For banks, MFIs and credit unions
A guide for risk, IT, compliance and legal: how a pilot is scoped, set up, reviewed and run, and what it takes to go live.
Are you a digital lender or credit provider?Find out if it fits
Built for institutions with committees and policies.
Who it is for
Banks, microfinance institutions, credit unions and cooperatives lending to people and small businesses with thin or no bureau files.
Where it sits
Beside your credit policy, not above it. AfriScore returns a score, a risk tier, a probability of default and reasons. Your credit committee keeps every decision.
What it replaces
Nothing at first. It runs next to your current process until you have the evidence.
Look before you talk
Read how the model is built and tested on the Technology and Research pages, and try the demo.
Agree the terms
More people at the table, so we plan for it.
Bring the right people early
Risk, IT, compliance and legal each have questions. A joint first call saves rounds of back and forth.
Scope the pilot
We agree the portfolio, the products and the outcomes to measure, so your committee knows what a go looks like.
Terms
Usually an enterprise subscription: a monthly commitment with lower per-score rates as volume grows. Final terms are agreed with you.
Paperwork
A service agreement and data processing terms for legal and compliance, plus model documentation for your validation team.
Set up and integrate
Deployment that your security team can sign off.
Choose where it runs
The hosted API, or Docker on your own servers when borrower data must stay inside your network.
Connect to your system
Through the API or a light adapter to your core banking or loan origination system.
Review with IT and security
Your security team reviews the integration and the deployment before real applications flow.
Test in shadow mode
Run real applications through it with no effect on a single decision.
Run the pilot, then go live
Evidence your committee can review.
Shadow mode
Every application is scored in parallel. Results are logged and never influence your decision.
Calibration
Real outcomes (approvals, rejections, repayments) tune the engine to your customers.
The report
A statistical analysis of how it would have performed on your portfolio, with a clear go or no-go for your committee.
Going live
Scoring moves from logged to used. Every decision still follows your credit policy and carries its reasons.
Who does what
Your teams own decisions and the loan. We own the model, its calibration and the scoring service.
Security and compliance
Built to be audited.
Auditable model
A random forest ensemble with isotonic calibration and SHAP explanations, documented for your validation team.
Stress-tested
Feature ablation, data sparsity, distribution shift and SHAP stability tests before it touches a loan.
Data sovereignty
Scoring can run in your environment. No borrower data leaves unless you choose a hosted deployment.
Regulators
Designed with CEMAC and regional data protection frameworks in mind. Where your supervisor requires notice or approval, we help you prepare it.
Trust Center
Security and compliance information is kept at trust.afriscore.africa.
Outcomes and partnership
What you take away, and what comes next.
Evidence for your committee
You see how it would have performed on your own portfolio before you commit to anything.
Consistency across branches
The same kind of score and reasons for every officer and branch.
Reach thin-file clients
People with no bureau file become scoreable from the digital footprints they already leave.
Where it is heading
Portfolio analytics and management, then credit information data. Both are planned.
Pricing
Pilot
Free
Shadow mode beside your current process. No cost and no commitment.
Pay per score
From $0.25
Per scoring request through the API, billed in your local currency.
Enterprise
Agreed with you
A monthly commitment with lower per-score rates as your volume grows.
What to bring
Tick what you already have. 0 of 6 ready.
Questions
Do you lend or take on risk?
No. AfriScore never lends and never holds the risk.
Who makes the credit decision?
Your institution. The score and its reasons inform your committee and your policy.
How does our validation team assess the model?
It gets documentation of the algorithm, calibration, explainability method and the stress tests we run.
Do we need regulator approval?
That depends on your regulator and policy. Shadow mode never influences a decision, so it is a lighter first step, but confirm with your compliance team.
Can it run inside our network?
Yes. AfriScore can be deployed with Docker on your own servers.
What does it cost?
The pilot is free. After that, usually an enterprise subscription with a monthly commitment and lower per-score rates at volume.
See it on your own portfolio
The pilot is free and runs beside your current process.