Credit Scoring and Automated Underwriting in Nigeria: Models and Decision Engines

A credit score estimates or organises risk according to a defined model; it is not the same as a complete lending decision. A useful automated-underwriting project connects data, approved policy, affordability, referrals and accountable review.

Wigmore Trading welcomes model-requirement, software-procurement and implementation-planning enquiries in Nigeria. Specific models, platform capabilities, integrations and professional responsibilities require confirmation. No approval engine or borrower-level credit decision is offered through this page.

Define the population and purpose

Business, SME and consumer scoring use different evidence and outcomes. State whether the project concerns application screening, pricing support, portfolio monitoring or another purpose. Explain what decision the output will inform and who owns that decision.

Assess data quality and lawful use

Document sources, coverage, missing values, errors and whether the data represents the intended borrowers. Alternative data is not automatically reliable, fair or lawful merely because it is available.

Review privacy, purpose, access, retention and automated-decision considerations with appropriate specialists. Consult NDPC resources for the Nigerian framework. Do not collect private contacts, device content or sensitive information without the required lawful basis and safeguards.

Test models before deployment

Use appropriate development, validation and monitoring methods. Evaluate predictive performance, calibration, bias, stability and the consequences of errors for the actual use case. An impressive vendor demonstration is not evidence that a model fits your portfolio.

AI and machine-learning tools require clear limitations, explainability and governance. Establish when a case must be referred and how corrections or challenges are handled. A probability of default estimate is not a certainty about an individual borrower.

Connect models to approved lending policy

Affordability, identity, fraud checks and product rules can remain important even with a score. The institution should define which inputs may affect a decision and who can approve exceptions.

Distinguish application scoring from IFRS 9 expected-credit-loss analysis. PD and LGD models may inform different tasks, but accounting and lending uses require their own scope, assumptions and professional review.

Compare the implementation package

Include data preparation, integration, testing, documentation, training, monitoring and change management. Ask who owns the models, rules and data, and what happens when a provider changes or exits.

Pilot with approved test data and a controlled release plan. Define service support and responsibilities for model drift, errors and security incidents.

Can automated underwriting remove all bad debt?

No. Models support decisions under uncertainty; they do not eliminate credit risk.

What should I send Wigmore?

Describe institution type, product, workflow, systems and model purpose. Avoid identifiable borrower data in the first enquiry.

Related underwriting guides

• Digital Insurance Underwriting in Nigeria: Software, Analytics and AI Workflows
• Business Loan Underwriting in Nigeria: Cash Flow, Appraisal and Repayment Capacity
• Credit Underwriting Controls in Nigeria: Policies, Outsourcing and Verification
• Loan Portfolio Reviews and IFRS 9 Modelling in Nigeria: Scope and Evidence
• Underwriting Services in Nigeria: Insurance, Financial and Support Roles

Prepared 4 November 2026. This guide does not establish a lending or insurance decision, a regulated appointment, bound cover, a funding commitment or a professional opinion.