong for an Indian retail bank; treasury gains a small single-digit % of PBT in a normal year. | Fee income is capital-light and multiple-accretive *if* annuity-like and customer-driven; worthless if it is bond profits from a falling-rate quarter or lumpy recovery income. Many banks bridge a weak quarter this way and still report "PAT growth". Strip both out and use **pre-provision operating profit ex-treasury** — the number that shows what the franchise earns before management discretion. |
| **SMA-1 / SMA-2 and Stage 2 pipeline** | India: special mention accounts, 31–60 and 61–90 DPD, disclosed in investor presentations and (for large exposures) reported to CRILC. DM: IFRS 9 Stage 2 balances and the migration matrix between stages. | SMA-2 low and falling; Stage 2 <8–10% and not rising. Judge the trend, not the level — definitions differ too much across banks for a hard band. | The pipeline into NPA, one step earlier than slippages. Rising SMA-2 with stable headline GNPA is the classic pre-break configuration. This disclosure is voluntary in places: a bank that *stops* publishing an SMA breakdown it used to publish is telling you something. |
| **Book value per share growth and dilution history** | BVPS (and tangible BVPS) CAGR over 5–10 years, alongside every equity raise and the P/B at which it was done; add back dividends for a total value-creation view. | BVPS compounding at or near ROE × retention, with no raises below ~1x book. | Aggregate book value growth flatters serial issuers. BVPS growth is the per-share truth, and raising equity below book is permanent, irreversible value destruction for existing holders. A bank that repeatedly issues at low P/B is telling you its ROA cannot fund the growth it reports. |
## How to value companies in this sector
Banks are valued on the **equity side only** and anchored to book value. Every EV- or FCFF-based method is structurally inapplicable.
**Primary — P/B, and P/TBV in developed markets.** Book value is a reasonable proxy for economic capital: bank assets are mostly financial and near fair value, and book is the regulatory constraint on how much business can be written. Use P/TBV wherever acquisition goodwill is material (US regionals, European banks); in India, where organic growth dominates, plain P/B is the convention.
The multiple is not a free parameter. It is anchored by the warranted-multiple identity:
> **Justified P/B = (ROE − g) / (COE − g)**
A bank earning 17% ROE with 12% growth against a 13% cost of equity justifies roughly 5x book; the same bank at 11% ROE justifies barely 1x. The practical implementation is a cross-sectional ROE-vs-P/B (or ROTE-vs-P/TBV) scatter regression across the peer set, with the residual explained by deposit-franchise quality, asset-quality track record and management. This is exactly why deposit quality translates so directly into multiple: it lowers COE and raises sustainable ROE simultaneously.
**India-specific — Price to Adjusted Book Value (P/ABV).** Adjusted book = net worth − net NPAs (or − the shortfall to a 70%+ coverage standard), sometimes also net of unprovided restructured assets and security receipts from ARC sales. Essential for PSU banks and any bank with a stressed book, because reported book contains losses recognised as NPAs but not yet provided. Two banks at 1.2x reported book can be at 1.3x and 2.4x adjusted book — the entire investment case can sit in that adjustment. Also deduct revaluation reserves and large deferred tax assets: book value that cannot absorb losses.
**Sum-of-the-parts — essential for large Indian banks.** Indian banking groups carry material listed and unlisted subsidiaries (life and general insurance, AMC, broking, NBFC arms). Value the standalone bank on P/ABV, value each subsidiary on its own sector metric (embedded-value multiple for life insurance, % of AUM or P/E for the AMC, P/B for the NBFC), apply a 15–25% holding-company discount, and back out the *implied standalone bank multiple*. Skipping this makes the consolidated P/B look expensive when the core bank is not.
**Excess return / residual income.** Value = current book value + PV of (ROE − COE) × book value each future year. This is the theoretically correct intrinsic model for banks: it works directly with the accounting book that regulation binds against, and it degrades gracefully because most of the value sits in today's book rather than a distant terminal value. It also makes the ROE–P/B link explicit.
**DDM / FCFE.** If a cash-flow model is required, define FCFE as net income minus the increase in regulatory capital needed to support RWA growth — earnings distributable *after* funding growth at target CET1. A three-stage DDM is standard for mature DM banks with stable payout ratios. Never a WACC-based DCF: WACC needs a cost of debt that, for a bank, is a cost of raw material.
**Secondary — normalised P/E.** Usable only after credit cost is normalised to a through-cycle level and treasury gains and recovery income are stripped. Reported P/E on trough provisions is the sector's classic value trap. Indian banks typically 12–20x normalised earnings, DM banks 8–13x — indicative, and cycle-dependent.
**M&A / control valuation.** Price per core deposit, price-to-deposits, and the core deposit intangible premium are the standard US bank M&A metrics: the acquirer is buying a funding franchise, not a loan book. In India, distressed bank resolutions have been done at or below adjusted book, with the acquirer effectively compensated for taking the liabilities.
**Cross-checks.** Dividend yield and payout sustainability against the stated capital plan; deposit market-share trend; and a reverse-valuation test — solve for the normalised credit cost (or the ROE) implied by today's multiple, then ask whether that number is plausible against the bank's own history and mix.
**Do not use:** EV/EBITDA, EV/Sales, EV/EBIT, FCF yield, PEG on reported EPS, or any WACC-discounted enterprise cash-flow model.
## Peer set construction
A valid comparable shares the *funding model, regulatory regime and asset mix* — not merely the label "bank". Mixing across these lines produces confidently wrong conclusions.
**Splits that must not be mixed:**
- **Indian private vs PSU banks.** Different cost of funds, different opex structures, different capital access (government recapitalisation vs market raises), different governance and appointment processes, and structurally different multiples. A PSU trading at 1.0x book is not "cheap versus" a private bank at 3x.
- **Universal / commercial banks vs small finance banks vs payments banks.** SFBs run 6–9% NIM, high credit cost and 50–60% cost-to-income by design; payments banks cannot lend at all and are fee/float businesses. Neither belongs in a large-bank scatter.
- **Banks vs NBFCs and HFCs.** No deposit franchise, no CASA, no CRR/SLR drag, no LCR regime in the same form — different funding risk and different multiple logic. Compare only on asset-quality and ROA discipline, never on P/B directly.
- **US money-centre / GSIB vs regional vs community banks.** GSIBs carry markets, trading and asset-management businesses and are valued partly on those; regionals are spread-and-CRE businesses. Different capital regimes (SCB and G-SIB surcharge vs Category IV).
- **Retail-led vs corporate-led vs treasury-heavy books.** Loss timing and volatility differ so much that identical GNPA today implies different futures.
- **Microfinance-heavy and unsecured-heavy lenders** deserve their own bucket — the loss distribution is fat-tailed and correlated to local events.
- **Consolidated vs standalone.** For groups with insurance and AMC arms, compare standalone bank to standalone bank, then handle subsidiaries in the SOTP. Comparing one bank's consolidated P/B against another's standalone is a common and material error.
**Also align:** fiscal year end (Indian banks are April–March, most US banks are calendar); accounting regime (Ind-AS/IGAAP-for-banks vs IFRS 9 vs US GAAP CECL — provisioning is not comparable across these); size band (a ₹50,000 crore balance sheet and a ₹20 lakh crore one face different growth ceilings); and stage of build-out (a bank three years into a branch expansion will look bad on cost-to-income for structural, not quality, reasons).
Aim for 5–8 peers. State the basis explicitly in the report, and benchmark every metric twice — against peers *and* against the bank's own 5–10 year record.
## Sector-specific red flags
- **Loan growth far above system (roughly 1.5–2x system credit growth) for several years, concentrated in one product.** The bad loan is written in the boom and recognised two to three years later. Almost every Indian NPA cycle (infrastructure 2010–13, unsecured retail and microfinance 2022–24) and every US regional failure was preceded by exactly this pattern.
- **RBI divergence in asset classification (India, non-negotiable).** SEBI requires disclosure of divergence from RBI's Risk Assessment Report where it exceeds 10% of reported PBT or 15% of reported GNPA. Any material divergence means management's own NPA numbers were wrong — the single highest-signal governance flag in Indian banking, and one that was raised repeatedly before at least one prominent private bank failure.
- **Falling provision coverage while GNPA is flat or rising.** Reserve releases manufacturing earnings growth. Always ask whether PAT growth came from pre-provision operating profit or from a lower provision line; the second kind reverses.
- **Heavy technical write-offs managing the headline ratio.** Reconcile: opening GNPA + slippages − recoveries − upgrades − write-offs = closing GNPA. If write-offs are doing most of the work, asset quality is not improving. The denominator trick is the twin: a rapidly growing book can drop GNPA% for two years while absolute GNPA rises.
- **Evergreening and disguised restructuring.** India: lending to a stressed borrower through an NBFC arm or co-lending partner, routing exposure via Alternative Investment Funds (RBI cracked down in December 2023), rolling working-capital limits to avoid a 90-day trigger, capitalising unpaid interest. DM analogue: IFRS 9 stage-migration management and "covenant amendment" extensions that keep Stage 2 out of Stage 3.
- **ARC sales where the bank retains security receipts.** The loan leaves the NPA line; the credit risk stays on the balance sheet as an investment. Check the outstanding SR book, its rating trajectory and its provisioning — the standard cosmetic tool of the 2015–18 Indian cycle.
- **Profit dependent on treasury gains, written-off-account recoveries, PSLC income or subsidiary stake sales.** Strip them out and look at pre-provision operating profit ex-treasury. A "record quarter" produced by a bond rally is not an improvement.
- **Deposit franchise deterioration.** Falling CASA, rising bulk deposits and CDs, deposit rates visibly above peers, rising top-20 depositor concentration, high uninsured-deposit share. This is a bank buying its funding; it precedes NIM compression and, in the tail case, a run.
- **ALM and duration mismatch with unrealised HTM losses.** Long-duration securities or fixed-rate loans funded by short, rate-sensitive deposits. Measure the HTM mark-to-market shortfall against CET1 — the loss is not routed through capital for many banks, which is precisely why the headline ratio can look fine while economic equity is gone.
- **Concentration.** Single-sector exposure (CRE above roughly 250–300% of capital is the classic US regional flag), a large top-20 borrower book relative to net worth, or geographic concentration in microfinance/agri lending exposed to one state's political cycle or monsoon.
- **Related-party and promoter-linked lending, and lending into the CEO's network.** This is the recurring cause of outright *failure* rather than mere underperformance across Indian private, co-operative and small banks. Read the related-party transactions note in full, every year.
- **Governance and regulatory signals.** RBI declining to extend an MD/CEO term, or granting a shorter term than the board sought; supervisory action such as a ban on new digital onboarding or new customer acquisition; monetary penalties for IRAC or KYC breaches; abrupt auditor resignation or statutory auditor change; sequential CFO or Chief Risk Officer exits; a CRO whose reporting line does not reach the board. These are visible well before the numbers break.
- **Repeated equity raises at or below book value.** Permanent per-share destruction, and a signal that ROA is too weak to self-fund reported growth. Track BVPS growth, not book value growth.
- **Book value quality.** A large share of net worth in deferred tax assets (worth something only if future profits arrive), revaluation reserves on premises, or acquisition goodwill. Recompute tangible book and re-run the multiple.
- **Poor-quality fee income.** Upfront recognition of loan processing fees, insurance mis-selling-driven distribution income, fees that vanish when disbursements slow. If fee income moves one-for-one with disbursement growth, it is not annuity income.
- **Rising SMA-1/SMA-2 or Stage 2 while headline GNPA is stable** — the pipeline, disclosed selectively. Withdrawal of a previously given disclosure is itself the signal.
- **Thinly provisioned restructured/standard-restructured book**, or a large "standard asset" exposure to a stressed group that peers have already classified as NPA.
- **Aggressive ECL assumptions (DM).** Optimistic macro scenario weights, extended cure periods, stale collateral valuations. Indian banks moving to RBI's expected credit loss framework face a one-time transition provision; a bank that has not quantified and disclosed its expected day-one ECL impact is deferring a known hit.
## Cycle and structural context
**Where you are in the credit cycle determines which metric lies to you.** Early cycle: credit cost is normalising down, GNPA falling, earnings growth flattered by provision reversals, P/E optically low — this is where value traps are bought. Late cycle: credit cost above trend, book value already impaired, P/B optically low, but the adjusted book is the real question. The correct posture is to hold credit cost at the through-cycle level regardless of where the reported number sits, and to explain the gap.
**Rate cycle.** Asset repricing is faster than liability repricing on the way up (NIM expands, then deposit beta catches up and NIM gives it back) and the reverse on the way down. In India, the share of loans on external benchmark-linked rates (EBLR, mostly repo-linked) versus MCLR determines how fast the asset side reprices — repo-linked books reprice almost immediately, MCLR books with a lag. A bank whose NIM expansion was purely mechanical repricing has not improved.
**Structural threats to score explicitly.** Deposit disintermediation into mutual funds, equity and higher-yielding alternatives — a live pressure in India, where retail savings have been shifting toward markets, forcing banks to compete harder for deposits. Fintech and NBFC capture of high-margin unsecured lending and payments. UPI making payments free and eroding transaction fee pools in India (with interchange economics far weaker than card-led markets). Digital-first challengers attacking the branch cost model. And, longer-term, account aggregator frameworks and open banking commoditising the customer relationship that CASA depends on.
**Regulation is a first-order value driver, not a footnote.** In India: CRR and SLR requirements (a direct NIM drag), priority sector lending targets and the PSLC market, risk-weight changes (the November 2023 increase on unsecured consumer credit and bank lending to NBFCs immediately repriced capital consumption), the LCR framework revision on digitally-accessible deposits, the coming ECL transition, and RBI's discretionary supervisory powers — which can halt a growth engine overnight without any legal process visible in the filings. In DM: Basel III endgame capital changes, stress-test regimes (CCAR/SCB in the US, EBA in Europe), deposit insurance reform after 2023, and the MDA trigger constraining distributions. Always check what regulatory change is in flight before extrapolating current ROE.
**Consolidation and licensing.** Indian bank licences are scarce and consolidation is state-directed for PSUs; US banking is fragmented with thousands of institutions and consolidation is the base case, which makes deposit franchises acquirable and supports takeout value for small, well-funded banks.
## India vs global notes
| Dimension | India | US / global |
|---|---|---|
| Regulator / supervisor | RBI (licensing, IRAC norms, supervisory action, MD/CEO approval); SEBI for disclosure; DICGC for deposit insurance (₹5 lakh per depositor) | Fed / OCC / FDIC (US), ECB-SSM and national regulators (EU), PRA (UK); FDIC insurance $250k per depositor per category |
| Statement format | Banking Regulation Act Form A (balance sheet) and Form B (P&L); exempt from Schedule III current/non-current classification; Ind-AS applies to most corporates but banks still follow RBI's prescribed formats and have not yet transitioned to full ECL | 10-K / 20-F on EDGAR; US GAAP with CECL (lifetime expected loss at origination, adopted 2020) or IFRS 9 three-stage ECL |
| NPA recognition | Rule-based IRAC: 90 DPD, hard classification into substandard/doubtful/loss with prescribed minimum provisioning. Less discretion, more comparability across banks | Judgement-based ECL staging and macro scenario weights; more forward-looking but far less comparable between banks |
| Distinctive disclosures | RBI divergence disclosure; SMA-1/SMA-2 and CRILC reporting; business per employee and per branch; PSL achievement and PSLC income; restructuring under RBI resolution frameworks; CARO does not apply to banking companies (they are exempt) but the LFAR — Long Form Audit Report — and the auditor's report do | Call Reports and FR Y-9C (US, granular and public); CCAR/DFAST stress test results; uninsured deposit disclosure; AOCI and HTM fair-value footnote |
| Units and reporting | ₹ crore and lakh; fiscal year April–March; quarterly results with an investor presentation and an analyst concall — treat the concall Q&A as a primary source for slippage guidance, restructuring and deposit strategy | $ millions/billions; calendar fiscal year for most banks; quarterly 10-Q, earnings call transcript, and a supplemental financial package with segment-level NIM and credit detail |
| Ownership | Promoter/founder shareholding matters and RBI caps voting rights and ownership in private banks; government holding dominates PSU banks and drives recapitalisation and appointment decisions; foreign ownership limits apply | Widely held; activist and institutional ownership matter more than a promoter concept; bank holding company structure is standard |
| Valuation convention | P/B and P/ABV; SOTP for groups with insurance/AMC arms; adjusted book is the standard analyst construct | P/TBV and ROTE dominate, because acquisition goodwill is pervasive; price per core deposit for M&A |
| Structural NIM | Higher gross yields but CRR (non-interest-bearing) and SLR requirements are a permanent drag; PSL obligations force lower-yield lending or PSLC purchases | No comparable reserve drag at current settings; NIM lower because loan yields and inflation are lower |
## Checklist
- [ ] Confirm it is a deposit-taking bank, not an NBFC/HFC — route to `nbfc.md` if not.
- [ ] Delete OPM, EBITDA, EV/EBITDA, net debt, D/E, ROCE, FCF, current ratio from the analysis and say why in the report.
- [ ] Build the bank DuPont: NIM + fees − opex − credit cost − tax = ROA; ROA × leverage = ROE.
- [ ] Check NIM level, trend and *composition* — CASA-driven or yield-driven? Confirm the denominator base before comparing to peers.
- [ ] Assess the deposit franchise: CASA%, deposit beta, bulk/CD share, cost of funds trend, top-20 depositor and uninsured-deposit concentration.
- [ ] Read the asset-quality flow, not just the stock: slippages, recoveries, upgrades, write-offs, and the GNPA reconciliation. Then SMA-2 / Stage 2.
- [ ] Recompute PCR excluding technical write-offs; check whether coverage is drifting down while GNPA is flat.
- [ ] Normalise credit cost to a through-cycle level and re-run EPS before applying any multiple.
- [ ] Strip treasury gains, recoveries and one-offs; compute pre-provision operating profit ex-treasury and judge growth on that.
- [ ] Check CET1, CAR, RWA density, and how much capital is DTA/revaluation reserve rather than retained earnings.
- [ ] Compare loan growth to system growth and to the bank's own deposit growth; flag any multi-year gap.
- [ ] Check LCR, NSFR, ALM 1–30 day bucket, and HTM unrealised loss as a % of CET1.
- [ ] Map loan mix and concentration — unsecured share, top-20 borrowers vs net worth, CRE vs capital, geographic concentration.
- [ ] Review operational KPIs: business per employee and per branch, deposits per branch, collection efficiency, digital share, cross-sell.
- [ ] India: read the RBI divergence disclosure, the related-party note, any RBI penalty or supervisory action, and MD/CEO term approvals.
- [ ] Compute P/ABV (net worth − net NPA, less DTA and revaluation reserves), not just P/B.
- [ ] For groups with insurance/AMC/NBFC arms, do the SOTP and report the *implied standalone bank* multiple.
- [ ] Anchor the target multiple with justified P/B = (ROE − g) / (COE − g) and sanity-check against a peer ROE-vs-P/B scatter.
- [ ] Track BVPS (and tangible BVPS) growth and every equity raise's P/B — flag any issuance below book.
- [ ] State where in the credit and rate cycle this sits, and what regulatory change is in flight.
- [ ] Peer set: same country, sub-sector, funding model, accounting regime, size band and consolidation basis — stated explicitly.
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| add-product-e2e-evalpaperclipai/paperclipAdd or extend a Paperclip full-stack runner E2E workflow, fixture, matcher, or report evidence path for local or Daytona execution. | Stars: 115k | Updated | Claude CodeCodex | Verified | Install |
| add-runner-evalpaperclipai/paperclipAdd or extend a Paperclip Runner protocol evaluation definition, roster, assertion, or report fixture with provenance and narrow validation. | Stars: 115k | Updated | Claude CodeCodex | Verified | Install |
| diagnose-why-work-stoppedpaperclipai/paperclipCompatibility entry point for failed, stalled, looping, or over-recovered Paperclip work. Use fix-failed-paperclip-run for the full diagnosis, regression, repair, real verification, and green-PR workflow; honor an explicitly requested diagnosis-only scope. | Stars: 115k | Updated | Claude CodeCodex | Verified | Install |
| fix-failed-paperclip-runpaperclipai/paperclipDiagnose and fix failed, stuck, slow, or incorrectly blocked Paperclip tasks and runs. Use with a task/run URL or failure report to trace the root cause, prove a regression red then green, check second-order effects, verify the real workflow, and deliver a PR with passing checks. | Stars: 115k | Updated | Claude CodeCodex | Verified | Install |