Satellite Credit Scoring: How AI Scores Unbanked Farmers From Space
September 2026
The Problem: 3 Billion People Without Credit
Globally, 3 billion people have no credit history. They have never had a bank account, never taken a formal loan, and never been scored by a credit bureau. In Mozambique, 95% of the economy is informal — meaning 9.5 million farmers are invisible to the financial system.
Traditional banks require:
- A bank account
- A credit history
- Collateral (property, vehicle)
- Payslips or tax returns
- A formal business registration
Mozambican smallholder farmers have none of these. They have land, crops, a mobile phone, and a cooperative. But they cannot access credit.
The Solution: Satellite Credit Scoring
Maputo Bridge Capital has built an alternative credit scoring system that uses six data sources — including satellite imagery — to generate a 300-850 credit score for unbanked farmers.
The Six Factors
1. Satellite NDVI (25% weight)
The European Space Agency's Sentinel-2 satellite flies over every farm every 5 days. It captures multispectral imagery at 10-meter resolution and calculates NDVI (Normalized Difference Vegetation Index).
NDVI measures vegetation density:
- 0.2-0.4 = bare soil, minimal vegetation
- 0.4-0.6 = sparse crops, early growth
- 0.6-0.8 = healthy, dense crops
- 0.8+ = very dense vegetation
For MBC's COJAZ cooperative in Zambézia:
- Current NDVI: 0.251 (September — dry season)
- Historical average: 0.394
- Expected rainy season: 0.6-0.8
The satellite provides objective, third-party verification that crops are actually growing. No one can fake it.
2. Mobile Money Activity (25% weight)
Mozambique has three mobile money networks: M-Pesa (Vodacom), e-Mola (Movitel), and mKesh (Tmcel). Together they serve 25 million accounts processing $35 billion annually.
MBC analyzes farmer mobile money transaction patterns:
- Transaction frequency
- Average transaction volume
- Consistency of inflows
- Payment behavior patterns
3. Weather Risk (15% weight)
Using the Open-Meteo Historical Weather API, MBC fetches:
- 30-day rainfall totals
- Historical rainfall comparison
- Temperature averages
- Drought event frequency
- Cyclone proximity
For COJAZ in Zambézia:
- 30-day rainfall: 27mm (rainy season starting)
- Historical average: matching
- Temperature: 23.7°C average
- Weather score: 80/100
4. Farm Profile (20% weight)
Each farmer's farm characteristics:
- Farm size (hectares)
- Years farming experience
- Crop types
- Cooperative membership
- Verification status
COJAZ founding members score 100/100 on farm profile (11.4 hectares, 5 years farming, cooperative members, verified).
5. Repayment History (10% weight)
Previous loan repayment behavior:
- Total funded
- Total repaid
- Repayment rate
- On-time payment history
For first-cycle farmers (no history), the score starts at 50/100 and improves with each successful repayment.
6. Community Verification (5% weight)
Verification by a trusted local agent:
- Identity verified (BI/NUIT document scan)
- Land boundaries confirmed
- Crop types verified
- Cooperative membership confirmed
COJAZ members were verified by Anselmo Boaventura (MBC founder and Mozambican lawyer), scoring 100/100.
The Score
COJAZ founding members currently score 608/850 (Grade C):
| Factor | Score | Weight | Weighted | |--------|-------|--------|----------| | Satellite NDVI | 25/100 | 25% | 6.25 | | Mobile Money | 30/100 | 25% | 7.50 | | Weather Risk | 80/100 | 15% | 12.00 | | Farm Profile | 100/100 | 20% | 20.00 | | Repayment History | 50/100 | 10% | 5.00 | | Community Verification | 100/100 | 5% | 5.00 | | Total | | | 55.75 → 608 |
How Scores Will Improve
| Factor | Current | How to Improve | Target | |--------|---------|-----------------|--------| | NDVI | 25/100 | Rainy season (Nov-Mar) | 70-80/100 | | Mobile Money | 30/100 | First M-Pesa transactions | 60-70/100 | | Weather | 80/100 | Continue good conditions | 80/100 | | Farm Profile | 100/100 | Already maxed | 100/100 | | Repayment | 50/100 | First loan + repayment | 70-80/100 | | Community | 100/100 | Already maxed | 100/100 | | Projected Score | 608 | After rainy season + first loan | 700+ (Grade B) |
The Loan Terms
Based on credit score:
| Grade | Score | Max Loan | Interest Rate | Term | |-------|-------|----------|---------------|------| | A | 720+ | $1,000 | 10% APR | 6 months | | B | 650-719 | $500 | 12% APR | 6 months | | C | 550-649 | $500 | 24% APR | 4 months | | D | Below 550 | Declined | — | — |
COJAZ members at Grade C qualify for $500 loans at 24% APR — lower than local moneylenders who charge 50-100%.
Self-Learning Credit Model
MBC's credit model is self-improving. Using L1-regularized logistic regression, the model trains on repayment outcomes:
- When a loan is repaid successfully, the model learns which factors predicted success
- When a loan defaults, the model learns which factors predicted failure
- Weights are adjusted automatically
- The updated model is stored in the CreditModelWeights entity
- Only deploys if accuracy improves by at least 2%
This means the credit scoring system gets smarter with every loan cycle.
Conclusion
Satellite credit scoring is not science fiction. It is operational today. The Sentinel-2 satellite is flying over Mozambican farms right now. The NDVI data is being processed. The mobile money transactions are being analyzed. The blockchain is logging everything.
No bank account needed. No collateral needed. No credit bureau needed. Just a satellite, a mobile phone, and a cooperative.
Maputo Bridge Capital. Satellite-verified. Blockchain-logged. M-Pesa-delivered.
Invest: maputobridgecapital.com/investments Donate: maputobridgecapital.com/donate Technology: maputobridgecapital.com/technology
