● Alt-Credit Engine ACTIVE
◆ Unbanked Profiles Scored 48,200 today
● Approval Rate +340% vs bureau-only
◆ Alternative Signals 2,400+ analysed
● Score Accuracy 94.7% predictive
◆ Avg Decision Time 1.8 seconds
● New Borrowers Reached 2.4M+ this year
◆ Default Rate Comparable to bureau
● Alt-Credit Engine ACTIVE
◆ Unbanked Profiles Scored 48,200 today
● Approval Rate +340% vs bureau-only
◆ Alternative Signals 2,400+ analysed
● Score Accuracy 94.7% predictive
◆ Avg Decision Time 1.8 seconds
● New Borrowers Reached 2.4M+ this year
◆ Default Rate Comparable to bureau
AI-Powered Core Banking · Zero-History Credit Access

No credit file?
No barrier.
AI sees more.

zung.ai's alternative credit scoring engine analyses 2,400+ behavioural signals — airtime top-ups, mobile money patterns, merchant payments, and digital footprint — to score borrowers who are invisible to traditional bureaus, with accuracy that rivals prime lending models.

Request a Demo → View API Docs
AO
MN
EK
+
Trusted by 50+ lenders reaching 2.4M+ previously unscored borrowers with AI
UDESLY-START-PHPecho udesly_get_image(_u('i2b04decc', 'img'))->altUDESLY-END-PHP
zung.ai AltScore
● LIVE
AO
Amara Osei · Accra, Ghana
Bureau Status: NO FILE FOUND
MTN MoMo
724
Alt Score
▲ Good
📱 Airtime top-up regularity
+88
💸 Mobile money behaviour
+76
🛒 Merchant payment history
+82
📊 Digital footprint stability
+65
✅
Decision — Pre-approved
GHS 2,400 LOAN · 12 MONTHS · 1.8s DECISION
724
Alt Score
1.8s
Decision
2,400+
Signals
Bureau blank — AI scores anyway
GHS 2,400 approved ✓ 1.8s
Data sources
By the numbers

Lending to the invisible majority

Africa has 350M+ adults with no formal credit file. zung.ai's alternative scoring engine turns their digital behaviour into a creditworthiness signal — unlocking lending markets that traditional models cannot reach.

350M+
Credit-invisible adults
in Africa alone
340%
More approvals vs
bureau-only lenders
94.7%
Alt-score predictive
accuracy (12-month)
1.8s
Average credit decision
with full alt-data
Core capabilities

Credit intelligence built
for Africa's real economy

A unified AI scoring engine that ingests 2,400+ alternative data signals and returns a predictive credit score for any borrower — with or without a bureau file — in under 2 seconds.

Mobile Money Behaviour Scoring
Transaction frequency, average balance, send/receive patterns, top-up regularity, and seasonal consistency across M-Pesa, MTN MoMo, Airtel Money, and 12 other African mobile money networks — aggregated into a high-signal creditworthiness indicator.
15 Networks · Real-time
Airtime & Utility Pattern Analysis
Airtime top-up cadence, prepaid electricity purchase patterns, utility payment consistency, and data bundle subscriptions — each revealing income regularity, bill payment discipline, and lifestyle stability that strongly predict repayment behaviour.
Airtime · Utilities
Merchant & Ecommerce Signals
Payment frequency at merchants, average transaction values, repeat purchase behaviour, and ecommerce activity on Jumia, Konga, and local platforms — revealing consumer reliability and household income stability beyond what bureaus can see.
Merchant Data
Digital Footprint & Behavioural AI
Telco behavioural signals — call patterns, location stability, network tenure, device consistency — combined with app usage patterns and digital financial behaviour, weighted by an ensemble AI model trained on 50M+ African repayment outcomes.
50M+ Training Records
SHAP Explainability & Transparency
Every credit decision explained in plain language — which signals drove the score up or down, with SHAP values surfaced to your credit officers and, where regulations require, to the borrower themselves. No black-box lending decisions.
Explainable AI
Bureau + Alt-Data Hybrid Scoring
Seamlessly blends bureau data where available with alternative signals — producing a unified hybrid score that outperforms either source alone. For bureau-file customers, alt-data catches risks the bureau misses. For thin-file customers, alt-data does all the work.
Hybrid Score
How it works

From digital footprint to
credit decision in 1.8 seconds

01
Consent-first data collection
With the borrower's explicit consent, zung.ai queries their mobile money transaction history, airtime top-up records, utility payments, telco behavioural signals, and digital payment activity — pulling 2,400+ data points in under 400ms via pre-built API connectors.
Consent-First
02
AI features extracted & weighted
The raw data is transformed into 200+ engineered features — transaction velocity, balance stability, payment regularity, income proxies, and behavioural consistency — and fed into an ensemble ML model trained on 50M+ African repayment records with documented outcomes.
50M+ Training Records
03
Score returned with full explanation
A 300–850 alternative credit score is returned in under 1.8 seconds — along with SHAP-based explanations identifying the top positive and negative drivers. Your credit officers see exactly why each score was generated, not just the number.
Explainable · 1.8s
04
Decision triggers loan workflow
The score feeds directly into your loan origination system — triggering automated pre-approvals for qualifying borrowers, human review queues for borderline cases, and clear decline reasons for rejections. The entire process from application to decision in under 2 minutes.
Auto-Decision
UDESLY-START-PHPecho udesly_get_image(_u('i2b04decc', 'img'))->altUDESLY-END-PHP
zung.ai AltScore
● LIVE
✓ Collect ⟳ Score Explain Decide
Data Collected — 400ms
Mobile Money Txns284 records
Airtime Top-ups47 events
Utility Payments12 months
Merchant Payments92 records
Consent captured2,400+ signals
AI Scoring — Ensemble Model
BUREAU
No File
→
ALT SCORE
724
200+ FEATURES · ENSEMBLE ML · 1.4s
SHAP Explainability — Drivers
Mobile money regularity
+88
Merchant payment history
+76
Airtime top-up cadence
+64
Balance volatility
-32
PLAIN-LANGUAGE REASON GENERATED
Loan Decision — Auto
✅ PRE-APPROVED
GHS 2,400 · 12 months 1.8s total
Loan triggeredReason generated
724
Alt Score
1.8s
Decision
94.7%
Accuracy
724 scored — no bureau needed
Approved GHS 2,400 ✓
Alternative data sources

2,400+ signals from
the life they actually live

Traditional bureaus see what people borrowed. zung.ai's alternative data engine sees how people actually behave with money every day — a far richer, more honest signal of creditworthiness.

📱
Mobile Money Data
Transaction frequency, balance patterns, send/receive ratios, and behavioural consistency across 15 African mobile money networks — the richest available signal of financial behaviour for unbanked populations.
Transaction velocity & frequency
Average balance stability
Inflow/outflow seasonality
Savings & goal-setting behaviour
28% model weight
📡
Telco Behavioural Signals
Airtime top-up cadence, data bundle subscriptions, call pattern consistency, network tenure, and device stability — each revealing income regularity and life stability in ways that strongly correlate with repayment.
Airtime top-up regularity
Network tenure (loyalty)
Data bundle subscription patterns
Device consistency over time
22% model weight
🛒
Merchant & Payment Data
Merchant payment history, repeat purchase behaviour, bill payment consistency for utilities and school fees, and grocery spend patterns — revealing household income, financial discipline, and lifestyle stability.
Merchant payment frequency
Utility & bill payment history
School fees payment discipline
Average spend per category
18% model weight
🌍
Digital Footprint & Identity
Location stability, digital activity consistency, app usage patterns, and social profile signals — aggregated into a digital stability score that indicates residential and employment consistency correlated with repayment.
Location & address stability
Digital activity consistency
Platform tenure & engagement
Device registration history
14% model weight
🏪
Ecommerce & Marketplace
Purchase history on Jumia, Konga, Kilimall, and other African ecommerce platforms — including order fulfilment, return rates, and review activity — revealing consumer financial discipline and disposable income levels.
Order fulfilment & returns
Purchase frequency trends
Average order value
Marketplace tenure
10% model weight
🏦
Bureau & Hybrid Blend
Where bureau data exists, it is blended with alt-data signals — with AI determining the optimal weighting based on data quality and recency. For thin-file customers, alt-data fills gaps. For prime borrowers, alt-data catches risks the bureau misses.
CRB Africa · TransUnion
Experian · Metropol
Dynamic bureau + alt weighting
Hybrid score outperforms both
8% bureau blend
Developer-first

Score any borrower
in one API call

Drop into your loan origination flow with a single API call. Score returned in 1.8 seconds with SHAP explanations, decision recommendation, and credit limit suggestion.

POST /v1/score
// Score a zero-history borrower
const score = await zung.credit.score({
  phone: "233241234567",
  consent_token: "cst_9kx2m",
  sources: ["momo", "telco", "merchant"],
  bureau_blend: true
});

// Response · 1.8s
{
  "score": 724,
  "band": "good",
  "suggested_limit": 2400,
  "shap_drivers": [{...}],
  "decision": "approve"
}
1.8s end-to-end scoring
Data collection, feature engineering, model inference, and SHAP explanation — all returned in a single response in under 1.8 seconds. Ready to embed in any loan origination flow.
SHAP explanations included
Every score includes a ranked list of positive and negative SHAP drivers — enabling human review, regulatory explanation, and adverse action notices for declined borrowers.
Model monitoring & drift alerts
Continuous model performance tracking — Gini, KS, PSI monitored weekly — with automatic alerts when population shift degrades predictive accuracy, and scheduled model retraining built in.
Responsible AI & compliance

Fair, explainable credit
by design

Every alternative credit score is generated from consented data, explained in plain language, auditable by regulators, and tested for demographic bias — because inclusive lending only works if it's fair lending.

Explicit Borrower Consent Required
SHAP Explainability on Every Decision
Adverse Action Notice Generation
Demographic Bias Testing (Fairness AI)
SOC 2 Type II · ISO 27001
CBK / CBN Credit Reporting Compliant
Consent Architecture
Granular, time-limited consent tokens — borrowers see exactly what data is used, can revoke consent at any time, and receive a plain-language summary of what drove their score.
Fairness Testing
Regular demographic parity, equal opportunity, and calibration checks run across gender, geography, and age cohorts — with documented bias testing reports available for regulatory review.
Regulatory Model Documentation
Full model documentation — training data lineage, feature definitions, validation methodology, performance statistics — ready for submission to CBK, CBN, or any other regulator requiring model approval.
Data Minimisation
Only the signals necessary for the scoring model are accessed — with automatic deletion of raw transaction data after scoring, and aggregated features retained rather than raw records.
Data connectors

80+ data sources,
one scoring API

Pre-built connectors to every major African mobile money network, telco, ecommerce platform, utility provider, and credit bureau — with new data sources continuously added to the model.

Lend to the invisible majority

350 million people
deserve a credit score

Join 50+ lenders using zung.ai to reach 2.4M+ previously unscored borrowers — with AI-powered alternative credit scoring that delivers 340% more approvals, 94.7% predictive accuracy, and a 1.8-second decision time.

Request a Demo → View API Docs
Live in 2 weeks 1.8s decisions Explainable AI SOC 2 Type II