Bank financial ratios

The panel is the heart of the playbook. There are twelve KPIs plus three early-warning instruments. Each one comes with its formula and its healthy and stress thresholds, and the mechanics are worked on numbers. Treat the thresholds as analysts' rules of thumb for Indian commercial banks in mid-cycle. Calibrate them to segment and cycle, and use the regulatory map for the US translations. All the small worked computations use Meridian Bank, a synthetic mid-size private bank whose full statements are in Worked Example 1. Its average assets are ₹3,00,000 cr, average interest-earning assets ₹2,80,000 cr and average advances ₹2,00,000 cr. Period-end gross advances are ₹2,04,000 cr and period-end deposits ₹2,45,000 cr. Equity is ₹30,000 cr.

How is each bank ratio calculated?

Now the mechanics, one at a time.

KPI 1. Net interest margin (NIM). NIM = NII ÷ average interest-earning assets, where earning assets means advances plus investments. Leave out cash and fixed assets. For Meridian, NII ₹9,860 cr ÷ ₹2,80,000 cr = 3.52%, which is healthy. India's private banks run at ~3.5 to 4.5% and PSU banks at ~2.7 to 3.2%, so a band of 3 to 4.5% is the sane range. Large US banks run at 2.5 to 3%, with more competition for assets and a different rate structure. Decompose NIM before you praise it. NIM ≈ yield on assets − cost of funds ± mix effects. Meridian's yield on advances is 9.0% and its cost of deposits 5.1%, a 3.9% spread. So why does NIM come in below that? Low-yield SLR investments dilute the asset side. It's also higher than it would otherwise be, because ₹30,000 cr of equity funds assets for free.

There's a trap, though. A rising NIM isn't automatically good news. It can mean the loan mix is shifting toward unsecured risk whose losses arrive later, as the credit-cycle section shows. A rising NIM alongside a rising unsecured share is deferred credit cost dressed up as alpha. Going the other way, NIM compression alongside 25%+ loan growth means the bank is buying growth on price, and that's one of the red flags for banks.

How do you read asset quality ratios?

KPIs 2 and 3. GNPA and NNPA. Under the RBI's IRAC norms, a loan becomes a non-performing asset once it's more than 90 days overdue. The GNPA ratio = gross NPAs ÷ gross advances, the fraction of the book that has already failed. NNPA nets out the provisions already held against those NPAs: NNPA = (GNPA − specific provisions) ÷ (gross advances − specific provisions). Think of NNPA as the unprovided rot, the part of the damage that would still hit equity if recoveries came to zero. Meridian's GNPA is ₹4,896 cr ÷ ₹2,04,000 cr = 2.4%, which is good. It holds provisions of ₹3,574 cr, so NNPA = (4,896 − 3,574) ÷ (2,04,000 − 3,574) = 1,322 ÷ 2,00,426 = 0.66%, also good. Good here means under 3% and under 1%. India's system GNPA peaked around 11% in FY18 after the corporate cycle, and it's near multi-decade lows of ~2.5% in the mid-2020s. Verify that in the RBI's latest Financial Stability Report.

KPI 4. Provision coverage ratio (PCR). PCR = provisions held ÷ GNPA. For Meridian, 3,574 ÷ 4,896 = 73%, above the 70% comfort line. PCR tells you how candid the bank is being about its own wounds. At 70%+ it's saying "we have largely paid for our known bad loans". At 45% it's saying "our reported profit is borrowing from future write-offs." There are two conventions. One is PCR on the balance-sheet GNPA. The other is PCR including technical write-offs, loans fully written off but still being pursued, a pool India calls AUCA. Banks quote whichever one flatters them, so compute both when the note allows. Falling PCR while GNPA rises is the red-flag combination. That's a bank under-providing just when it should be over-providing, usually to protect reported ROE, and Worked Example 2 shows it in numbers.

KPI 5. Credit cost. Credit cost = P&L provisions for the period ÷ average advances. State your denominator. Some analysts use average assets, but advances is the cleaner convention, and it's the one used here. For Meridian, ₹1,300 cr ÷ ₹2,00,000 cr = 0.65%, normal and mid-band (0.5 to 0.8%). Credit cost is the flow that repairs the stock, which is PCR. Over a full cycle it has to average the true loss rate of the book, and any single year is a management estimate. A retail-heavy book might normally run at 0.8 to 1.2%, with higher gross yields paying for it. A mortgage book might run at 0.2 to 0.3%. So judge credit cost against the yield the book earns. Unsecured lending at a 13% yield with 2% credit cost is fine. Corporate lending at an 8% yield with 2% credit cost is a disaster.

KPI 6. Slippage ratio. Slippages = standard loans that became NPA during the period, and slippage ratio = slippages ÷ opening standard advances. For Meridian, ₹3,000 cr ÷ ₹1,88,000 cr of opening standard book = 1.6%, which is benign (<2%). Slippage is the freshest asset-quality number a bank publishes. If GNPA is the reservoir, slippage is the pipe filling it. The full plumbing is the NPA roll-forward, and you should rebuild it every quarter from the investor presentation:

Closing GNPA = Opening GNPA + slippages − recoveries − upgrades − write-offs

Be suspicious of write-offs. A bank can manufacture a falling GNPA ratio by writing off aggressively. Those loans leave GNPA, but the losses were real. So always read GNPA alongside slippage and write-offs. If GNPA is falling purely through write-offs while slippages stay high, the bank is bailing water and the leak is still there.

How do you read profitability ratios?

KPI 7. ROA. PAT ÷ average total assets. For Meridian, ₹4,620 ÷ ₹3,00,000 = 1.54%, which is excellent (>1% is good, >1.5% excellent). ROA is the franchise's truth with leverage stripped out, so it's the number to compare across banks with different capital levels. The best Indian private banks sustain 1.8 to 2.2%. PSU banks have historically run at 0.5 to 0.9%. Anything negative for consecutive years used to attract PCA attention, as the regulatory map explains.

KPI 8. ROE. PAT ÷ average net worth. For Meridian, ₹4,620 ÷ ₹30,000 = 15.4%, right at the 15%+ bar. Valuation runs on ROE, but never read it alone. ROE = ROA × leverage, and the leverage term is capital adequacy read upside down. A 16% ROE at 8× leverage (2% ROA) is a franchise. At 18× (0.9% ROA), the same 16% is a coiled spring. Insist, too, that ROE clears the cost of equity, ~13% for Indian banks on the CAPM build from the cost-of-capital work. An ROE of 11% destroys value at every growth rate, as the value-driver rule says.

KPI 9. Cost-to-income. Opex ÷ (NII + other income). For Meridian, ₹5,400 ÷ ₹12,860 = 42%, inside the <45% bar. Efficiency is strategy you can see. HDFC Bank-class franchises run at ~40%. SBI runs at ~50%+, carrying pension costs and legacy branches. A digital-heavy challenger may run at 65% while it builds. Watch the jaws, meaning income growth minus opex growth. Positive jaws compound PPOP year after year without any fuss. Negative jaws mean the bank is running to stand still, and Practice 5 drills it.

What do the funding and capital ratios show?

KPI 10. CASA ratio. (current + savings deposits) ÷ total deposits. For Meridian, ₹98,000 ÷ ₹2,45,000 = 40%, right at the >40% bar. Every point of CASA is a durable funding-cost advantage. Shift 10 points of deposits from term (7%) to CASA (~2.2% blended) and you save ~48 bp on total funding. At 10× leverage that's roughly 5 percentage points of ROE before competitors claw some back, and Practice 2 puts a number on it. CASA is also the sticky money. It's transactional and habit-bound, and in small units it's insured. But mind the composition. Corporate current accounts are CASA that behaves like bulk money in a panic.

KPI 11. CD ratio. Advances ÷ deposits. For Meridian, ₹2,04,000 ÷ ₹2,45,000 = 83%, at the top of the comfort band (75 to 85%). Above ~90% the bank is lending money it hasn't durably raised. Its growth is funded by borrowings, or by hoping deposits turn up later. HDFC Bank spent the years after its 2023 merger grinding an inherited ~100%+ CD ratio back toward the low 90s, and the market graded it every quarter on this one number. Verify the current figure in its investor presentation.

KPI 12. CRAR and CET1. CRAR = (CET1 + AT1 + Tier 2 capital) ÷ risk-weighted assets. In India, assets are weighted by riskiness under Basel III's standardised approach. G-secs carry 0%, corporates 20 to 150% by rating, retail/mortgage loans 35 to 75% by product and LTV, and unsecured consumer credit 125%. RBI raised that last weight in November 2023, so verify the current weights.

India's minima are stricter than global Basel. CET1 has to be at least 5.5%, Tier 1 at least 7% and total CRAR at least 9%. On top of those comes a capital conservation buffer (CCB) of 2.5%, so the operative floors are CET1 8.0% and CRAR 11.5% (as of mid-2026; check against the RBI's Basel III Master Direction). The three designated D-SIBs (SBI, HDFC Bank, ICICI) carry 0.2 to 0.8% extra. Check RBI's current D-SIB list. And there's a ≥3.5% leverage ratio, 4% for D-SIBs.

For Meridian, CET1 ₹27,000 ÷ RWA ₹2,10,000 = 12.9% and CRAR ₹31,500 ÷ ₹2,10,000 = 15.0%, both comfortable. Two readings matter. First comes distance to minimum. Work from the unrounded ratio: 12.8571% − 8.0000% = 4.8571 points, or 486 bp of CET1. Then take the absorption straight off the balance sheet: 27,000 − 0.08 × 2,10,000 = ₹10,200 cr above the floor. If you multiply out the displayed 12.9% instead, you get 4.9 points and ₹10,290 cr. That's ₹90 cr of pure display rounding, and it's how a capital plan picks up headroom it doesn't have. Second comes growth headroom. Capital consumed ≈ RWA growth × ~11.5%, and a bank at the floor can't grow without dilution.

That's how weak banks get trapped. They have too little capital to grow out of their problems and too little profit to build capital.

What are the early-warning indicators?

EW1. SMA buckets, the smoke detector. RBI's Special Mention Account ladder tags loans before they become NPAs. SMA-0 is 1 to 30 days overdue, SMA-1 is 31 to 60 and SMA-2 is 61 to 90. At day 91 the loan is an NPA. So SMA-2 is next quarter's slippage, already standing in the doorway. Banks disclose SMA data unevenly. Large borrowers' SMA status flows to the RBI's CRILC registry, and investor presentations increasingly show SMA-1+2 for the whole book. Where it's disclosed, SMA-2 above 2% of advances and rising is the best public early warning in Indian banking. Rising SMA-2 together with rising slippage is the first flag on the red-flag panel. The pipe is filling and the reservoir is already rising.

EW2. Restructured book. These are loans whose terms were eased under RBI resolution frameworks (tenor extended, rate cut, moratorium granted) without the loan being classified NPA. Economically they're wounded loans on parole, and historically ~25 to 40% of Indian restructured loans eventually slipped. Add the restructured book to GNPA for a "stressed book" total. A bank advertising 2% GNPA with 3% restructured is a 5%-stress bank in a party dress. India's 2012 to 2015 restructuring boom is the canonical case. It had its CDR cell, and its "evergreening" meant fresh loans made to pay old interest. What ended the masquerade was the Asset Quality Review, covered in the credit-cycle section.

EW3. Divergence. RBI inspects every bank once a year and re-marks its loan classifications. Sometimes its assessed GNPA or required provisioning exceeds the reported figures by more than the materiality thresholds. Currently those are additional provisioning above 10% of pre-provision profit, or additional GNPA above 15% of incremental GNPA. Verify the live circular. When that happens, the bank must publicly disclose the divergence. Read that disclosure as the referee caught the scorekeeper. Yes Bank reported GNPA of ₹2,018 cr for FY17. RBI's assessment was ₹8,373 cr, a 4× divergence. It was disclosed, and still widely rationalised at the time. Divergence is never a technicality. It's an integrity variable with a number attached.