With the expected credit loss (ECL) guidelines set to take effect from April 2027, banks are gearing up to overhaul their data architecture to meet the Reserve Bank of India’s requirements. A key challenge will be integrating data from multiple systems to calculate credit losses.
Avinash Raghavendra, group head – IT & retail operations at Axis Bank, said, “In order to meet the regulatory expectations, changes are envisaged in the non-performing asset (NPA) management software. Discussions are currently underway on where changes are required and around third-party support.”
“ECL is far more than an accounting change. It represents a large-scale transformation across data, analytics, risk management, and technology,” said Dhruv Parikh, partner – financial services risk consulting, EY India.
To estimate credit losses using forward-looking information, banks will need to draw data from core banking systems, loan origination platforms, collateral management and recovery systems, credit bureaus, macroeconomic feeds and finance systems.
The RBI guidelines call for sufficiently granular credit-risk data, robust processes to identify and manage data-quality risks across internal and external datasets, and adequate historical loss data covering representative periods and variations across business cycles. Banks will also have to guard against material inconsistencies or selective use of data during aggregation, which could distort ECL estimates.
Recent tenders floated by Indian Bank, Indian Overseas Bank (IOB) and the National Bank for Agriculture and Rural Development (NABARD) show regulated entities preparing technology systems to handle probability of default (PD), loss given default (LGD), exposure at default (EAD), staging, macroeconomic scenarios and model recalibration.
Sagar Lakhani, partner at Uniqus Consultech, said many Indian banks’ core banking software was not designed to retain the detailed historical data required for credit-risk assessment, making significant data-cleaning and correction work necessary.
Banks with multiple main banking systems could also face problems in making data consistent and matching records across systems. Older systems that cannot easily exchange data may require information to be transferred through files, while collateral data may be held separately and need to be organised before it can be reliably used to calculate expected credit losses.
“Realistically, banks should budget 6 to more than 12 months for the integration layer alone,” Lakhani said.
Given the scale and complexity of these requirements, analysts said banks will need robust end-to-end technology platforms to automate calculations, maintain data quality, run risk models and generate consistent and timely results across millions of accounts in line with the RBI’s ECL guidelines.