Safety stock, explained: how much buffer inventory you actually need

Why a flat reorder point runs out

A reorder point built on a single number — average daily sales times lead time — works exactly as well as that average holds true. Most weeks, demand doesn't sit at the average; it moves above and below it, sometimes by a wide margin, and a reorder point sized only for the typical week has nothing left over for the weeks that aren't typical. Lead time isn't fixed either: a supplier that usually ships on schedule can run late without anyone doing anything wrong, and a plan built for the average lead time runs out during the extra days a late shipment adds. The result is a stockout that looks like bad luck but is really just variance the plan never accounted for. A flat reorder point isn't wrong on average — it's wrong on the weeks that matter, which are exactly the weeks an empty shelf or a sold-out listing costs a sale. That's true whether the reorder point was set by hand or generated by a spreadsheet formula — the number itself isn't the problem, the absence of any room for deviation is.

What safety stock actually absorbs

Safety stock is the extra units held above what average demand and average lead time would require — a deliberate buffer, not a rounding error. It exists to absorb two separate sources of uncertainty: demand variability, which is how much sales in a given period swing above or below the forecast, and lead-time variability, which is how much a supplier's actual delivery time swings around its average. Either one alone can eat into a plan that has no cushion; together, a demand spike that lands during a slow supplier shipment is the scenario a flat reorder point is least prepared for. How much cushion to hold is a choice, not a fact — it's set by the service level you're targeting, the probability you're willing to accept of not stocking out before the next delivery. A higher service level means carrying more buffer for the same product; accepting a lower one frees up capital but accepts more stockout risk. There's no single right answer, only a deliberate tradeoff instead of an accidental one. Neither source of variance is a flaw to eliminate — demand moves because customers do, and lead times move because suppliers, carriers, and customs each add their own noise. Safety stock is the acknowledgment that some of that noise will show up in the numbers no matter how good the forecast is.

The inputs that size it

Sizing a safety stock number starts with the same three inputs, whatever formula or tool ends up doing the arithmetic. Demand variability — how much a product's sales swing from period to period, not just its average rate — sets how much of a demand spike the buffer has to absorb. Lead time, and specifically how much that lead time varies rather than just its average length, sets how long a shortfall can run before a new order arrives to close it; a supplier with an erratic ship time demands more buffer than one who is slow but consistent. And the target service level converts both of those into an actual unit count, translating how much risk you're willing to carry into how many extra units that means. Change any one of the three and the number moves: a fast-moving, volatile product from an inconsistent supplier needs meaningfully more buffer than a steady seller from a reliable one, even if their average sales look identical on a report. None of the three stays fixed for long, either — a product can go from steady to volatile after demand shifts, and a supplier's lead time can drift as its own backlog changes — so a buffer that was right last quarter isn't guaranteed to still be right this one.

From a buffer to a reorder decision

None of that has to be a spreadsheet you maintain by hand. Assisty forecasts what each product — and each variant — will sell over the horizon you choose, and flags which items are at risk of running out before that period ends; those at-risk items carry straight into the Smart Buying Table, so spotting the risk and acting on it happen in the same motion. Inside the buying table, every suggested order quantity is computed from sales velocity, lead time, and incoming stock already on an open purchase order, then rounded to your case sizes and capped at your maximum stock level — so the risk you were reasoning about in the abstract becomes a concrete number of units to order, one you can check and adjust before turning it into a purchase order. And because that forecast recomputes when you ask a what-if question about demand trending up, the quantity it suggests tracks the demand you're actually planning for, not a static average. It's also not a black box: you can ask why a forecast is moving and get the drivers back in plain language, so the number you're ordering to is a response to something you can see and question, not one you just have to trust.