In the past, for many lenders in Kenya, digital lending involved using an algorithm to approve a loan. This algorithm analyzed how often a borrower tops up airtime, how many apps they have installed, whether they charge their phone at night, and how many contacts they have saved.
In under three seconds, it has declared the person creditworthy. The loan has been disbursed, and the fees have been applied. And nobody, not the algorithm, not the lender, not the regulator, has asked the most fundamental question in all of lending: can this person actually afford to repay?
This is not a fringe scenario. It describes how credit has been issued to millions of people across Africa over the past decade. And regulators, watching default rates pile up on one end while borrowers spiral into debt traps on the other, have finally decided it needs to stop.
Let's explore what happened.
The Regulation That Changes Everything
In March 2026, Kenya's Central Bank, Capital Markets Authority, and Communications Authority jointly released a Financial Consumer Protection Framework draft that is direct in its language and far-reaching in its implications. A Financial Services Provider shall not provide a credit product unless they have first undertaken a reasonable assessment to confirm the retail consumer's ability to repay the credit without financial hardship.
More significantly, the framework defines what that assessment must involve. Lenders must base that assessment on "appropriately reliable information" about a borrower's financial position, including income, expenses, and existing obligations. The requirement sets a baseline for how credit is issued, limiting the use of models that rely primarily on behavioral data or predictive scoring.
That last clause is the one that should stop every digital lender operating in Africa in their tracks. Because behavioral and predictive scoring, including airtime patterns, app metadata, repayment loops, and device signals, is exactly what the industry has built its entire credit infrastructure on.
Nigeria has been moving in the same direction. The FCCPC's Digital, Electronic, Online, or Non-Traditional Consumer Lending Regulations, 2025, require lenders to conduct adequate credit assessments to ensure borrowers can repay loans sustainably, with fines of up to ₦100 million or 1% of turnover for non-compliance.
Both markets are telling lenders the same thing: proxy data is no longer sufficient. You need to know what someone actually earns and owes.
What Behavioral Scoring Gets Right And Where It Breaks Down
To understand the gap Kenya's new framework is trying to close, it is worth being clear about what alternative credit scoring measures and what it cannot.
Behavioral models that use mobile phone data are genuinely good at predicting whether someone who behaved a certain way in the past will behave similarly in the future. Airtime recharge patterns, mobile money activity, and consistent repayment on previous loans all carry real predictive signals.
Research has validated that a credit scoring model using mobile phone usage patterns, including call frequency, data usage, and mobile money transactions, achieved 90% prediction accuracy compared to less than 80% when using traditional methods.
But here is where it breaks down. Behavioral scoring predicts repayment willingness, the likelihood that a person with a given pattern of behavior will choose to pay. What it cannot measure is repayment capacity, whether a person, at the moment a loan is issued, actually has the income to service it.
For example, a market trader who topped up ₦500 in airtime every three days for six months is probably disciplined with little money. That pattern tells you something meaningful about her character. It tells you very little about whether her income this month can cover a new loan instalment, her rent, her children's school fees, and her restocking costs simultaneously.
Conventional models typically rely on demographic and financial data, which may not accurately reflect a borrower's ability to repay loans, especially for those in the informal economy.
This is not a minor flaw in the model. It is a structural blind spot. And over years of high-volume, low-scrutiny lending, that blind spot has accumulated into a crisis.
The Data Behind the Crisis
Data from the Central Bank of Kenya show that loans below KES 1,000 had default rates of more than 80%, while loans between KES 1,000 and KES 5,000 had default rates of about 69%. Overall default rates for digital lenders have been reported as high as 40%, more than double those in the banking sector.
The 2021 FinAccess Household Survey in Kenya found that default rates among borrowers using mobile banking loans were 50.9%, significantly higher than those at microfinance institutions (30.8%), banks (22.1%), and SACCOs (16%).
In Nigeria, the picture is similar. Nigeria's banking industry's NPL ratio stood at about 7% in 2025, above the 5% threshold, with industry executives openly acknowledging that members are shifting away from high-risk, small-ticket nano loans toward quality and customers with verifiable income to reduce non-performing loans.
Across the industry, the tiny loans that built Nigeria's loan app ecosystem are quietly disappearing, as digital lenders retreat from ₦5,000 to ₦10,000 loans and shift toward larger credit products and borrowers with more predictable income.
These default rates are not just a financial problem for lenders. They represent millions of people pushed further into financial hardship by credit that was never properly calibrated to what they could afford. The borrowers who took small loans they could not repay did not do so recklessly; they were approved by systems that never asked whether repayment was realistic.
In addition to over-indebtedness and high default rates, illegal debt-collection methods and violations of data privacy laws have been reported as further risks to consumers. The debt crisis and the data abuse crisis are the same crisis, flowing from the same root: a credit system that expanded access without ever building the infrastructure to confirm affordability.
The Three Questions Every Lender Must Now Answer
Kenya's new framework, and Nigeria's equivalent in the FCCPC DEON Regulations, effectively require every lender operating in these markets to answer three questions before issuing any credit:
1. What does this borrower actually earn?
Not what their spending patterns suggest they earn, or what the model predicts their income tier is based on app usage. What they actually receive into their account, at what frequency, from what sources, and how stable is it over time?
2. What are their existing obligations?
Not a credit bureau check alone, which, as we have written about at length, covers only a fraction of Africa's borrowing population. What recurring commitments — rent, utilities, prior loans, school fees, dependants — does this person already carry, and how do those obligations shift what they can actually repay?
3. What does their real financial behavior show?
Not proxy behavior, but actual transaction behavior. This includes income inflows, expense patterns, savings accumulation or erosion, seasonal variation, overdraft frequency, and the delta between what comes in and what goes out after necessary costs are deducted.
The problem is not that lenders do not want to answer these questions. Many do. The problem is that, until recently, the data infrastructure to answer them reliably at speed, at scale, and for borrowers who have never had a traditional financial profile has not existed in these markets.
That has changed.
Why Behavioral Proxies Became the Default
It is worth being fair about why African digital lending built its models the way it did. The alternative, waiting for traditional credit infrastructure to develop, would have left the majority of the continent without access to credit for another generation.
BCG's 2026 analysis of Africa's second fintech wave notes that large transaction datasets, expanding digital ID programs, and emerging alternative lenders provide the raw material for scalable credit, but weak credit bureaus, thin-file SMEs, and high capital costs continue to limit risk-based lending at scale.
When lenders entered markets where 70% of borrowers had no bureau history, no payslip, and no formal employment record, they had to work with the data that existed. Mobile phone data, app metadata, and repayment history on small airtime loans existed.
The problem was not in using that data, but in treating it as a complete picture of creditworthiness rather than a partial signal of willingness to repay. Presently, lenders increase borrowing limits based on a customer's repayment history, without fully assessing their ability to take on additional debt.
A borrower who repaid a ₦5,000 loan gets a ₦15,000 offer. Then ₦50,000. At no point is anyone measuring whether their income has grown proportionally to their approved credit limit.
This is how you get a population of borrowers who are technically creditworthy by the models' definition — they have always repaid before — and deeply over-indebted by any reasonable affordability calculation. The models were optimizing for the wrong outcome.
What "Appropriately Reliable" Income Data Means in Practice
Kenya's framework deliberately uses the phrase "appropriately reliable information". It does not mandate that every lender conduct a full audit of a borrower's finances before every loan. It asks lenders to use data sources that are verifiable, current, and directly reflect a borrower's actual financial position, rather than being inferred from proxy signals.
In practice, for the vast majority of African borrowers who are not formally employed, this means one primary source: bank and mobile money transaction data.
Equity Group Holdings demonstrated this at scale in 2025, accounting for nearly 45% of Kenya's MSME loan disbursements by deploying machine-learning credit models that analyzed real-time transaction data from its digital merchant ecosystem. This allows algorithms to parse daily cash flows and allocate capital in seconds, often bypassing the need for traditional collateral or land titles.
Analysts conclude that consistent cash flow has become the new collateral. This is the infrastructure gap that the Kenyan regulation exposes for every lender that is not Equity Bank, and that is most of them.
To answer the affordability question compliantly, lenders need access to verified transaction data that shows real income and real expenses. That requires open banking infrastructure: a standardized, consent-based mechanism for pulling transaction data directly from a borrower's bank account, in real time, without asking them to submit documents that can be manipulated or forged.
The Infrastructure That Closes the Gap
This is what Zeeh's product stack is built to deliver, and why the regulatory shift in Kenya, combined with the equivalent pressure in Nigeria, makes it directly relevant to every lender operating on either side of this transition.
Zeeh Connect addresses the affordability data problem directly. Using Nigeria's open banking framework, which went live in 2025, making Nigeria the first African country to operationalize it, Connect pulls verified transaction data from a borrower's bank account with their explicit, revocable consent.
Then, Zeeh Insights converts that raw transaction data into structured credit intelligence. This income categorization distinguishes consistent salary inflows from irregular trading income, cash flow stability scores that capture whether income is trending up or down, spending pattern analysis that surfaces recurring obligations, and affordability calculations that answer how much a borrower can realistically repay without financial hardship.
Cash flow underwriting using real transaction data allows lenders to detect potential cash flow issues early and mitigate risks effectively, leading to better-informed underwriting decisions and a more accessible lending environment.
Together, these two products give lenders the infrastructure to answer the three questions that Kenya's new framework demands, at the speed African digital lending requires, without reverting to the document-heavy processes that excluded borrowers in the first place.
Final Thoughts
Kenya's regulation is a signal that will travel. Nigeria's FCCPC DEON Regulations arrived first. Ghana, South Africa, and the broader ECOWAS region are watching what both markets produce. BCG identifies a critical imperative for Africa's second fintech wave: transaction data must become credit infrastructure, enabled by secure data-sharing frameworks, AI-enabled underwriting, and emerging open banking reforms.
The markets that build this infrastructure first will not only be compliant earlier than their competitors but also have better credit portfolios. They will approve more of the right borrowers, including millions of traders, gig workers, and informal operators whose transaction histories demonstrate real financial discipline that proxy models miss, and decline more of the wrong ones. The accuracy advantage compounds over time as the models learn from verified outcomes rather than proxy predictions.
Ability to repay was never an unreasonable ask. It was simply a question that the industry lacked the data infrastructure to answer. That infrastructure now exists.
If your lending model still depends on behavioral scoring alone, the shift has already begun. The good thing is Zeeh helps you move from assumptions to verified affordability, using real financial data, structured insights, and compliant infrastructure built for African markets.
Talk to our team to see how Zeeh Connect and Zeeh Insights give lenders the data infrastructure to answer the affordability question compliantly and accurately.
