The credit officer's blank-page problem.
An institutional credit officer evaluating an SME application typically has 60 to 90 minutes per file. The pack arrives with a 12-month bank statement, two years of unaudited financials, a tax-clearance certificate, and a one-page profile. The officer is being asked to reconstruct the business from a thin set of artefacts. The OECD has named information asymmetry as the binding constraint behind the global MSME finance gap for over a decade.
An SME applies for a working-capital facility. The application form arrives at the lender with a 12-month bank statement, two years of unaudited financials, a tax-clearance certificate, and a one-page business profile. The credit officer has, in most institutions, between 60 and 90 minutes to evaluate it.
In those minutes the credit officer is doing one thing: reconstructing the business from a thin set of artefacts. The artefacts arrive in different formats, from different time windows, with no narrative tying them together. The officer is making a credit decision on a partial reconstruction of operational reality.
The diagnosis is not new
The IFC estimates the global MSME finance gap at $5.2 trillion. The OECD's annual Financing SMEs and Entrepreneurs Scoreboard, tracking roughly fifty countries, has named information asymmetry between SMEs and lenders as a binding constraint behind that gap across multiple editions. The World Bank's SME Finance Forum, the EBRD's Small Business Initiative, and the AfDB's SME programme reviews repeat the finding in different language. The diagnosis has been stable for fifteen years.
What has not been stable is the operational response. Most policy answers - credit guarantees, blended finance, capacity-building grants - operate on the capital side of the asymmetry rather than the information side. The capital is delivered. The information layer that would let lenders read SMEs at decision-relevant granularity remains scattered across forms, spreadsheets, certificates, and credit-officer reconstructions.
What the application form does not capture
The application form is a snapshot. It does not capture the business as a moving system. Specifically, it consistently misses five categories that credit officers and risk teams care about most.
- Operating cadence: is the business running a monthly financial close, or producing financials only when asked? The single most predictive operational signal in SME lending, and the one application packs almost never expose.
- Customer concentration over time: how dependent is revenue on a single buyer, and how long has that buyer been a customer? A snapshot revenue split is not a customer-concentration history.
- Compliance freshness pattern: is the tax clearance current because the operator runs a monthly compliance cadence, or current because they scrambled to renew it for the application? A point-in-time document tells you nothing about the pattern.
- Forward-looking risk indicators: cash buffer in months of overhead, runway visibility, next major contract renewal date, foreign-exchange exposure. Backward-looking financials do not surface forward risk.
- Post-funding behaviour: how has this operator behaved after previous capital events? Did discipline improve, deteriorate, or stay the same? The application pack has no slot for the longitudinal answer.
None of these are unknowable. All of them require a record that existed before the application, kept consistently, available for read. The application form is not that record.
A credit officer who can see the operating cadence, the customer history, and the post-funding behaviour is making a different decision than one reconstructing the business from a PDF.
The regulatory direction is sharpening
Two regulatory developments are tightening the floor on what counts as adequate credit-decision data. The EU AI Act classifies credit scoring and creditworthiness assessment as high-risk AI uses, with documentation, risk-management, and human-oversight requirements that take effect in phases through 2026 and 2027. The US Federal Reserve's long-standing model risk management guidance (SR 11-7) sets standards for model validation, monitoring, and governance that institutional lenders increasingly apply to SME decision systems by analogy.
The Bank for International Settlements has published a series of reports on AI and machine learning in banking that converge on the same operational requirement: lenders need explainable, auditable inputs to credit decisions, not opaque scores. Thin application packs do not satisfy explainability or auditability. Structured evidence layers, with confidence-graded signals attributable to source, do.
The evidence layer lenders actually need
A useful evidence layer for SME lending contains five elements: graded evidence confidence per claim (self-reported vs document-verified vs live-data-verified vs third-party-verified), longitudinal operational signals (financial rhythm, cash buffer, document freshness), customer and offtake history with verifiable supporting documents, sector-specific risk markers (climate exposure for agri, supply concentration for manufacturing, regulatory dependence for professional services), and a structured channel for post-funding monitoring.
The lender does not need to build this layer. The SME does, with an instrument that captures the operator's reality in structured form before the application is needed. The application pack then becomes a derived artefact of the underlying record, not a hurried reconstruction.
What changes for the credit officer
Two things change. First, the application pack is no longer the totality of available information. The officer reads the longitudinal record behind the application - three years of monthly close, two years of compliance freshness, the customer-relationship history at confidence-graded rungs. Second, the post-funding view is built in. The same evidence layer that supported the application continues to update after disbursement.
The decision quality improves because the data quality improves. The relationship quality improves because the operator stops being a stranger every time the lender re-evaluates them. Post-funding surprises, the cost of which DFI portfolio reviews have repeatedly documented, become recoverable because the early-warning signals are already in the lender's view.
Where this leads
A platform-grade evidence layer is not a credit bureau, not a substitute for underwriting, and not a model that decides whether to lend. It is an instrumentation layer. It captures operational reality in structured form. The lender continues to decide on credit. The platform continues to capture the operational record. The two work together because they do different things.
The OECD has been recommending this operational fix in different language for over a decade. The Mothusi Growth Score is the published methodology that operationalises it.
- [1]IFC (2017, updated 2022). MSME Finance Gap: Assessment of the Shortfalls and Opportunities in Financing MSMEs in Emerging Markets.
- [2]OECD (2024). Financing SMEs and Entrepreneurs: An OECD Scoreboard - information asymmetry as binding constraint.
- [3]World Bank Group. SME Finance Forum - credit conditions and MSME finance data.
- [4]EBRD. Small Business Initiative - SME credit and advisory frameworks.
- [5]European Union. Artificial Intelligence Act - high-risk classification for credit scoring and creditworthiness assessment.
- [6]US Federal Reserve. SR 11-7: Guidance on Model Risk Management.
- [7]Bank for International Settlements (BIS). Reports on artificial intelligence and machine learning in banking.
- [8]IFC Independent Evaluation Group. Reviews of SME financial-intermediary portfolio outcomes.