Summary
Four shops and one question from the owner: where is money being left on the table? The till, stock sheets and a weather feed answered it with three findings, and a fourth figure shows what asking the data looks like day to day.
The station shop works hardest.
The Market shop takes the most over a whole day, but the Station shop takes the most for each hour it is open. It opens only for the commute, so a longer morning there is worth more than a longer day anywhere else.
Two peaks, then a long dip.
People buy before work and at lunch. The afternoon dip is where waste begins: shelves are stocked for a third peak that never comes.
Croissants are baked for the morning and thrown away at night.
One early bake is sized to the morning peak. A second, smaller bake at midday would cover lunch and leave far less in the bin.
| Shop | Dry | Rain | Cold |
|---|---|---|---|
| North | Same | Same | Quieter |
| Market | Busier | Quieter | Same |
| Station | Same | Busier | Busier |
| Harbour | Busier | Quieter | Quieter |
Rain moves people indoors, and to the station.
On wet days the harbour empties and the station fills. Stock can follow the forecast: move the morning delivery a day ahead when rain is due.
- Question
- Which supplier raised the price of butter this season?
- Answer
- The dairy supplier, on its latest invoice. The other suppliers held their prices.
- Sources
- The dairy invoices and the order history it was checked against.
An assistant that shows its sources.
The owner asks in plain words. The assistant answers from the bakery's own invoices, recipes and orders, names the documents it used, and is built to say so when the answer is not in them.
What we would do next.
- Bake croissants in two batches, and watch what is left at closing for a month.
- Put four measures on one weekly dashboard: sales for each hour open, waste at closing, footfall against the forecast, and supplier prices.
- Load the till, stock and weather data into one warehouse every night, so the dashboard is current each morning.
Want a report like this on your own data?