Credit Scoring with Alternative Data in Africa

Millions of adults across Africa have income, spending patterns and reliable habits, but no formal credit history. Traditional scoring sees them as invisible. Alternative data, combined with machine learning, has changed that, and it has put credit in the hands of people banks once ignored.

What counts as alternative data

• Mobile money and bank transaction histories.
• Airtime and utility payment patterns.
• Merchant and supplier payment records for small businesses.
• Device and application data, where lawfully collected with consent.

Why it works

Regular income, stable spending and consistent repayment of small obligations are strong signals of creditworthiness. Models can learn these patterns and make decisions in minutes, which is what made mobile lending possible at scale.

Where it goes wrong

From my work in debt recovery, I see the other side of fast lending. Models trained on short histories can overestimate repayment capacity when conditions change. Aggressive lending to people without the means to repay creates portfolios that later need recovery, and aggressive collection practices damage borrowers and the industry's reputation. Regulators in Nigeria have responded with stronger oversight of digital lenders and their conduct.

Building responsible models

• Consent and transparency: borrowers should know what data is used.
• Fairness testing: check outcomes across groups to avoid hidden bias.
• Affordability: score capacity to repay, not only willingness.
• Monitoring: watch model performance and recalibrate when the economy shifts.
• Humane collections: professional, lawful engagement when loans go bad.

Affordability comes first

Being able to predict who will repay is not the same as lending responsibly. A model might find that a customer repays on time, while missing that they are borrowing from several apps to do so. Responsible lenders check affordability: income stability, existing obligations and the effect of the new loan on the borrower's budget. Credit bureau checks, where available, help prevent stacking of loans across multiple lenders.

Explaining decisions to customers

When an application is declined or a limit is low, customers deserve a clear explanation and a path to improve: repay existing loans, build a transaction history or provide additional information. Explainable models make this possible. They also help lenders respond to regulators and resolve complaints quickly.

A checklist for responsible digital lending

• Use only data customers have consented to share, and explain how it is used.
• Test models for unfair outcomes by gender, age and location.
• Show the full cost of the loan before acceptance, including fees.
• Use respectful collection practices; never shame borrowers or contact their friends.
• Report repayments to credit bureaus so good borrowers build a record.
• Monitor model performance monthly and retrain when conditions change.

Digital lending has brought millions of Africans into formal credit. Doing it responsibly ensures that inclusion leads to opportunity, not debt traps.

A leader's view: credit is a long-term relationship

Digital lenders that focus only on approval speed and loan volume often see losses rise sharply as they scale. Those that treat credit as a long-term relationship perform better. They start with small loans, increase limits as customers repay, reward good behaviour with better terms and help struggling customers early rather than pushing them into default.

Leaders should watch the right metrics: not only approval rates and disbursement volume, but also repayment rates by cohort, the share of customers borrowing from multiple lenders, complaints, collection practices and customer lifetime value. A lending book that grows quickly with deteriorating cohorts is a warning sign.

Partnerships can strengthen responsible lending. Sharing data with credit bureaus helps prevent over-indebtedness. Working with regulators on standards for digital lending builds trust in the whole sector. Collaborating with employers, cooperatives and merchant platforms can provide richer data and more stable repayment channels. As someone who has worked in both credit recovery and fintech, I see clearly that the cost of irresponsible lending is borne twice: first by borrowers who fall into debt, and then by lenders who must recover it. Responsible design prevents both.

The prize

Done well, alternative data can bring tens of millions of people and small businesses into formal credit. Done badly, it creates debt traps. Lenders that combine data science with sound credit discipline will capture the opportunity and keep the trust of regulators.

Common questions

What is alternative data in credit scoring?

Information beyond traditional credit files, such as mobile money activity, bill payments, transaction histories and business sales data, used with consent to assess creditworthiness.

Is alternative data credit scoring fair?

It can widen access to credit, but models must be tested for bias, check affordability and be explainable to customers and regulators.