How to reduce SME non-performing loans with dynamic credit scoring

For many banks in the UEMOA region, the SME book is one of the hardest portfolios to manage. To reduce SME non-performing loans, tightening lending criteria is not enough: it often means turning away viable businesses and walking away from a considerable market. The real lever lies elsewhere, in how risk is assessed at the outset and then monitored over the whole life of the loan.

Why SME portfolios generate non-performing loans

SME non-performing loans rarely have a single cause. Yet the same patterns keep coming back:

• A static assessment at origination: the decision rests on a file frozen at the time of application, and is never examined in depth again.
• Outdated or missing balance sheets: where they exist, they describe the previous financial year. And 85% of African SMEs have no complete balance sheet.
• Weak monitoring: between disbursement and the first missed instalment, the bank has few signals. A missed payment is a lagging indicator.
• Heavy periodic reviews: manual and infrequent, they cannot track each borrower closely.

The consequence: by the time a payment is missed, the business has often been struggling for a long while, and there is little room left to restructure.

From a frozen score to a living one

A dynamic score does more than rate the business on the day it applies. It updates as new data comes in, and every score comes with its main explanatory factors. A change in the score is therefore not an abstract number: it shows what has shifted in the borrower’s profile.

The ENGINE federated scoring engine is built on Federated Learning: the model is trained inside each institution, on its own data, and only encrypted, anonymised parameters are aggregated into an improved global model. No raw data leaves the bank, which still benefits from what the whole network has learned.

PI-SPI flows as early warning signals

PI-SPI, the BCEAO’s interoperable instant payment platform, connects banks, electronic money issuers and MFIs across the eight UEMOA countries, notably via interoperable QR code. When an SME collects payments through the FLOW module, with its consent, its flows become dated, structured data. Read over time, they reveal useful signals:

• a sustained drop in collections;
• unusual irregularity in the pattern of incoming payments;
• growing reliance on a small number of customers;
• a break from the business’s usual seasonality.

To explore this lever further, read our analysis on financing African SMEs with PI-SPI and Federated Learning.

Credit risk monitoring and early warnings

The RISK module for portfolio monitoring and early warnings turns these signals into concrete action:

• Continuous monitoring of the SME portfolio, instead of one-off reviews.
• Early warnings when a borrower’s profile deteriorates, together with the factors behind the change.
• Prioritisation of files for relationship managers and collections teams.

An early alert opens options that vanish once a payment is missed: reaching out to the business owner, understanding the difficulty, adjusting a repayment schedule, calling on a guarantee. This monitoring complements credit bureau data from providers such as Creditinfo West Africa and guarantee schemes such as the GARI Fund or the African Guarantee Fund.

Keeping decisions human and governance strong

An alert is not an automatic penalty. The score remains a decision-support tool, and every action on a loan is a matter for the bank and its committees. This governance rests on technical and regulatory safeguards:

• role-based access (RBAC) and exchanges protected by TLS 1.3;
• every operation logged and auditable, with human validation;
• data kept within the UEMOA region, in compliance with BCEAO regulations and under ARTCI oversight (Ivorian Law No. 2013-450);
• client consent collected by the partner institution.

Putting it into practice

• Target a segment of the SME portfolio where monitoring is weakest today.
• Connect your systems: CONNECT’s API connectors and a documented REST API plug into your core banking and existing tools.
• Define the alert workflow: who receives alerts, how quickly, and which actions are available.
• Measure the trend in arrears and restructured loans on the pilot segment, then scale up.

SCORE360 is preparing a six- to nine-month pilot phase with one bank and two microfinance institutions, targeted for November 2026. Explore our offer for banks.

Want to spot deterioration in your SME portfolio earlier? Book a SCORE360 demo to see the dynamic score and RISK alerts in action.