Behavioral Credit Scoring Engine for Instant, Explainable Lending

Most existing customers have no bureau-grade credit history, so lending still runs on salary slips and manual appraisal. Stixor's engine turns the bank's own account data into explainable scores and pre-approved offers, delivered in under a second inside the bank's perimeter.

<1s

Decision Latency

100%

Decisions with Reason Codes

+18%

Approval Rate Uplift at Same Risk Appetite

15%

Default Rate Reduction on the Scored Book

Business Overview and Strategic Direction

EXECUTIVE SUMMARY

Banks hold the strongest predictor of repayment they have: how each customer uses their account. Yet consumer and SME lending still depends on salary slips, manual appraisal and conservative limits, because most existing customers have no bureau-grade credit history. Good customers are declined or under-served, turnaround takes days, and transaction data sits unused in core banking.

Stixor built a behavioral scoring engine that scores every existing customer from their own account behavior, pre-approves limits for personal loans, cards and SME working capital, and returns each decision in under a second with reason codes. It runs on the bank's data inside the bank's own environment, and every score is explainable, monitored and documented under model governance.

Operational Challenges

Lending to existing customers should be the bank's easiest growth channel, but the data and process were not set up for it. Key challenges:

PROBLEM STATEMENT

Thin-File Customers

Most existing customers have no bureau-grade credit history, so salaried and self-employed customers are declined or given low limits despite years of healthy account activity.

Slow, Paper-Based Appraisal

Lending relies on salary slips, manual review and branch visits, so a simple personal loan or card decision takes days instead of seconds.

Unused Behavioral Data

Transaction, salary-credit, balance and channel data that best predicts repayment sits in core banking and is not used in credit decisions.

Regulatory and Explainability Demands

Any automated decision must be explainable to customers and auditors, monitored for drift and documented under the bank's model risk governance.

Behavioral Scoring Platform for Pre-Approved Lending

An end-to-end scoring platform that turns account behavior into explainable credit decisions and pre-approved offers across app, branch and SME channels.

Our Solution

Behavioral Risk Modeling

Gradient-boosted models and points-based scorecards trained on transaction, salary-credit, balance and channel behavior estimate each customer's probability of default (PD), including thin-file and salaried customers.

Instant Decisioning and Offers

A real-time decisioning API returns a score, a risk-based limit and the top reason codes in under a second, feeding pre-approved offers for personal loans, credit cards and SME working capital in the app and branch.

Governance and Monitoring

Every score carries reason codes, and models are tracked on Gini, KS and population stability index (PSI), with validation reports, model documentation and audit trails that support model risk management (MRM) and regulator review.

Behavioral Feature Engineering

Salary-credit regularity, average and minimum balances, inflow-outflow ratios and channel usage built from core banking data
Early-warning signals of financial stress, such as falling balances or missed salary credits, that bureau data often misses
Separate segment models for salaried, self-employed and SME customers

Decisioning and Limit Setting

Risk-based limits and score cut-offs calibrated to the bank's risk appetite
Pre-approved offers pushed to the mobile app, internet banking and branch staff
Champion-challenger testing before any credit policy change goes live

Explainability and Compliance

Reason codes on every approval, decline and limit, ready for customer and auditor queries
Independent validation, model documentation and full audit logs
Fairness and stability checks across customer segments, plus behavioral scores for early-warning and IFRS 9 staging