🏪 Ranchi Dukaan Pro — Predicting Stock Demand with TabPFN-TS

Description and use

Ranchi Dukaan Pro uses TabPFN-TS to forecast seven‑day demand and generate restock recommendations for small retailers, showing better MAE and WAPE than a simple baseline on synthetic data, though real‑world validation is pending.

Problem

Small retailers need to decide what to reorder, how much stock to keep, and when to place the next order, but lack tools to forecast demand and generate inventory recommendations.

Outcome

In a synthetic test with three products, TabPFN-TS achieved MAE 1.29 and WAPE 9.18%, outperforming a simple weekday‑one‑week‑earlier baseline (MAE 1.62, WAPE 11.56%). The model did not beat the baseline for Bread.

Workflow

• Enter daily sales history for each product
• Use TabPFN-TS to generate a seven‑day demand forecast
• Check expected demand during the supplier’s lead time
• Combine that estimate with current stock, incoming stock, and safety stock
• Display a suggested reorder quantity and its estimated cost

Evidence

I generated synthetic daily sales data for three example products: Milk 500 ml, Bread, and Rice 1 kg.

Source

The results from this single test window were: Mean Absolute Error (MAE) 1.29 units for TabPFN-TS vs 1.62 units for baseline.

Source

Important limitation: these results come from synthetic data and one seven-day holdout.

Source