Demand Forecasting for Shopify, Tuned to Every Product

Stockful fits five candidate models to every SKU at every location: a flat baseline, trend, day-of-week, annual seasonality, and an intermittent-demand model for slow sellers. Each model is backtested against that SKU's own sales history, and a richer model is adopted only when it beats the simple baseline - so a forecast never gets worse by being fancier. Accuracy is then re-measured continuously against actual sales.

It works from day one - Stockful pulls up to 2 years of your existing sales history from Shopify when you install, so forecasts don't start from zero. Products with longer history get richer models: annual seasonality, for example, needs enough history to see the pattern. As history accrues, forecasts deepen automatically.

Yes. Seasonality is learned per product from your own sales history - Christmas lines, summer lines and BFCM spikes each get their own learned pattern. There's no manual 'same period last year' setup, and reorder suggestions take the same season last year into account.

Slow and sporadic sellers get a dedicated intermittent-demand model instead of a noisy average. These are the SKUs where naive forecasting is worst - a couple of random sales can double a simple average - so they're modelled separately.

No. Forecasting is self-tuning: models are selected, backtested and adopted per SKU automatically. Settings are available if you want them - supplier lead time, service level, and lookback period can all be adjusted, including per-SKU overrides.