🇪🇦 South Africa · Databricks Azure · Confluent Group

DebtBusters Intelligence Platform

Enterprise data warehouse and ML solution for South Africa's leading debt counselling group. 80,000 clients · 3.96M rows · Medallion architecture · 5 validated ML models · 30-figure ebook.

Databricks Azure Delta Lake PySpark MLflow XGBoost LightGBM CatBoost Power BI Kimball Star Schema Folium

🌎 View Province Map 📚 Download Ebook 🔗 View on GitHub

Platform Scale

One Platform. 3.96 Million Data Points.

80,000

Total Clients Modelled

Across 9 SA provinces

3.96M

Total Rows in DW

8 entities, Bronze → Gold

500K

Lead Interactions

XGBoost AUC 0.74

1.5M

Payment Transactions

PDA collection tracking

ML Models Validated

XGBoost · LightGBM · CatBoost · RF · GBM

0.74

Best AUC Score

Lead Conversion — XGBoost

0.89

Credit Score R²

3/6/12-month forecast

Ebook Figures

Each with Audience/What/Why/Solutions

Geographic Intelligence

South Africa — Debt Counselling Province Map

Interactive province map showing average DTI severity, client volume, and risk scores across all 9 South African provinces. Click any bubble for full province stats. Bubble size = client volume.

● ≥75% DTI — Severe over-indebtedness

● 68–75% — High DTI

● 62–68% — Elevated

● <62% — Manageable

Machine Learning

5 Validated ML Models — Full MLflow Tracking

Lead Conversion

XGBoost · Hyperopt tuned

0.74 AUC

Prioritises call centre outreach — identifies 74% of convertible leads from 500K interactions.

Payment Default

LightGBM · SHAP explained

0.73 AUC

Early-warning system — flags at-risk clients 30 days before missed PDA payment.

Product Recommendation

Random Forest + Isotonic

85% Accuracy

Routes 85% of clients to the correct NCR product — reduces mis-sells and creditor disputes.

Credit Score Forecast

Multi-output Gradient Boosting

R² = 0.89

Predicts credit score at 3, 6, and 12 months — powers the client counselling roadmap.

Client Churn

CatBoost + Optuna

0.72 AUC

Detects withdrawal risk 30 days out — enables targeted retention calls for the top 200 at-risk clients monthly.

Deliverables

Everything Included. Click to Download or Explore.

📚

Intelligence Ebook (v7)

Full platform narrative — 30 embedded figures, each with Audience / What It Shows / Why It Matters / Solutions & Recommendations. Includes the Thandiwe data story, 10 strategic recommendations, and geographic intelligence maps.

⭳ Download Ebook (.docx)

📊

Executive KPI Report

907 KB Excel workbook with full KPI dashboards across Marketing, Operations, Finance, Credit Risk, Client Outcomes, and Creditor Management. Built for C-suite and Board presentation.

⭳ Download KPI Report (.xlsx)

🤖

ML Validation Report

Full ML validation workbook — AUC/R² scores, feature importance tables, cross-validation results, confusion matrices, and SHAP value summaries for all 5 models.

⭳ Download ML Report (.xlsx)

🌎

Interactive Province Map

Folium-powered interactive HTML map of all 9 SA provinces. Click any province bubble for full DTI, client volume, average debt, over-indebtedness rate, and risk score. Colour-coded by DTI severity.

👀 Open Full Map

💾

GitHub Repository

Full source code: 26 Databricks notebooks (Bronze/Silver/Gold/ML), Python data generation scripts, DAX measures, Kimball star schema SQL DDL, chart generation, and ebook generation.

🔗 View on GitHub

📋

Star Schema DDL

Kimball-compliant DDL — 5 dimension tables and 6 fact tables with SHA-256 surrogate keys, Z-ORDER clustering on Delta Lake, and NCA-aligned field naming for the debt counselling domain.

👀 View Schema SQL