Imagine you lend ₹10,000 to a friend. He's been a bit erratic with money lately. Late on rent. Skipping work. You start getting worried.
Question: when do you mentally write off some of that money as "probably gone"?
If you're like most people, you wait until your friend actually disappears with the cash. Or you keep telling yourself "he'll pay me back" right up to the day the WhatsApp goes blue tick.
That's basically how Indian banks have been running their books for decades. They wait until a loan actually goes bad before setting aside money to cover the loss.
The RBI just decided that's no longer good enough.
On April 27, 2026, the Reserve Bank dropped 14 final directions that fundamentally change how Indian banks recognise bad loans. The shift is from "wait and see" to "look ahead and prepare." The framework goes live on April 1, 2027.
And the market noticed instantly. Bank stocks took a hit the very next session, with the Sensex closing 0.5% lower at 76,887 as investors rushed to figure out who gets hurt and who doesn't.
Let's break it down.
What just happened?
On Monday, April 27, the RBI issued final guidelines for adopting an Expected Credit Loss (ECL) based loan loss provisioning framework. Implementation date: April 1, 2027.
Banks had been begging for an extension. They didn't get one.
Here's the simplest way to understand the change.
Under the old system, called the "incurred loss" model, banks set aside money for bad loans only after the loan had actually gone bad. A borrower defaults. A bank books a loss. Money gets allocated to cover it. By that point, the damage is already done.
Under the new ECL model, banks have to look ahead. They estimate, in advance, how much they're likely to lose on every loan based on the borrower's risk profile and the broader economic environment. They set aside the money up front. If things go wrong later, they're already prepared.
It's the difference between an umbrella you carry just in case versus one you only buy after you've been soaked.
The three-stage system
The new framework introduces a three-stage classification for every loan a bank holds.
Stage 1: Healthy loans. No significant increase in credit risk since the loan was given out. The bank sets aside provisions equal to 12 months of expected losses. For standard corporate and retail loans, the minimum provision is 0.40%.
Stage 2: Warning loans. The borrower is showing signs of stress. Maybe credit rating got downgraded. Maybe macroeconomic conditions are deteriorating. The bank now has to provision for lifetime expected losses on these loans. Minimum provision: 5%.
Stage 3: Credit-impaired loans. The loan has gone bad. The bank also has to take lifetime expected losses, with significantly higher provisioning depending on how long it's been in default.
The clever part? A loan can move between stages. If a borrower's situation worsens, the loan slides from Stage 1 to Stage 2 even before they actually default. The bank starts setting aside more money immediately. That's the forward-looking bit.
How is this calculated?
Banks now have to use three technical parameters to figure out their expected losses.
Probability of Default (PD). Basically, what's the chance this borrower will not pay back?
Loss Given Default (LGD). If they don't pay back, how much do we actually lose after recovery from collateral?
Exposure at Default (EAD). What's the outstanding amount we'd have lent to them by the time they default?
You multiply these together, weight them across multiple economic scenarios (some good, some bad, some catastrophic), and you get the expected credit loss number.
Sounds technical because it is. Banks now need sophisticated, data-driven models to do this. They need clean data going back years. They need risk teams that can build and validate these models. They need finance and risk departments to actually talk to each other.
For a lot of Indian banks, that's a serious upgrade. Not impossible. Just expensive.
Who gets hit?
This is the part everyone's asking about. And the answer isn't pretty for everyone.
Analysts at Nomura flagged that ECL implementation will lead to higher upfront provisioning, especially in unsecured retail loans, MSME exposures, and corporate loans. The pain isn't evenly distributed.
Public sector banks. PSU banks generally hold lower additional provisions today. They'll have to increase that buffer significantly. This means lower reported profits in the immediate transition period.
Mid-tier banks. Same story. The impact on these banks is expected to be higher because their existing buffers are thinner.
Large private banks. Less affected. HDFC, ICICI, Axis, Kotak already maintain provision buffers of 2-4% of net worth. They've been preparing for this for years. The transition will sting but won't bleed them.
This is why bank stocks fell on Tuesday. Investors did the math. Bandhan Bank rallied 9% on a different earnings story, but the broader banking sector felt the heat. PSU banks, in particular, have been quietly under pressure as the market re-prices their future earnings.
The decline followed the RBI's confirmation of its ECL framework and final asset classification norms, said analysts at Bajaj Broking, with concerns centring on higher provisioning requirements.
The prudential floor twist
Here's where the RBI threw a curveball.
Even after banks calculate their expected credit losses using fancy models, the central bank has imposed minimum provisioning floors. These are non-negotiable. No matter what your model says, you can't provision below these floors.
Why? Because the RBI doesn't fully trust banks to model their own losses honestly. If you let banks calibrate everything themselves, the temptation to underestimate risk is enormous. Lower provisions equal higher reported profits equal happier shareholders equal higher executive bonuses.
So the RBI said: fine, use your models. But you're never provisioning less than this.
Banks asked for these floors to be reduced. The RBI didn't agree. According to Suresh Ganapathy, MD at Macquarie Research, the central bank held its line on the floors despite industry pushback.
The product-wise floors apply across retail, corporate, MSME, agriculture, and real estate exposures. Each category has its own minimum. Even Stage 1 corporate and retail loans need at least a 0.40% provision. Stage 2 jumps to 5%. Stage 3 goes much higher depending on how long the loan has been in default.
Translation: the RBI is treating the ECL framework as a floor, not a ceiling. Banks can be more conservative if they want. They cannot be less conservative than the regulator demands.
The governance overhaul nobody's talking about
Beyond the math, the RBI quietly added something powerful: a governance framework that puts senior management on the hook for ECL implementation.
Every bank now needs a Board-approved committee, including the Chief Financial Officer (CFO) and Chief Risk Officer (CRO), to oversee the ECL framework. This committee has to ensure data integrity throughout the entire computation lifecycle, maintain effective control frameworks, and guarantee complete independence of internal model validation.
There's a three-tier model risk management structure spanning business, risk, and audit functions. If something goes wrong, there's a clear chain of accountability all the way up.
This sounds bureaucratic. It actually matters a lot. One of the consistent failures in Indian banking has been weak internal risk modelling, where models said one thing and reality did another, and nobody was clearly responsible. The new framework forces banks to put a name and a face on every modelling decision.
What stays the same
Here's an important detail most coverage is missing.
The 90-day rule for classifying a loan as a non-performing asset (NPA) is unchanged. If a borrower doesn't pay for 90 days, the loan becomes an NPA. This rule has been around for years, and the RBI is keeping it.
What's changed is the provisioning approach, not the classification approach.
So when you read headlines saying "RBI tightens bad loan rules", that's slightly misleading. The definition of a bad loan hasn't changed. What's changed is when and how much money banks have to set aside for loans that might go bad.
Subtle but critical difference.
What it means for you
If you own bank stocks. Expect short-term pain in PSU and mid-tier names as they ramp up provisions. Expect resilience in large private banks like HDFC, ICICI, Axis, and Kotak that already have buffer cushions. The next two to three quarters could see lumpy earnings as banks transition. But beyond that? Cleaner balance sheets, more comparable financials across the sector, and stronger investor confidence over time. Well-managed banks may see valuation re-rating, according to India Infoline's analysis.
If you're a borrower. The bank now has a stronger incentive to stress-test you before sanctioning a loan. Expect tighter underwriting, especially for unsecured personal loans, credit cards, and MSME loans. The era of "approved in 30 seconds" personal loans isn't going away, but the credit checks behind those approvals will get more sophisticated.
If you're a depositor. No direct impact on your savings or FD rates. But indirectly, banks with cleaner books and better risk management are safer custodians of your money. So that's a quiet win.
If you work in banking, finance, or compliance. Job market just got more interesting. Banks are going to need risk modellers, data engineers, model validators, and compliance officers who understand IFRS 9 and ECL methodologies. Salaries in these roles are already moving up.
If you're an equity analyst. Earnings models for the next two years just got harder. Provisions will be lumpy. Comparability across years will break temporarily. The banks that report transparently and explain their ECL transitions clearly will be rewarded by the market. The ones that hide behind jargon will get punished.
The bigger picture
This isn't just a bank regulation story. It's an Indian banking maturity story.
For years, Indian banks operated in a regulatory environment that lagged global standards. The "incurred loss" model was scrapped in most major economies after the 2008 financial crisis showed how badly it underestimated risk. IFRS 9, which mandates ECL, became the global gold standard. India is now finally aligning with that.
Why does this matter? Because when international investors look at Indian bank stocks, they want to compare them to global peers. Cleaner, forward-looking provisioning makes that comparison possible. It builds confidence. It opens up capital flows.
It also makes the Indian banking system more crisis-resilient. The 2008 crisis happened partly because banks underestimated future risks. By the time losses showed up, capital had already been distributed as profits and bonuses. The ECL framework is designed to prevent that exact failure mode.
For Indian banks, the next 11 months are about preparation. Building models. Cleaning data. Training risk teams. Setting up governance. April 1, 2027 is the deadline. Whoever's ready by then will sail through. Whoever isn't will scramble.
Sometimes regulation hurts before it helps. This one is going to sting in the short run for banks that haven't done the work. But ten years from now, when the next financial crisis hits and Indian banks come through cleaner than the rest of the world, this is the regulation people will point to.
The friend you didn't lend ₹10,000 to? You'll thank yourself later.
