SME early warning indicators — the move from past-due to behavioural

The classical credit monitoring metric is days past due. It has the considerable virtue of being unambiguous. A borrower has either paid on the due date or they have not, and once a payment is missed the metric increments daily. It feeds directly into regulatory delinquency definitions, into impairment staging under IFRS 9, and into the bank’s collections and remediation workflows. Within those workflows it remains essential.

As an early warning indicator, however, days past due is structurally lagging. By the time it triggers, the borrower’s underlying difficulty has typically been visible for weeks, and often months, in data the bank already holds. The transactional velocity in the borrower’s operating accounts has compressed. Working-capital cycles have lengthened. Trade finance utilisation patterns have shifted in ways that suggest cash-flow stress rather than business growth. Disclosures have begun to arrive late. None of these signals constitutes a missed payment, and none of them would necessarily prompt action under a monitoring framework keyed to delinquency. All of them tend to precede the missed payment by a meaningful interval.

By the time delinquency-based EWIs trigger, the borrower’s underlying difficulty has typically been visible for weeks in data the bank already holds.

The supervisory direction of travel is to expect early warning frameworks that incorporate these forward-looking signals, particularly for SME portfolios where transactional data is rich and the borrower-bank relationship is operationally embedded.

What the supervisory expectation actually is

The Basel Committee’s update of its credit risk management principles, published in 2025, makes the expectation explicit. The principles call for banks to enhance early-warning and remediation capabilities through predictive analytics, behavioural indicators, and automated escalation protocols designed to flag deteriorating exposures and activate timely interventions. The same principles emphasise end-to-end data traceability and close-to-real-time monitoring of exposures — the operational substrate without which behavioural EWIs cannot be implemented.

The EBA Guidelines on loan origination and monitoring (EBA/GL/2020/06), which apply in EU jurisdictions and which the CBUAE has referenced in its own approach to credit risk monitoring, set similar expectations. The guidelines require institutions to maintain ongoing monitoring frameworks proportionate to portfolio risk, with indicators capable of identifying deterioration before contractual default occurs.

The CBUAE Risk Management Regulation does not specify a list of required early warning indicators, but it does require, under Article 2, that risk management functions develop metrics relevant to the risk appetite statement, monitor those metrics, and escalate breaches. For credit portfolios, an EWI framework is the operational instrument through which deterioration metrics are monitored against credit risk appetite — and the supervisory dialogue increasingly tests whether the institution’s EWI framework actually performs that function.

The behavioural signals that tend to precede delinquency

Where institutions have invested in behavioural EWI frameworks for SME portfolios, the indicators that empirically lead delinquency tend to fall into a small number of categories. The specific calibrations differ across portfolios, jurisdictions, and economic cycles, but the underlying categories recur.

Transactional velocity in operating accounts. The most consistent leading indicator across SME portfolios is a sustained compression in the velocity of operating-account turnover relative to the borrower’s prior pattern and to the seasonal baseline. A business under cash-flow stress will, almost without exception, show this signal first in its current accounts. Detecting it requires the bank to compare current-period velocity against a moving baseline that controls for seasonality — not against a static threshold. The discipline is straightforward; the data is held by the bank in any case; the implementation tends to lag for reasons that have more to do with system architecture than with analytical complexity.

Trade finance utilisation pattern shifts. For SMEs engaged in international trade, the utilisation pattern of trade finance facilities — letter of credit issuance, document collection timing, bill discounting requests — often signals stress before the operating accounts do. A borrower that has historically utilised trade finance in line with its commercial calendar and then begins requesting facilities at unusual intervals, or at unusual amounts relative to its commercial flow, is signalling either a change in business model or a working-capital problem. Both warrant a relationship-level review.

Cheque issuance patterns and same-day mirroring. Cheque return signals — particularly returns for insufficient funds, but also the broader pattern of mirrored same-day deposits and withdrawals that suggest the borrower is managing tightly — tend to precede payment default by a measurable interval. These are conduct-level signals that fall out of routine account monitoring; the question is whether the bank’s monitoring infrastructure surfaces them to the credit team in time to act.

Disclosure timing. SME financial disclosures arrive on a covenant schedule. When they begin to arrive late — particularly when the lateness pattern correlates with quarter or year-end periods — it tends to indicate either deteriorating reporting discipline or reluctance to disclose. Both warrant inquiry. The signal is among the easiest to monitor and among the least systematically tracked.

External and market signals. Public information about the borrower’s customers, suppliers, or sector — particularly in concentrated sectors where SME exposures cluster — provides context for borrower-level signals. A behavioural EWI for an SME in a sector under stress will produce different staging implications than the same EWI for an SME in a sector in expansion. The framework needs to be able to make that distinction.

The link to IFRS 9 SICR — frequently weaker than it should be

A properly designed EWI framework should feed directly into the IFRS 9 staging assessment. IFRS 9 paragraph 5.5.17 requires institutions to use reasonable and supportable forward-looking information in assessing whether credit risk has increased significantly since initial recognition. Behavioural EWIs are precisely the kind of forward-looking information the standard contemplates — and yet, in practice, the link between the EWI framework and the SICR assessment is often weaker than the standard implies.

Three operational issues tend to weaken the link.

First, the EWI framework and the SICR assessment are often owned by different teams — the credit risk monitoring function and the IFRS 9 modelling function respectively — with limited two-way information flow. EWI triggers do not automatically inform staging; staging movements do not automatically trigger EWI review. The result is that a borrower can be flagged by the EWI framework and remain in Stage 1, or be migrated to Stage 2 by quantitative model deterioration without any corresponding EWI flag. Either disconnect makes the framework less defensible in audit and supervisory dialogue.

Second, the EWI framework’s outputs are often qualitative — a watch-list classification, a relationship-level note — when the SICR assessment requires evidence that is at least semi-quantitative. Translating an EWI signal into the form the IFRS 9 model can consume requires that the EWI framework produce, for each flagged exposure, something more than a categorical flag. The mapping from behavioural signal to PD movement is non-trivial; in our experience, the institutions that handle it best have invested in the mapping explicitly rather than treating it as an emergent property of two parallel systems.

Third, the time-horizons of the two frameworks frequently differ. An EWI flag is a near-term signal — typically a thirty-to-ninety-day window in which intervention is meaningful. The SICR assessment looks across the lifetime of the exposure. The challenge is to translate a near-term behavioural signal into evidence of a significant increase in credit risk over the lifetime of the instrument. The translation is doable, but it requires deliberate methodological work and documentation; where the work has not been done, the EWI framework and the SICR framework tend to coexist without genuinely informing each other.

Where the investment tends to be most productive

Across engagements, the institutions that have built the most useful SME EWI frameworks have tended to invest in three areas in approximately this order.

First, in the data infrastructure that allows behavioural signals from operating accounts, trade finance, cheque clearing, and disclosure systems to be aggregated at the borrower level on something close to a real-time basis. Without the aggregation, the analytical work cannot be done.

Second, in the analytical work that converts raw signals into stable, calibrated indicators — controlling for seasonality, controlling for borrower-specific patterns, and producing alerts at thresholds that the relationship and credit teams find actionable rather than overwhelming.

Third, in the operational workflow that translates an EWI alert into a relationship-level review, with a defined escalation path, a defined intervention toolkit, and a defined feedback loop into both the credit monitoring framework and the IFRS 9 SICR assessment.

The institutions that have invested in all three tend to report meaningfully lower realised loss rates per cohort — not, primarily, because they collect more effectively, but because the dialogue with the borrower starts earlier, when the range of available restructuring and remediation options is still wide. By the time delinquency-based EWIs trigger, that range has usually narrowed.

For a more detailed treatment of EWI framework architecture for GCC SME portfolios — including specific indicator calibration patterns, the link to IFRS 9 SICR, and the workflow integration that distinguishes effective frameworks from analytically sophisticated but operationally inert ones — see our discussion in the Library.