Espero AI
Insights  /  Credit & Risk

Credit's new frontier: lending to the un-scored

Thin-file borrowers aren't un-scorable — they're un-scored. Here's the data, the models, and the governance that let East African institutions extend responsible credit to millions who've never held a formal credit history.

E
Espero ResearchCredit & Risk
20 AUG 2026
9 MIN READ

Across East Africa, the single biggest barrier to credit isn't willingness to lend — it's visibility. A bank or SACCO can only price the risk it can see, and for tens of millions of people, the traditional file is nearly empty: no salaried history, no collateral registry, no bureau score. So the default answer becomes no.

That answer is increasingly expensive. Mobile money, agricultural cooperatives, utility payments, and merchant transactions now generate rich, continuous records of financial behaviour — they simply don't live in the places a conventional scorecard looks. The institutions that learn to read these signals responsibly won't just grow their books; they'll reach a market their competitors have written off.

Where the signal actually lives

Alternative data isn't a single dataset; it's a portfolio of behavioural traces, each with different predictive power and different governance weight. In our work across the region, four categories consistently carry signal:

  • Mobile-money velocity and stability — not just how much flows, but how regularly, and how resilient the pattern is to shocks.
  • Cooperative and group history — repayment behaviour inside SACCOs and chamas is among the strongest predictors of individual reliability.
  • Merchant and till records — for the informal-sector entrepreneur, transaction consistency stands in for a payslip.
  • Bill and utility payments — a quiet but durable indicator of financial discipline.
The core insight

A thin credit file and a thin behavioural record are not the same thing. Most 'un-scorable' borrowers are generating a rich data trail every single day — it just isn't being read.

From signal to score, responsibly

Turning these traces into a lending decision is where discipline matters most. A model that maximises approval accuracy while quietly encoding bias — against a region, a gender, an income band — is not an asset. It's a liability waiting for a regulator, and a breach of the trust an institution owes its members.

100%
of decisions carry human-readable reason codes
<90
days from pilot to a monitored production decision
4x
reviewable checkpoints across the model lifecycle
Illustrative: score distribution before and after adding alternative-data features — the thin-file population becomes assessable
Regulated finance can't run on black boxes. If a model can't explain a decline, it isn't a model — it's a risk.
Espero AI — model governance stance

What this means for your institution

The opportunity is concrete and near-term. A SACCO can responsibly extend its lending base to members it previously turned away. A micro-finance institution can price small-ticket loans that were never viable to underwrite by hand. A bank can meet the informal-sector customer where their financial life already happens.

None of it requires a moonshot. It requires the right data pipeline, models built to survive scrutiny, and the governance to stand behind every decision. That's the work — and it's exactly the work Espero was built to do.

E
Written by

Espero Research

The Espero AI team builds and deploys production-grade credit, fraud, and customer-intelligence systems for financial institutions across East Africa — with explainability and governance built in from the first line.

Espero AI Weekly

One sharp read on AI for finance, every week.

Join the leaders shaping AI at Africa's banks, insurers, and SACCOs. Research, playbooks, and field notes — no noise. Or book a consultation and we'll bring the thinking to you.

Double opt-in · one-click unsubscribe · we respect the Kenya Data Protection Act