AI-Powered Risk Analytics — Ujjivan Small Finance Company
Visual: field-app snapshots, branch dashboards, and risk heatmaps.
Client Overview
Ujjivan Small Finance Company is a leading microfinance & small finance institution serving underbanked urban and rural customers across India. High-volume microloan origination and limited digital footprints for many customers created a challenge to scale credit decisions while managing portfolio risk.
Challenge
Traditional credit assessment relied on manual forms and limited bureau coverage. Rapid portfolio growth risked increased NPLs (non-performing loans) without better risk differentiation. The client needed an affordable, scalable scoring solution that worked with limited data and supported agent workflows in low-connectivity environments.
Solution — Hybrid Credit Scoring & Portfolio Monitoring
Medro Hi Tech Symbol delivered a hybrid risk analytics platform combining alternative data, field-app behavioral signals, and traditional bureau information to generate dynamic risk scores and portfolio-level alerts.
Core capabilities
Approach
- Data readiness audit and integration of minimal viable signals from field agents.
- Model development with emphasis on explainability for agents and credit officers.
- Pilot on one state with retrospective validation and live A/B testing against legacy rules.
- Rollout with agent training and weekly performance dashboards for branch managers.
Technology stack
Implementation — Phased Rollout
Phase 1 — Data & Pilot (Weeks 1–8)
Collected historical loan outcomes, patched field-app event logs, and defined pilot cohorts by product and geography.
Phase 2 — Model Build & Field Validation (Weeks 9–20)
Built ensemble models and tested their predictive lift against legacy rules in a controlled live pilot.
Phase 3 — Branch Rollout & Monitoring (Weeks 21–36)
Deployed scoring in production, added automated early-warning alerts, and trained credit teams to interpret scores.
Phase 4 — Continuous Learning (Months 9+)
Weekly retraining with new performance labels, causal analysis for cohort drift, and new feature experiments.
Impact & Results
22%
Reduction in 30–90 day delinquencies in pilot cohort
18%
Increase in approved-but-safe disbursals (better risk differentiation)
2x
Faster credit decisions at branch level
Ongoing
Continuous model calibration
Qualitative outcomes
- Credit officers trusted the score due to transparent feature explanations.
- Agents used the mobile insights to prioritize collections and tailored counselling.
- Improved portfolio resilience without restricting responsible lending growth.
Client Testimonial
Key Highlights & Learnings
- Alternative data and field signals are highly valuable in microfinance contexts with limited bureau coverage.
- Explainability helps field agents and officers adopt scoring tools faster.
- Small, frequent retraining cycles manage cohort drift in high-turnover portfolios.