The Challenge
An e-commerce retailer was struggling with inventory management — frequently running out of stock on high-demand products while over-ordering others. Their existing approach was based on gut feeling and manual spreadsheet estimates, leading to lost sales and excess storage costs.
They needed a data-driven 6-month forecast for their top 3 product lines to plan purchases and negotiate with suppliers more effectively.
My Approach
Data Collection & Cleaning
Gathered 36 months of historical sales data from the client's Excel exports. Cleaned missing months, fixed date formatting, and normalized units.
Exploratory Analysis
Identified strong seasonality patterns (Q4 peaks), a slight upward long-term trend, and one promotional anomaly that needed to be treated separately.
Model Selection
Compared ARIMA, Exponential Smoothing, and Prophet. Prophet performed best on seasonal data with the holiday effect built in, achieving the lowest MAPE score.
Forecast Delivery
Delivered monthly forecast values for 6 months with confidence intervals, trend charts, and an Excel file ready for the client's procurement team.