Restock: next week's order for small shops, from the sales file they already have

Description and use

Restock uses TabPFN to forecast next week's demand for small shops, calculates a safe order amount, and produces a supplier message.

Problem

Small shops run out of best sellers on busy days, losing sales because owners guess orders.

Outcome

On sample bakery data, TabPFN was 24% more accurate than the simple "same as last week" rule, reducing forecast error from 5.7 to 4.3 units per item per day.

Workflow

• Upload CSV of past sales
• Restock cleans data (date, item, units, stock)
• TabPFN forecasts next 7 days per item
• Calculate safe amount (busy-day forecast sum)
• Compute order = max(0, safe amount – in stock)
• Display order message and forecast table

Evidence

Restock takes the sales file a shop already has, from a POS or Excel, and gives back: how much of each item it will likely sell next week

Source

Before showing a forecast, Restock tests itself on your own sales. It hides your last 7 days, predicts them, then checks how close it got, and compares that with the simple rule most owners use

Source

On the sample bakery data, TabPFN was off by 4.3 units per item per day, against 5.7 for "same as last week" (24% better).

Source