The ECL Shift: From Reactive to Predictive Banking

Introduction

When the Reserve Bank of India released its draft Directions on Expected Credit Loss (ECL) on October 7, 2025, it started one of the biggest updates to India’s banking rules in decades.

The draft replaces the old “incurred loss” system with a forward-looking, data-based approach to credit risk. The goal is simple: help banks spot stress early, protect capital, and make the system stronger.

The framework will start on April 1, 2027, with a transition period until March 31, 2031. For lenders facing pressure from unsecured retail, microfinance, and small-business loans, this move comes at the right time.

Why the Change Matters

India still uses the incurred loss model, where banks set aside money only after a borrower defaults. This often delays recognition of stress and makes downturns worse.

Global standards such as IFRS 9 and CECL already use an expected loss model, which anticipates losses much earlier. RBI’s draft ECL Directions bring India in line with those standards while fitting local needs.

The timing is important. Credit growth is solid, but defaults in unsecured loans, microfinance, and MSMEs have been rising. Predictive provisioning helps banks react faster and keeps profits steadier.

The draft also replaces more than 70 circulars and master directions issued since the 1990s. It is a complete reset of how India manages loan loss provisions.

What the Draft Directions Introduce

The new framework gives banks a clear structure for measuring expected losses.

Key features:

• A three-stage ECL model

○       Stage 1: Performing assets, 12-month expected loss

○       Stage 2: Assets with significant increase in credit risk (SICR), lifetime expected loss

○       Stage 3: Credit-impaired assets, lifetime expected loss with higher minimums

• Prudential floors as minimum coverage levels — 0.25 to 1 % for Stage 1, 1.5 to 5 % for Stage 2, and up to 100 % for Stage 3.

• Governance: Board-level oversight, model validation, and independent audit

• Data: At least five years of historical loss and recovery data

• Macroeconomic scenarios: Use of probability-weighted outcomes under different conditions

• Disclosure templates: Detailed reconciliations and macro assumption reporting

Governance, Data, and Model Requirements

The draft Directions formalize a comprehensive governance framework for ECL:

• Board responsibility: Overall approval and periodic review of ECL methodology.

• CFO + CRO subcommittee: Oversee implementation, controls, and disclosures.

• Model validation: Mandatory independent validation before rollout and periodic revalidation.

• Audit trail: End-to-end traceability of assumptions, data, and model results.

• Data requirement: At least five years of granular default and recovery data.

• Collateral valuation: Regular updates; once at classification and every two years thereafter for large exposures.

• Borrower linkage: If one exposure of a borrower is in Stage 3, all exposures are treated as Stage 3.

This is a major step toward data-driven governance - shifting compliance from manual provisioning to continuous, analytical control.

Practical Implementation and Transition Planning

RBI has chosen a gradual approach so banks can adjust smoothly without any sudden strain on their capital reserves.

• The new system begins April 2027, with transitional relief until March 2031.

  
  

  

  
  

• Banks may add back part of the initial ECL shortfall to core capital (CET 1) during this period.

• RBI is also revising the Standardised Approach for Credit Risk, expected to reduce risk weights for MSME and housing loans.

These measures work together. Early provisioning will raise short-term costs (especially driven by Stage 2 assets), but lighter risk weights and capital relief will balance it out. By 2031, India’s banks will have cleaner balance sheets and capital buffers that match their true risk.

How ECL Actually Works

Under ECL, banks estimate losses for every loan when it is made

Formula: ECL = PD × LGD × EAD

Example: If a borrower has a 2% chance of default, 50% loss given default, and ₹10 lakh exposure, the expected credit loss = ₹10,000. That amount is provisioned immediately, even if the borrower is currently performing.

Loans move between stages as risk rises, using triggers like payment delays or rating downgrades. This impacts ECL through the underlying components of PD and LGD.

The Directions specify that if a borrower’s payment is more than 30 days past due (SMA 1 asset), the loan is automatically presumed to have undergone a significant increase in credit risk (SICR). This means it must shift from Stage 1 to Stage 2 unless the bank has documented evidence to rebut this assumption. This rule creates a uniform trigger for early warning and lifetime provisioning across the industry.

Expected and Unexpected Losses – A Clear Split

The ECL model formalizes the divide between expected and unexpected losses.

• Expected losses are regular defaults that can be predicted. They are now covered fully through ECL provisions shown in the profit and loss account.

• Unexpected losses are rare shocks such as recessions or sudden failures. They are absorbed by capital buffers under Basel rules.

This split helps banks plan better and price risk more fairly. Capital now protects against only extreme events, while provisions handle the usual level of credit loss. The prudential floors in the draft act as guardrails so that models do not underestimate expected losses.

Together, they create a two-layer defense:

1. Provisions for normal, expected risks.

2. Capital for rare, unexpected shocks.

   
   
   
   

Industry Impact – Winners, Challenges, and Opportunities

The new model will reshape how banks lend and monitor credit.

• Credit evaluation will depend more on data — borrower behavior, repayment trends, and forward-looking indicators.

• Profit stability will improve because losses are recognized early instead of in sudden waves.

• Portfolio management will become dynamic, with frequent review of SICR triggers and segment performance.

• Governance and data will take center stage: boards must approve models, internal teams must validate them, and all data and assumptions must be traceable.

Who benefits

●       Large banks with diversified portfolios and strong buffers are well positioned.

●       Sectors like MSME and housing benefit from lower risk weights under revised Basel norms.

Who faces pressure

●       Lenders heavy in unsecured retail, microfinance, and small business loans will see higher provisions and model costs.

●       NBFCs and co-lenders will need to provide higher data quality to partner banks.

Where new opportunities arise

●       Growing demand for ECL models, validation tools, and provisioning automation.

●       New business for fintechs offering credit analytics, scenario engines, and data enrichment.

OneFin’s View – Ready for Predictive Provisioning

At OneFin, we see ECL as the start of a wider move toward predictive and robust risk management built on the foundations of a strong data and modeling ecosystem

Our platform already supports:

• ECL-ready, stage-based provisioning models and rule configurations able to handle complex DPD transition scenarios using historical patterns.

• Scenario simulation features that use forecasted macroeconomic indicators to determine base, optimistic and pessimistic future states and estimate probability-weighted losses.

• Integrated origination, loan management, and collections modules, giving unified data for credit monitoring.

• Audit and reporting tools that create traceability and data integrity for governance reviews.

OneFin, based on its extensive experience in risk modelling and risk management has built out a sophisticated suite capable of handling all the implementation nuances and edge cases for ECL as detailed in the pointers above.

While each bank will still design its own regulatory models, OneFin offers the flexible, low-code base needed to build, test, and deploy these capabilities quickly.

The Road Ahead

The draft ECL Directions turn India’s credit system from reactive to predictive. Once final, they will replace nearly all provisioning and asset-classification rules issued since the 1990s - a true modernization of banking regulation.

The next 18 months are the time to prepare:

●       Start ECL pilot projects on selected portfolios.

●       Improve data capture and quality across products.

●       Build governance and validation teams around risk models.

●       Plan capital and disclosure processes for the transition period.

With careful preparation, lenders can meet new standards and gain an edge in credit discipline. For those ready to act early, ECL is the path to smarter, steadier growth.

To know more, schedule a Demo.