What can lenders do to adapt?

This is part 2 of 2 article series. The first part is published here.

If this framework is directionally correct, the next edge in Indian lending will not come primarily from faster disbursement or wider distribution or sophisticated statistical models making approval/rejection decisions. They are necessary but not sufficient conditions.

It will come from improving how lenders detect early liquidity stress and, more importantly, how they translate it into collection actions before delinquency shows up in traditional metrics.

This is not a reinvention of underwriting or portfolio risk management. Most lenders already use bureau data, bank statement analytics, and behavioural signals for credit approval/rejction decisions. The gap is not entirely in data availability – it is in how these signals are operationalised beyond approval /rejection scorecards into portfolio management decisions.

Across retail and MSME credit, stress is increasingly showing up as liquidity compression rather than outright default. Practically, this appears as:

• rising intra-month account volatility (salary credit → rapid depletion cycle shortening)
• increasing EMI-to-inflow ratio drift over 3–6 months
• higher frequency of small-ticket short-term borrowing to bridge cash gaps
• delayed payments that self-cure within cycle (not yet DPD breach, but weakening timing discipline)
• shrinking “idle balance days” in bank account behaviour

Individually, these are weak signals. But at portfolio level, persistent clustering across cohorts is where signal becomes usable.

Where this actually changes lender behaviour

As I have mentioned, the real shift is not “better prediction” credit behaviour affecting approval/rejection decisions for new loans. It is earlier portfolio intervention thresholds for loans already made because the changes AFTER the loan can be quite dramatic. Practically, this changes three layers:

1. Early-warning portfolio segmentation (not credit scoring and decision): Instead of re-scoring borrowers, lenders may need parallel cohort tagging, such as:

• Stable liquidity cohort
• Volatile but self-correcting cohort
• Rising liquidity stress cohort (pre-DPD risk)

2. Exposure management rules (hard operational lever): Once a borrower enters a “rising stress pattern cluster”, actions are not binary approval/rejection — they are:

• freeze on top-up / incremental exposure
• tighter renewal underwriting even if account is current
• reduction in pre-approved limits (even without delinquency)
• shortened review cycles (e.g., 90-day vs 180-day monitoring)
• conditionality shifts (auto-debit strengthening, repayment alignment to inflows)

This is where most current systems are still underdeveloped — they detect risk, but don’t systematically change exposure behaviour early enough.

3. Collections re-segmentation (before delinquency): In India, collections strategy is typically days past due or worse, Non Performing Asset classification-led. A more advanced structure introduces a pre-delinquency engagement layer, where:

• frequent micro-delinquency cured within cycle triggers “soft watchlist”
• repeated liquidity stress patterns trigger proactive restructuring offers
• early restructuring is used selectively for “cash-flow stressed but structurally viable” borrowers

This reduces the binary jump from “current to delinquent” that most systems still rely on.

MSME reality: why this matters more here

In MSME lending, financial statements and bureau signals lag operational reality. Stress typically builds through:

• working capital cycle elongation
• input cost volatility (fuel, logistics, energy, and weather-linked disruption in agri supply chains)
• uneven receivable conversion
• informal liquidity substitution between household and business cash flows

Importantly, outcomes are not linear – the same stress pattern can lead to either recovery or deterioration depending on demand recovery, pricing power, and local liquidity conditions. So the objective is not prediction – it is earlier recognition of weakening cash-flow resilience and hence decisions on active portfolio, rather than default probability affecting lending decisions in future.

Portfolio risk is shifting from borrower risk to ecosystem risk

Correlation risk is increasingly driven by shared external dependencies:

• fuel and logistics cycles
• discretionary consumption demand
• export-linked industries
• gig/platform income concentration
• construction and real estate cycles
• refinancing-heavy credit ecosystems

Climate variability adds another correlated layer – not as a direct credit input, but as a background amplifier of income volatility in specific ecosystems (especially agri-linked and rural consumption portfolios). Its practical value is not just in borrower scoring, in fact it has greater role to play in an existing portfolio. Eg: Its role could be critical in explaining why stress clusters appear simultaneously within similar geographies and cohorts, even when borrower profiles differ.

What actually matters in implementation

As of now, the constraint is not data or statistical modelling sophistication. It is operating discipline:

• defining when a “pattern becomes a cohort risk”
• deciding when monitoring intensity changes
• ensuring exposure rules actually trigger action (not just dashboards)
• avoiding false positives where seasonal or cyclical volatility is misclassified as stress

Most models do not fail because they cannot detect signals – but because they do not consistently translate signals into consistent portfolio action changes.

Now, all of this is easier said that done. One reasonable argument would be “it would increase costs of operations”. Well, yes! It would. Do you prefer higher credit costs instead?

The next phase of Indian lending will be defined by how lenders can detect liquidity stress patterns early, and how consistently they convert that detection into exposure discipline and intervention before delinquency appears in traditional metrics. So, the advantage will sit less in modelling sophistication, and more in operational willingness to act on weak but persistent signals at portfolio scale.