E-Commerce: Predictive Churn & Strategic Retention
E-Commerce: Predictive Churn & Strategic Retention
Embedded Files
Read Case Study (Medium) -> link to Medium Article
View Source Code (GitHub) -> Link to GitHub Repository
For my Data Science Master's Capstone, my team partnered with Ora Fashion to solve a critical e-commerce bottleneck: high Customer Acquisition Costs (CAC) and unpredictable churn. While my partner developed the ML predictive models in Python, I architected the Strategic Command Center in Power BI to translate those probabilities into immediate marketing solution.
- Advanced Analytics: Engineered a robust Star Schema to compute dynamic RFM (Recency, Frequency, Monetary) Segmentation and Cohort Analysis, visually isolating 'At-Risk' customers and mapping the exact drop-off curve.
- Geo-Strategic Action Plan: Moved beyond generic dashboards by cross-referencing demographics, purchasing behavior, and ML churn probabilities to prescribe tailored retention plays across 4 distinct European clusters.
- Business Impact: Transitioned the company from retroactive reporting to proactive Customer Lifecycle Management, providing a data-driven blueprint to slash CAC and protect profit margins."
The VIP Paradox: Decoding Supply Chain Chaos with Python & Power BI
The VIP Paradox: Decoding Supply Chain Chaos with Python & Power BI
Read Full Article (Medium) -> Link to Medium article
View Source Code (GitHub) -> Link to GitHub Repository
A predictive analysis project addressing the counterintuitive high delay rates in priority shipments.An e-commerce giant faced a silent crisis: nearly 60% of shipments were delayed. My goal was to build a predictive ML model, but the data revealed a fractured operation where operations defied mathematics.
- The ML Diagnosis: After testing 5 algorithms (including XGBoost and Gradient Boosting), the ~67% accuracy ceiling proved that delays were chronic and chaotic, not logically patterned.
- The VIP Paradox (Power BI): Interactive dashboards revealed that 'High Importance' products and expensive Air Freight suffered the highest delay rates (>60%).
- Business Impact: Shifted the strategy from deploying a predictive algorithm to conducting an emergency warehouse audit and renegotiating SLAs, saving premium costs on failing services.
Predicting Credit Default: Saving Millions with a 61% Recall Rate
Predicting Credit Default: Saving Millions with a 61% Recall Rate
Read Case Study (Medium) -> Link to Medium article
View Source Code (GitHub) -> Link to GitHub Repository
In banking, lending blindly is catastrophic. I analyzed a portfolio of 30,000+ loans with an alarming 21.5% default rate to build an end-to-end risk mitigation solution.
- The ML Strategy: Prioritized Recall over simple Accuracy. Visualized complex data overlaps and deployed a k-NN model that successfully flagged 61% of actual defaulters before approval.
- The Power BI Diagnosis: Mapped the risk visually, revealing that 'G' grade loans are a near-guaranteed loss (98.4% default) and that Renters seeking Medical/Debt Consolidation loans drive the highest financial hemorrhage.
- Business Impact: This dual approach (ML gatekeeper + BI dashboard) allows institutions to transition from reactive debt collection to proactive, profitable lending
Corporate FP&A: Real Estate Financial Management (Real Case: Brazil)
Corporate FP&A: Real Estate Financial Management (Real Case: Brazil)
Read Case Study (Medium) -> link to Medium Article
View Source Code (GitHub) -> Link to GitHub Repository
Contracted as a Data Consultant, I transformed a massive residential condominium's retroactive, spreadsheet-based accounting into a proactive Financial Command Center using Power BI.
- Data Architecture: Designed a scalable Star Schema, implementing a dynamic Calendar Table (Time Intelligence) to seamlessly bridge daily operational cash flows with 2024/2025 strategic budgets.
- Business Logic (DAX): Centralized complex financial metrics into categorized measure tables, automating real-time calculations for Actuals, Forecasts, and critical Variance analysis.
- Business Impact: Eliminated manual data entry and delivered an interactive FP&A ecosystem. Leadership can now visually track 'Actual vs. Budget' discrepancies, transforming budget meetings into proactive capital allocation strategies.
Stay Tuned for next Projects
Stay Tuned for next Projects
Page updated
Google Sites
Report abuse