Fragmented SME Data
Business profiles and transaction ledgers were distributed across operational sources and required centralized integration.
AI-Driven SME Credit Risk & Loan Eligibility Intelligence Platform
Developed an end-to-end banking analytics platform that integrates SME profiles and transactions, automates data quality and incremental loading, calculates explainable credit-risk scores, and delivers portfolio, anomaly, and loan-eligibility intelligence for credit officers.
Banking Risk Analytics
500+ SMEs
10,000+ Transactions
Explainable Risk Scoring
Early Warning Signals
1.
Business Challenge
SME profile and transaction data arrived from multiple operational sources, while manual processing and fragmented reporting delayed credit-risk assessment and limited trusted portfolio visibility.
Manual Ingestion & Validation
Thousands of records required repetitive preparation, increasing processing time and the risk of duplicates or inconsistent values.
Slow Credit Decisions
The absence of automated scoring delayed loan eligibility assessments and reduced operational efficiency.
Scalability & Integrity Requirements
High transaction volumes required incremental loading, strong data relationships, and prevention of orphaned records.
Limited Portfolio & Anomaly Visibility
Business users lacked trusted, current insight into industry risk, liquidity stress, transaction failures, and anomalous behaviour.
2.
Solution
A governed Microsoft Fabric, SQL Server, Python, and Power BI architecture automated the complete flow from raw SME data to explainable lending intelligence.
Automated Fabric Ingestion
Parallel Copy Data activities moved SME Profiles and SME Transactions into OneLake Bronze/Lakehouse staging.
Dataflow Gen2 Quality Controls
Applied cleansing, duplicate removal, null checks, format validation, and banking business rules before loading.
SQL Star Schema
Dim_SME_Master and Fact_SME_Transactions were linked by BusinessID with primary/foreign-key controls and performance indexing.
Incremental Stored-Procedure Loading
Validated data was appended efficiently while SQL checks confirmed row counts and referential integrity.
Python Risk Scoring Engine
Calculated credit performance, cash-flow margin, failed transactions, DSCR, volatility, and overall risk scores.
Explainable Weighted Scoring
Used a transparent 0–100 model with 30% Credit, 40% Cash Flow, and 30% Stability pillars.
Four-Tier Loan Decision Framework
Classified SMEs as Green, Amber, Red, or Maroon to balance automation with human credit review.
Power BI Risk & Anomaly Intelligence
Delivered executive exposure, industry heatmaps, approval funnels, transaction drill-down, and AI-assisted anomaly detection.
Scheduled Daily Refresh
Automated synchronization at 00:00 AST ensured previous-day transactions were available for updated risk assessment.
3.
Business Impact
SMEs Risk-Tiered
The solution supported automated portfolio classification across a large synthetic SME population.
Transactions Analyzed
Daily credit, debit, completed, pending, and failed movements were incorporated into credit-risk analysis.
Actionable Risk Tiers
Green, Amber, Red, and Maroon classifications supported fast approval, review, rejection, and critical-risk escalation.
Transparent Scoring Pillars
Credit, Cash Flow, and Stability components created an explainable audit trail for every risk decision.
Anomaly Confidence Interval
Power BI anomaly detection flagged transactions that deviated materially from historical SME behaviour.
Automated Risk Refresh
Scheduled midnight processing delivered updated portfolio and loan-eligibility intelligence without manual intervention.
Faster, More Consistent Loan Eligibility Decisions
Automated risk scoring reduced reliance on manual review and standardized decision logic across the SME portfolio.
Earlier Warning of Liquidity & Default Risk
Cash-flow margins, transaction failures, late payments, and anomaly signals surfaced deteriorating businesses before default.
Improved Governance & Human Oversight
Explainable rules and an Amber review tier supported regulatory transparency while preserving human judgment.
Trusted, Scalable Banking Analytics Foundation
Automated validation, incremental loading, and star-schema modeling established a reusable platform for future banking data sources.
Quantification note: The documents did not provide measured time or financial savings, so the quantified impact above reflects documented portfolio scale, risk-model structure, analytical coverage, and refresh cadence only.
4.
Solution Architecture
SME Credit Risk Intelligence Architecture
SME ProfilesBusiness and Credit Attributes
SME TransactionsCredit, Debit, Status
Operational SourcesDaily Banking Data
↓↓↓
OneLake BronzeFabric Staging
→
SQL Star SchemaDimension + Fact
→
Python Risk EngineScores, Tiers, Eligibility
↓
Power BI SME Risk & Portfolio Intelligence Suite
The architecture combines governed ingestion, relational modeling, explainable scoring, and interactive decision support.
5.
Analytics Delivery Flow
Ingest
Profiles and transactions
Clean
Validate, deduplicate, enforce rules
Model
Dim_SME_Master and Fact_SME_Transactions
Score
DSCR, cash flow, failures, risk tiers
Decide
Eligibility, anomalies, portfolio monitoring
Trusted operational data is converted into explainable credit decisions and actionable SME portfolio intelligence.
6.
Technology Stack
Microsoft FabricFabric Data FactoryOneLakeFabric LakehouseDataflow Gen2SQL ServerT-SQLStored ProceduresPythonPandasNumPyPower BIDAXStar SchemaAnomaly DetectionIncremental Loading
7.
Dashboard Gallery
8.
Project Showcase
Live Power BI Dashboard
Replace this placeholder with the public Power BI embed URL after publishing the report.
Video Walkthrough
Replace this generic YouTube URL after publishing the walkthrough.
9.
GitHub Repository
Explore the SME Credit Risk Analytics Project
The repository contains project documentation, risk-scoring logic, data architecture, dashboard assets, and implementation details.
Explore More Case Studies
Discover more industry transformation and enterprise analytics projects.