● Alt-Credit Engine SCORING LIVE
◆ Mobile Money Signals 2,400+ patterns
● Thin-File Approval Rate +67% vs bureau
◆ Score Latency 1.8 seconds
● Default Prediction 97.3% accuracy
◆ Data Sources Live 500+ signals
● Bureau-Free Scoring Active
◆ SHAP Explainability Always-on
● Alt-Credit Engine SCORING LIVE
◆ Mobile Money Signals 2,400+ patterns
● Thin-File Approval Rate +67% vs bureau
◆ Score Latency 1.8 seconds
● Default Prediction 97.3% accuracy
◆ Data Sources Live 500+ signals
● Bureau-Free Scoring Active
◆ SHAP Explainability Always-on
AI-Powered Core Banking · Alternative Credit Scoring
Credit scores
for the 2 billion
the bureau missed
zung.ai's alternative credit scoring engine uses mobile money history, sales data, behavioural patterns, and 500+ AI signals to score any borrower — even those with zero credit history — in under 2 seconds.
JK
AM
TO
+
Trusted by 100+ lenders scoring 30M+ thin-file borrowers annually
742
LOW RISK
ALTERNATIVE CREDIT SCORE
M-Pesa History
91%
Sales Patterns
84%
Behaviour Score
78%
Utility Payments
88%
Fraud Risk
5%
1.8s
Score Time
97.3%
Accuracy
+67%
Approvals
Score: 742 · Low Risk
Zero bureau data needed
Data sources
M-Pesa
Airtel Money
MTN MoMo
Open Banking
Experian
GSM Data
By the numbers
Credit intelligence that includes everyone
Two billion people are creditworthy but invisible to traditional bureaus. zung.ai's alternative credit engine sees them — and turns data signals into profitable, low-default lending.
97.3%
Default prediction model accuracy
67%
More approvals vs. bureau-only scoring
30M+
Thin-file borrowers scored annually
500+
Alternative data signals per score
Core capabilities
Score anyone, anywhere,
in any market
A modular AI scoring engine that combines every available data signal into a single, explainable credit score — bureau-augmented or bureau-free — in under 2 seconds.
Mobile Money Scoring
Analyse 24 months of M-Pesa, Airtel Money, and MTN MoMo transaction history — income regularity, merchant diversity, savings behaviour, and network centrality — to build a rich repayment probability model.
M-Pesa · MTN · Airtel
SME Sales Data Intelligence
For merchants and SMEs, our engine analyses POS transaction history, inventory turnover, supplier payment regularity, and seasonal revenue patterns to generate a business creditworthiness score.
SME-Ready
Behavioural Risk Modelling
Device usage patterns, app session behaviour, form-fill speed, and interaction signals feed our behavioural biometric model — catching fraud and predicting repayment intent before any money moves.
Passive Signals
Open Banking Score Fusion
When open banking data is available, our model fuses real-time bank transaction analysis — income verification, recurring payments, debt service ratios — with alternative signals for maximum accuracy.
Bureau + Alt Fusion
SHAP Explainability — Always On
Every score surfaces the top positive and negative factors with SHAP values — enabling credit officers to understand decisions, regulators to audit outcomes, and borrowers to receive actionable decline reasons.
Regulator Ready
Self-Learning Portfolio Models
Models continuously retrain on your own origination and default data — improving prediction accuracy with every loan cycle. Champion-challenger testing keeps your scorecard at peak performance automatically.
Auto-Retrains
How it works
From raw data to credit score
in under 2 seconds
01
Alternative data ingested
Mobile money APIs, POS sales feeds, open banking connections, and behavioural data streams are pulled via our secure data fabric — with customer consent and zero manual document uploads required.
Real-time Ingestion
02
AI extracts 2,400+ signals
ML models process raw transaction and behavioural data into 2,400+ predictive features — income regularity, payment velocity, cash flow patterns, seasonal rhythms, and digital footprint characteristics.
2,400+ Features
03
ML ensemble scores the borrower
A gradient boosting and neural ensemble model — trained on 50M+ loan outcomes — synthesises all signals into a single credit score, risk band, and recommended lending limit in under 1.8 seconds.
1.8s Average
04
Explainable decision delivered
The API returns the score, risk band, recommended limit, top signal drivers, and a regulator-ready adverse action notice — all in one response, in plain language your team can act on immediately.
SHAP Explained
✓ Data In
⟳ Signals
Score
Explain
Data Sources — Connected
M-Pesa API24mo ✓
POS Sales Feed18mo ✓
Open Banking12mo ✓
Digital FootprintLive ✓
No bureau neededConsent-based
Signal Extraction — 2,400+ Features
Income Reg.
94%
Pay Velocity
87%
Revenue
82%
Behaviour
79%
Digital
83%
PROCESSING 2,400+ SIGNALS...
ML Score Output
742
Score · out of 850
RISK BAND
Low Risk
MAX LIMIT
KES 120K
Decision: APPROVE ✓
1.8s · No bureau
SHAP Explainability — Top Drivers
↑M-Pesa consistency (24mo)+142pts
↑Revenue growth (6mo)+98pts
↓No formal bureau history-18pts
1.8s
Score Time
97.3%
Accuracy
2,400+
Signals
Scored — No bureau
+340% approvals ✓
Alternative data sources
Every signal that tells the real story
zung.ai ingests and models data from every source that reflects true creditworthiness — going far beyond the thin data that bureaus hold on most African borrowers.
Mobile Money
24 months of mobile wallet history revealing true income, spending discipline, and financial behaviour — higher predictive power than most bureau scores.
Income regularity & velocity
Float management patterns
Bill payment consistency
Peer transfer networks
Sales & Revenue Data
POS transaction records, e-commerce, and market stall data reveal an SME's actual business performance — the most predictive signal for merchant lending.
Monthly revenue consistency
Growth trajectory (12mo)
Seasonal pattern analysis
Customer repeat rate
Behavioural Signals
App usage patterns, transaction timing, device consistency, and geographic stability create a rich fingerprint that predicts repayment probability with high accuracy.
Transaction time patterns
Device & location stability
App engagement depth
Social graph signals
Open Banking
Live bank account data via PSD2 and local open banking APIs across 12+ markets — income verification, liability detection, and cash flow scoring in real time.
Salary & income verification
Existing debt obligations
Cash flow surplus scoring
Bank relationship depth
Digital Footprint
Consented digital behaviour data — device type, connectivity, usage consistency — enriches the profile of borrowers with limited financial history.
Device fingerprint stability
Network & connectivity
Email account age
Phone plan tenure
Bureau Augmentation
When traditional bureau data exists, our AI blends it with alternative signals — creating a hybrid score up to 40% more accurate than bureau-only models.
TransUnion · Experian · CRB
Hybrid model blending
+40% predictive accuracy
Thin-file override logic
Developer-first
One API call — complete credit intelligence
Single REST endpoint returns score, risk band, recommended limit, and SHAP explanations in under 2 seconds. SDKs for Node, Python, Java, and Go.
POST /v1/score
// Request alternative credit score
const score = await zung.credit.score({
customer_id: "cust_9x21k",
sources: ["mpesa", "pos", "behaviour"],
include_shap: true
});
// Response · 1.8s avg
{
"score": 742,
"band": "low_risk",
"max_limit": 120000,
"decision": "approve",
"top_drivers": [{...}],
"bureau_used": false
}
const score = await zung.credit.score({
customer_id: "cust_9x21k",
sources: ["mpesa", "pos", "behaviour"],
include_shap: true
});
// Response · 1.8s avg
{
"score": 742,
"band": "low_risk",
"max_limit": 120000,
"decision": "approve",
"top_drivers": [{...}],
"bureau_used": false
}
1.8s average response
Complete score, risk band, recommended limit, and SHAP explanations in a single API response — no polling.
Configurable data sources
Choose which alternative data sources to use per customer segment or product — mobile money only, sales augmented, or full hybrid.
Model drift detection
Continuous monitoring against actual loan outcomes — with automatic retraining alerts when score drift is detected on your portfolio.
Trust & compliance
Explainable, fair, and regulation-ready
zung.ai's alternative scoring models are designed for regulatory scrutiny — with built-in bias detection, full explainability, and adverse action compliance from day one.
SHAP Explainability — Every Decision
Fair Lending / Disparate Impact
GDPR & Consent Framework
Adverse Action Notice Generation
SOC 2 Type II · ISO 27001
CBK / CBN / CMA Compliant
SHAP Explainability
Every score shows ranked signal drivers. Regulators and customers always know why a decision was made.
Bias Detection
Automated disparate impact analysis across gender, ethnicity, and geography — with real-time monitoring and alerts.
Consent Management
Built-in consent capture and audit trail for every data source — GDPR, NDPR, and local data privacy law compliant.
Adverse Action Notices
Auto-generated, regulator-compliant decline notices with top reasons and improvement guidance for every borrower.
Integrations
Connects with your entire data stack
Pre-built connectors for every mobile money API, open banking provider, POS system, and core banking platform — live in weeks, not months.
M-Pesa API
MTN MoMo
Airtel Money
Plaid
TrueLayer
TransUnion
Experian
Temenos
Mambu
Salesforce
Square / SumUp
nCino
Start scoring the unscored
Lend to the 1.4 billion
with no credit history
Join 100+ lenders using zung.ai to unlock thin-file markets, reduce defaults, and score borrowers that traditional bureaus can't reach — with AI models that only get smarter over time.
No bureau required
Live in 4 weeks
SOC 2 Type II
SHAP explainable