A mid-sized bakery network automated daily delivery planning for thousands of store-product combinations - replacing a manual, error-prone process with a statistical forecasting model available as a simple web app.
Problem
The bakery operates a network of several dozen retail points and delivers fresh products daily - including bread, rolls, and pastries with very short shelf life.
Every working day, someone had to decide exactly how much of each product to deliver to each store for the next morning.
That "someone" was actually three people, working full time, scrolling through Excel spreadsheets and manually entering delivery quantities for roughly 4,000 contractor-product combinations. Every day.
The cost wasn't just payroll. With that many manual decisions, errors were inevitable - underestimates meant stockouts at the client, overestimates meant returns and waste. Different people applied different gut-feel heuristics, so results were inconsistent. And if the business added new clients or products, the only answer was hiring more people to scroll through more spreadsheets.
Challenge
The core difficulty was building a model that could reliably forecast next-day delivery quantities across thousands of combinations with very different demand patterns. The solution had to be accurate enough to trust, simple enough to use, and fast enough to run daily without disrupting existing workflows.
Solution
We developed a statistical forecasting model that automatically generates next-day delivery recommendations for every client-product pair.
The model is deployed as a web API with a clean browser interface. A user uploads the current Excel file with today's data and gets it back with the delivery tomorrow column filled in - ready to act on.
Performance is validated against historical data using standard forecasting metrics.
Results
| Metric | Impact |
|---|---|
| Manual work replaced | 3 full-time roles automated |
| Decisions per day | ~4,000 client-product pairs, handled in seconds |
| Consistency | Same logic applied across all combinations - no more gut-feel variance |
| Availability | 24/7 via web browser, no Excel gymnastics required |
| Validation | Backtested on historical data with standard forecasting metrics |
Let's talk about the process your team is still doing in Excel.
