Why Nigerian lenders should replace harassment with algorithms

Temidayo Akindahunsi

There is a unique paradox in Nigeria’s digital lending industry. It is among the most technologically advanced areas of the financial services ecosystem, offering real-time credit, immediate payouts, and convenient mobile platforms. Yet once loans are due and payments default, most lenders turn to harassment, mass messaging, aggressive calls, public humiliation, and indiscriminate contact-list shaming.

An industry capable of building real-time underwriting engines continues to enforce repayment with methods that predate data science.

The use of coercion is not the most effective way to collect loans. It is popular because it requires little analytical ability, is easy to deploy, and produces immediate signals of activity.

In the absence of behavioural insight, pressure becomes the default substitute for understanding. Having spent years reviewing delinquency dashboards and sitting in escalation meetings across Nigeria’s digital lending sector, I have watched well-intentioned teams confuse intensity with effectiveness. The pattern is consistent: harassment feels decisive but teaches nothing. Algorithms feel slower to trust, but they compound insight with every cycle.

The Problem
One of the most damaging simplifications in Nigerian digital lending is the assumption that borrowers are either “good payers” or “defaulters.” This binary framing collapses a wide spectrum of repayment behaviour into a single category, undermining collection effectiveness. A critical distinction most systems fail to make is between temporary delinquency and structural default. Temporary delinquency arises when borrowers intend to repay but face short-term liquidity shocks, delayed salaries, irregular income cycles, or delays in inventory turnover. Structural default reflects a deeper inability or unwillingness to repay.

In Nigeria, liquidity timing plays a decisive role. A large proportion of borrowers earn income weekly, biweekly, or irregularly. Salary earners may be paid late; informal workers experience uneven cash inflows. Repayment probability fluctuates predictably around these cycles. Yet coercion-first collections ignore timing entirely, escalating pressure immediately rather than aligning interventions with expected liquidity windows. A trader waiting for inventory turnover and a borrower with no intention of repaying are treated identically: both get harassed from day one.
Coercion-based collections appear effective because some borrowers pay following aggressive contact. But this is misattribution. Many would have paid anyway once liquidity constraints eased. Meanwhile, early aggression triggers avoidance, SIM changes, app deletion, and number blocking, particularly among digitally savvy borrowers. What might have been recoverable short-term delinquency becomes permanent loss.

One of the most damaging extensions of this model is the use of borrowers’ mobile contact lists as a collections tool, contacting friends, relatives, and workmates to apply social pressure. This practice substitutes reputational damage for behavioural understanding and exposes lenders to significant regulatory risk. Harassment also destroys customer lifetime value. Repeat borrowing, cross-sell opportunities, and referrals form a substantial portion of long-term portfolio profitability, yet borrowers who repay under duress rarely return. Coercion-based collections have no feedback loop; they extract value without improving future outcomes.

The regulatory environment has shifted decisively. The Federal Competition and Consumer Protection Commission’s Digital Lending Regulations now expressly prohibit harassment, unauthorised third-party contact, and excessive messaging. Between August and September 2023, the FCCPC pushed back against illegal loan apps operating without approval and asked platforms like Google to remove them from app stores, leading to dozens of apps being delisted as non-compliant.

The Algorithmic Alternative
Algorithmic collections replace pressure with structured decision-making: using data to determine whom to contact, when, how, and critically, when not to act at all. The premise is that repayment behaviour is predictable, heterogeneous, and responsive to timing and context.

At the foundation is a segmentation framework I call the Willingness-Capacity Matrix. Borrowers with high willingness and high capacity need minimal intervention, often just a well-timed reminder. High-willingness but low-capacity borrowers benefit from deferrals or restructuring. Low-willingness but high-capacity borrowers require targeted escalation. Low-willingness and low-capacity cases should be deprioritised early. This framework ensures resources are deployed where the marginal recovery probability is highest.

Propensity-to-pay models sit at the centre of this approach. They predict repayment likelihood using inputs such as repayment history, proxies for income regularity, past response to contact, and failed payment attempts. The most powerful application is not deciding whom to contact, but who not to contact. Suppressing unnecessary activity reduces cost, prevents borrower fatigue, and preserves goodwill without sacrificing recovery.

Beyond segmentation and propensity scoring, algorithmic systems treat Days Past Due as a trajectory rather than a status, optimise contact timing around liquidity windows, match channels to borrower preferences, and interpret failed payment attempts as intent signals rather than failures. Each cycle refines the model.

What I Have Seen Work
When algorithms lead the collection strategy, contact volumes fall sharply while recovery per attempt rises. In one portfolio I managed, we deployed an algorithm segmenting SMS campaigns by prior repayment behaviour. Without changing message content, targeting alone delivered more than 100% improvement in conversion rates compared to untargeted mass campaigns.

Applying the willingness-capacity matrix revealed that roughly one in four customers with broken payment arrangements had actually attempted repayment but failed. When this group received appropriately timed, lower-friction contact, same-day response rates were eight times higher than those from generic escalation. Aligning contact with borrowers’ historical payment windows improved promise-to-pay fulfilment by 5 to 15 percent.

The operational impact extended beyond recovery metrics. Collections teams shifted from chasing noise to resolving cases with realistic recovery chances. Attrition rates fell. Borrower hostility declined as interactions became predictable and proportionate. Trust was preserved even during delinquency, making reactivation and repeat borrowing viable rather than exceptional.
In practice, the most effective implementations are hybrid. Algorithms determine prioritisation, contact timing, and channel selection, while human agents handle negotiation, hardship cases, and high-value customers. This balance matters in an economy where income volatility, informal employment, and sudden shocks are common. Human judgment adds value when context is messy; algorithms ensure that judgment is applied selectively, not indiscriminately.

The Choice Ahead
Early in my career, I believed collections were fundamentally about pressure, that non-payment was a behavioural weakness that could be rectified through escalation. Prolonged exposure to actual portfolios changed that view. The same borrowers unreachable at day five would repay voluntarily at day twelve. What appeared effective in isolation looked destructive in aggregate. Cohort analysis revealed that each escalation round pulled forward cash at the expense of future recoverability.

Nigerian digital lenders now face a choice that cannot be deferred. Rising acquisition costs, tighter regulation, and increasingly sophisticated consumers make coercion unsustainable. Algorithms are not a buzzword or moral posture; they are an operational discipline that clarifies whom to contact, when to wait, and when to walk away. The future of credit enforcement belongs to lenders who value intelligence over activity.

Temidayo Akindahunsi is a Data Professional working at the intersection of fintech, credit risk, and behavioural analytics. He develops machine learning models and algorithmic decision frameworks that improve recovery outcomes for digital lending portfolios.