← Back to Portfolio

📈 Demand Forecasting with Prophet

Python Prophet pandas NumPy SQL

65%

Reduction

23 → 8

Emergency Orders/Month

$18K

Annual Savings

The Problem

Emergency procurement was happening 23 times per month because demand wasn't being predicted. The procurement team was reactive rather than proactive, causing stockouts and premium shipping costs.

Architecture Overview

📊 Historical Data → ⚙️ Feature Engineering → 📈 Prophet Model → 🎯 Daily Forecast → 📋 Reorder Alerts

18 months data → Rolling averages, days of stock → MAPE 12.3% → 65% reduction → $18K saved

The Solution

Built a demand forecasting model using Python Prophet. Analyzed 18 months of historical procurement data. Engineered features including rolling averages, days of stock, and seasonal indicators.

Established cross-functional review cadence between procurement and data teams. Automated daily forecasts that generate prioritized reorder lists.

Results

Emergency procurement: 23 → 8 per month (65% reduction)
Model accuracy: MAPE of 12.3%
Cost savings: $18,000 annually from reduced emergency orders
Process change: Shifted from reactive to proactive purchasing

Key Takeaway

The model's accuracy mattered less than the cross-functional review process. Consistent weekly meetings between procurement and data teams drove the real behavior change.

🔗 View on GitHub 🏠 Back to Portfolio