Nigerian researcher develops data-driven model to strengthen fraud detection

Three years before a Nigerian bank collapses, the evidence is already sitting inside its own data. Irregular journal entries posted at midnight. Payments approved by the officer who raised them. Loan valuations that quietly drift from market reality quarter after quarter. The fraud is never invisible. It is simply that nobody is looking at everything, […]

Three years before a Nigerian bank collapses, the evidence is already sitting inside its own data. Irregular journal entries posted at midnight. Payments approved by the officer who raised them. Loan valuations that quietly drift from market reality quarter after quarter. The fraud is never invisible. It is simply that nobody is looking at everything, all at once, all the time. That structural blindness has defined Nigerian financial auditing for decades and it has cost this country more than any single administration has ever publicly admitted. A Nigerian finance professional at Barclays in the United States has now published a peer-reviewed, internationally recognised model for eliminating it completely.

Chukwudera Obumneke Anunagba, based at Barclays in Whippany, New Jersey, is the co-author of a landmark academic paper titled “Advanced Conceptual Model for Strengthening Audit Quality Using Data Analytics Across Financial Institutions,” published in the International Journal of Advanced Multidisciplinary Research and Studies, Volume 5, Issue 6, 2025, pages 2246 to 2268. His collaborators are David Excel Ozowara of Western Illinois University in Macomb, Illinois, and Abolaji Adebayo of East Tennessee State University, three Nigerian scholars whose combined expertise in financial governance, data science, and institutional risk management has produced one of the most practically useful audit reform frameworks to emerge from the African diaspora in a generation.

The paper begins with a diagnosis that every Nigerian banker, regulator, and depositor will recognise. Conventional auditing selects a small sample of transactions from what is often millions of daily entries, tests those samples manually, and extrapolates conclusions across the entire institution. It is a method designed for a slower, simpler financial world that no longer exists. In today’s banking environment, where electronic payments clear in seconds, where trading positions shift in milliseconds, and where sophisticated financial crime has evolved to exploit precisely the gaps that sampling leaves open, this approach is not merely outdated. It is, the authors argue, negligent.

Their solution replaces periodic, partial testing with a continuous, whole-population analytics framework that monitors every transaction an institution processes, every day, without exception. Imagine replacing a security guard who checks one in every five hundred visitors with a complete biometric surveillance system that tracks every single entry in real time and raises an alarm the instant something deviates from acceptable behaviour. That is the essential shift this model proposes, applied not to physical security but to the vast, complex flow of money moving through a financial institution at every hour of every working day.

The framework operates through layered intelligence. All financial data generated by the institution, including payments, loan records, collateral registers, trading positions, approval logs, and internal communications, is first subjected to strict quality controls before analysis begins. A detection engine then simultaneously applies rule-based testing to catch clear policy violations, statistical analysis to identify transactions that deviate from established peer patterns, machine learning models trained on records of historical financial misconduct, and process mining tools that reconstruct the actual path each transaction took through the institution’s systems. Any transaction that bypassed a required approval checkpoint, was posted outside authorised hours, or bears the fingerprint of a known fraud pattern is flagged immediately, not in the next quarterly audit cycle, but within hours of occurring.

Critically, Anunagba and his colleagues insist that every flag the system generates must be explainable in plain language. No black box. Every alert must state precisely which rule was violated, which statistical threshold was crossed, and which approval step was skipped, in terms that a Central Bank examiner, a board member, or a court of law can read, interrogate, and act upon. The paper includes a full implementation roadmap with phased rollout guidance, infrastructure requirements, staff training protocols, and alignment pathways for existing regulatory frameworks, making it immediately actionable for any institution willing to commit.

For Nigeria, the research lands at a moment of acute institutional urgency. Every major Nigerian banking crisis of the past three decades shared a common feature: the warning signals were present in the data long before the collapse became news. The difference between an institution that catches fraud in week two and one that discovers it in year four is, almost entirely, the sophistication of its audit architecture. Anunagba’s published framework gives Nigerian banks, the Central Bank, the NDIC, and the Securities and Exchange Commission a clear, tested, internationally peer-reviewed roadmap for building that sophistication now.