What is it?
A predictive reorder tool built to solve a real operations problem in the food industry.
Context
A case study on building a predictive reorder tool from scratch to solve a real operations problem in the food industry.
The Problem
Running a restaurant means managing hundreds of ingredients across multiple suppliers, each with different lead times, units, and minimum order quantities. The standard approach? Spreadsheets, gut feeling, and the occasional panic run to the store.
I lived this problem firsthand. Every week I'd spend hours cross-referencing what we had on the shelves with what we'd sold, trying to guess how much to order. Despite the effort, stockouts still happened. A missing ingredient during a Friday dinner rush is the kind of failure that compounds: lost revenue, wasted prep time, frustrated staff, unhappy guests.
Problem & Impact
| Problem | Impact |
|---|---|
| Manual order building | 3–4 hours/week spent comparing stock vs. sales by hand |
| No demand signal | Orders based purely on experience, not actual consumption data |
| Stockouts | Running out of key ingredients mid-service |
| Supplier fragmentation | Items split across 3+ suppliers with different catalogs, units and prices |
| No cost visibility | Couldn't easily see per-plate ingredient costs or spot price changes |