Built From the Inside
What Seven Years of Waiting Taught Me About Building AI It was 2017 and I was in San Francisco to speak at a fraud analytics conference. I had a confirmed slot on the programme, I knew what I was going to say, and I had arrived well in advance. What I hadn’t anticipated was what […]
AI Charged-Off Debt Recovery for US Banks: Improving ROI on Legacy Portfolios
TL;DR A charged-off portfolio on a US bank’s balance sheet has a specific financial character. The accounts have been written down. The provision has been taken. The loss is already on the books. Recovery at this stage is pure upside against a cost base that has already been absorbed. The standard model for working that […]
AI Self-Learning Models for NBFC Collections: How the Technology Works
TL;DR A propensity model deployed on a personal loan portfolio in January was trained on 18 months of historical payment data. At deployment, its scores were accurate. Recovery rates improved through the first quarter. By August, the picture has changed. Payment rates in the mid-propensity band have dropped. The model continues to route accounts to […]
MLOps for SA Credit Providers: What AI Collections Implementation Looks Like
TL;DR Building an AI collections model and running one are two different disciplines. A model that performs well in development can fail in production for reasons that have nothing to do with the quality of the underlying algorithm. The training data becomes stale as the portfolio evolves. The portfolio composition shifts toward product types the […]
Credit Union Collections AI: Balancing Member Relationships and Recovery
TL;DR The member who is 45 days past due on their auto loan is also the member who has held a savings account since 1998. Their spouse has a mortgage with the credit union. Their two adult children opened their first checking accounts at the same branch three years ago. The family’s total deposit relationship […]
AI for Early Bucket Collections: How US Community Banks Reduce 30-60 DPD Loss
TL;DR A US community bank with a $180 million consumer loan portfolio sees between 800 and 1,200 accounts enter the 30-60 DPD bucket every month. The collections team has three full-time agents. With current tools and contact infrastructure, they can realistically work about 600 accounts in a month. The remaining 200 to 600 accounts carry […]
Propensity Scoring for South African Collections: A Practical Introduction
TL;DR A collections manager at a South African retailer credit provider works a 30-60 DPD portfolio of roughly 8,000 accounts every month. Her current system ranks them by outstanding balance. Her team contacts the highest balances first and works down the list. By month end they have reached about 60% of the accounts. The rest […]
Champion-Challenger Testing for SA Banks and Credit Providers
TL;DR A collections model has been running at a South African bank for ten months. It was built on 24 months of payment data, the bulk of it pre-pandemic. It performed well in the first two quarters after deployment. Recovery rates in the 30-60 DPD bucket improved. Cost per recovery came down. In the last […]
Real-Time Collections Decisioning for India’s Digital Lending Market
TL;DR At 2:17pm on the 5th of the month, a borrower receives a salary credit notification. Her personal loan EMI was due on the 1st. She is now four days past due. At the moment her salary lands, her propensity to pay is at its highest point for the rest of the month. She has […]
Collections AI for South African Credit Providers: How the Technology Works
TL;DR A collections supervisor at a South African credit provider starts most Monday mornings the same way. She opens the delinquency report. She looks at the accounts between 30 and 90 days past due. She divides them into calling batches for the week, organised by balance with the larger balances worked first. Her team moves […]
