From your own completed orders, per product and per variation. It takes daily sales over a window you set (30 days by default) and turns them into a units-per-day velocity: a simple average while there is under a month of data, a weighted average that favours recent days after that, or exponential smoothing if you choose it. Days left is stock divided by velocity; the 30, 60 and 90-day figures are velocity times days. No AI service, no market data — arithmetic on your own history.

It reports days of stock remaining and expected sales at 30, 60 and 90 days, per product and per variation. It is a projection of the current rate, not a prediction about the market.

More than you think. A product with three weeks of data will produce a number, but the weighted average only kicks in after 30 days of orders, and a seasonal line needs a season before the number means anything.

Lead time per product, on the product edit screen, with a store-wide default (7 days) for everything else. Safety stock is a store-wide percentage of the demand expected during that lead time — 25% by default. Reorder point = velocity × lead time, plus that margin; the suggested order quantity covers the lead time plus 30 days.

Yes, per variation. This matters: a parent product can look healthy while the only size anyone buys is about to run out.

No. There is no external service — the forecast runs on your own installation from your own orders. Some forecasting tools send your order history to their own servers to compute the same numbers; StockPulse reads it in your own database.

A product that is in stock but has not sold in the last 60 days, listed with its retail value so the most expensive shelf comes first. The 60 days is fixed. Slow movers are cash sitting on a shelf, and they are much easier to ignore than a stockout because nothing goes wrong when you do.