Most stores handle restocking with a spreadsheet, last month's numbers, and a feeling. That works until it doesn't — overstock ties up cash, and understock costs the sale and sometimes the customer along with it. Demand forecasting replaces the feeling with four numbers, and each one answers a different question.
Reorder point: when to order
This is the stock level that should trigger a new order. It's not "zero" — by the time a product hits zero, the order is already late, because your supplier's lead time still has to pass before more stock arrives. The reorder point is set far enough above zero that an order placed the moment stock crosses it still arrives before the shelf actually empties.
Safety stock: the buffer
Safety stock is the cushion built into that reorder point to cover a busier-than-usual stretch — a product that sells at double its normal rate for a week shouldn't be able to blow through your stock before the reorder even triggers. How conservative that buffer is is something you set, not something fixed for every store; a business that can't tolerate stockouts sets it wider than one that can absorb an occasional short wait.
Order size: how much to buy
Order size balances two costs against each other: the cost of placing an order (which favors ordering less often, in bigger batches) against the cost of holding stock (which favors ordering more often, in smaller batches). You provide both costs, and the batch size that balances them is what gets suggested — it's a calculation, not a guess.
Days left: the plain-English version
Days left translates all of the above into the question a store owner actually asks out loud: how long does current stock last at the forecast rate. It's the number worth glancing at daily; the other three are what it's built from.
One honest caveat
A forecast is only as good as the history behind it. A product with a thin or noisy sales record gets a plainer, trend-based estimate instead of a precise-looking figure, and every view says which method produced it and how reliable it's been historically. A confident-looking number built on three weeks of sales data is worse than an honest one that says "not enough history yet" — so that's the one it shows.