Scoring 99 Unbanked Farmers in 4.2 Seconds: Satellite Credit Model Hits 632 Average

COJAZ Cooperative Farmers is a cashew farm located in Mocuba, Zambézia province, Mozambique. MBC deployed its satellite-based alternative credit scoring model across 99 smallholder farmers in Zambézia province, achieving a 632 average credit score with 85%+ model accuracy — unlocking $39,500 in investor capital through a single API call.

99 farmers in the MBC database lacked traditional credit histories. No bank statements, no FICO scores, no collateral records. Standard financial institutions reject these farmers automatically. Without a credit score, no investor will deploy capital. The challenge: build an alternative credit model using satellite NDVI data, mobile money transaction history, and farm size metrics to produce a bankable credit score on a 300-850 scale.

MBC's credit scoring engine ingests three data layers: (1) Sentinel-2 satellite NDVI crop health indices averaged over 12 months, (2) M-Pesa mobile money transaction velocity and repayment patterns, and (3) farm size and crop type verification from agent-assisted onboarding. The model applies SHAP explainability to generate transparent reason codes for every score — satisfying Basel III compliance requirements. Each farmer receives a risk grade (A-D) and a maximum loan amount. The entire pipeline runs as a single API call: farmer data in, credit score out. 12% of scored farmers have now been formally verified and funded, with the remaining 88% in the verification pipeline.