wsery558/agri-lending-alt-data-briefing

Briefing: alternative data (seasonal + crop-health signals) for African agri micro-lending # Alternative Data for Agricultural Micro-Lending -- Briefing Published by Engineer Tsai Studio. A briefing note, not a claim of an operating credit-scoring product. ## The real problem Agricultural micro-lenders in Sub-Saharan Africa underwrite almost entirely on thin mobile-money transaction histories, and since only roughly 4% of global AI training data is African, models routinely misread normal seasonal farm income patterns as risk. Near-zero-cost smartphone crop disease/yield-risk detection exists but isn't yet fed into these credit models as a complementary signal. Source: ezbob -- SME Lending in Africa: Alternative Data Drives Credit Decisions. ## Who this is for Agri-fintech lenders serving smallholder farmers (e.g. Emata-style mobile-money-linked lenders). ## Briefing points 1. **Seasonal income is the model failure mode, not lending risk itself** -- a model trained mostly on non-agricultural, non-seasonal income patterns will systematically misread a normal harvest-cycle dip as distress unless seasonality is an explicit input. 2. **Crop-health signal is cheap now, not five years ago** -- smartphone- based crop disease/yield-risk detection has collapsed in cost, which is exactly why it's underused as a credit input today rather than structurally unavailable. 3. **The gap is integration, not data existence** -- both mobile-money histories and crop-health signals already exist independently; the real product gap is combining them into one underwriting model. ## What this is NOT Not a claim that any current NC-affiliated capability operates a credit model today -- a briefing for a lender's own data science team.