The challenge
An operations team was ordering stock largely on instinct and last month's numbers. The result was the worst of both worlds: popular items ran out and lost sales, while slow movers piled up and tied up cash. Nobody had a forward-looking view of demand.
The approach
- Pulled together sales history across products and locations into one clean dataset.
- Built a demand forecast using statistical and machine-learning methods, accounting for seasonality and trend.
- Translated forecasts into reorder signals — clear guidance on what to order and when, per product.
- Surfaced it in Power BI so planners saw projected demand next to current stock at a glance.
What I built
A weekly demand forecast feeding an inventory dashboard that flags items heading for a stock-out and items sitting as dead capital. Planners moved from reacting to last month to planning for next month.
Stock-outs on key items fell by around 18%, while excess inventory dropped roughly 12% — recovering working capital that had been sitting idle on shelves.
The outcome
Ordering shifted from gut feel to evidence. The business held less of what didn't sell, more of what did, and freed cash that could be put to better use elsewhere.