By Winston Osuchukwu
Digital transformation has modernised the front end of the credit
process in Nigeria, streamlining customer journeys and shortening the
path from application to disbursement. However, this progress has not
reached the core of the credit process. While digital application flows
are now standard, the underlying risk infrastructure remains
underdeveloped. Following the withdrawal of the Central Bank of
Nigeria’s forbearance measures, the sector’s non-performing loan (NPL)
ratio climbed to 8.03% – well above the 5% regulatory limit.
The deeper, structural flaw is that banks still run on legacy risk
models and backward-looking data: an approach that leaves existing
portfolios exposed while shutting out the vast retail market. To scale
retail and SME credit safely, forward-looking institutions must close
three critical gaps in their core credit infrastructure.
1. THE BUREAU AND DATA BLIND SPOT
Many institutions rely on a fragmented view of borrower risk. Internal
transaction data offers a deep but narrow view of a borrower’s behaviour
within one institution, while periodic credit bureau reports provide a
broad but shallow, “negative-only” history across other lenders. Because
credit bureau coverage in Nigeria remains relatively low and data
sharing is often inconsistent, neither source effectively captures how a
borrower actually earns, spends, and repays. Resolving this requires
unifying the data architecture, integrating internal behavioural signals
with diverse external streams such as payroll, utility, and alternative
financial data, to build a continuous, real-time picture of cash flow
and true repayment capacity.
2. STATIC RISK ACCEPTANCE CRITERIA
To assess a borrower’s credit eligibility, banks apply internal risk
acceptance criteria that are often static. In a volatile macroeconomic
environment marked by shifting interest rates and inflation, a
borrower’s financial reality changes rapidly, rendering these rigid,
point-in-time benchmarks obsolete. Furthermore, out of caution, these
inflexible thresholds often default to conservative rejections for
unfamiliar applicants, such as new salaried employees or thin-file
borrowers – those with little or no formal credit history for a bureau
or bank to draw on – leaving profitable loans on the table.
Transitioning to a predictive model changes risk management into a
continuous, data-driven cycle. By ingesting high-frequency behavioural
data, risk systems can dynamically govern their acceptance criteria in
real-time, allowing them to adjust parameters, optimize pricing, and
deploy interventions well before a default occurs.
3. THE COLLECTIONS DISCONNECT
In many institutions, collections teams operate in silos downstream of
the credit department, meaning critical recovery performance data rarely
gets fed back to front-end risk models. Consequently, underwriting
systems fail to learn from actual repayment behaviours – repeating the
same structural pricing mistakes. Integrating these functions via a
direct data pipeline creates a self-learning loop, routing recovery
outcomes back into the origination engine. This empowers the risk engine
to dynamically update models, continuously refining underwriting
criteria based on real-world results to prevent future defaults and
capture lost basis points
THE BOTTOM LINE
Closing these gaps requires intentionality: moving away from
‘set-and-forget’ tools to systems that actively manage risk. It means
moving beyond fragmented data toward an integrated intelligence layer
that learns from borrower behaviour to govern automated decisions with
precision. The lenders that lead over the next year will be those that
treat credit not as an isolated transaction, but as a continuous,
dynamic process. At Mathesis, we have spent years building the engine
that makes this possible, powering over eight million loans for two plus
million Nigerians. The future of credit belongs to those who adopt this
predictive approach – and we have the proven tools and expertise to help
you get there.
