The client
A microfinance lender offering working capital loans to small businesses and salary-backed loans to employees of partner organisations, through 12 branches and a team of field officers.
The challenge
Applications were completed on paper, with copies of identity documents, bank statements and guarantor forms. Credit officers assessed each application manually and loan committees met twice a week. The average time from application to disbursement was five working days, and much longer during busy periods.
Repayment performance was also deteriorating. Portfolio at risk over 30 days (PAR30) had risen to 9.8%, and collections were handled informally by loan officers alongside their sales targets.
Management set three goals: make faster decisions, make better decisions and collect more consistently.
What we deployed
- Digital loan origination through a mobile app, web portal and field officer tablet app, with document capture and electronic signatures.
- Bank statement analysis using AI document processing to extract income, regular expenses, existing loan repayments and bounced items.
- Credit scoring combining application data, bank statement indicators, bureau data and repayment history, with reason codes for every decision.
- Decision rules that approve low-risk applications automatically, decline those outside policy and refer the rest to a credit officer.
- Payroll deduction integration with partner employers for salary-backed loans.
- Structured collections with automated reminders before and after due dates, payment links, a collections team worklist and promise-to-pay tracking.
- Regulatory returns and portfolio reports generated from the live loan book.
How we delivered
Digital origination for salary-backed loans launched first, after three weeks, because the data was cleaner and employer confirmation reduced risk. Small business loans followed with an initial expert scorecard. After six months of performance data, a statistical scorecard was built and tested against the expert rules before replacing them.
Results after twelve months
| Measure | Before | After |
|---|---|---|
| Average time from application to decision | 5 working days | 4 hours |
| Applications decided automatically | None | 62% |
| PAR30 | 9.8% | 5.1% |
| Monthly disbursement value | Baseline | 2.3 times baseline |
| Credit team size | 9 | 9 |
| Collections contact within 24 hours of a missed payment | Inconsistent | 100% |
Reporting that changed decisions
The weekly portfolio report, emailed to the managing director and board credit committee, shows PAR by product, branch and officer, vintage analysis by month of disbursement, and collections performance.
Vintage analysis showed that loans approved under a temporary relaxed policy performed significantly worse, and the policy was reversed. Branch-level PAR identified two branches where field officers were approving applicants outside policy. Automated rules and reason codes now make those decisions visible and consistent.
What made the difference
- Speed for good customers. Automatic approval for low-risk applicants means good borrowers get funds the same day, which improved customer satisfaction and repeat borrowing.
- Consistency. Policy rules and scores apply the same way at every branch.
- Early collection. Automated reminders and same-day follow-up of missed payments prevent small arrears from becoming defaults.
- Explainability. Reason codes help credit officers, customers and auditors understand every decision.
Client identity withheld under a confidentiality agreement. Figures are drawn from the client's before-and-after reporting for the engagement period; outcomes for your business will depend on your starting point and scope.