How To Calculate Safety Stock

Safety stock is the buffer sized so demand does not run out during a lead time. At a ninety five percent service level, a demand standard deviation of eight units and a lead time of six days gives thirty two units, on top of a reorder point of two hundred and forty units made up of average demand during the lead time.

The service level is the probability that demand during the lead time does not exceed the buffer. Ninety five percent is the usual target in retail, ninety nine percent is common where a stockout is expensive or dangerous, and ninety nine point nine is used for spare parts that halt production.

Higher service levels get expensive quickly. The factor is 1.645 at ninety five percent and 2.326 at ninety nine, so going from ninety five to ninety nine raises the buffer by about forty one percent. Raising it to ninety nine point nine multiplies it by nearly two again, which is why the last few percentage points are rarely bought.

The square root of lead time is the part that surprises people. Doubling the lead time does not double the required buffer, it increases it by about forty one percent, because random variation accumulates as a square root rather than in a straight line.

The formula assumes demand is roughly normal and that daily demand figures are independent. Neither is exactly true: demand has seasonality, promotions cause spikes and some products are ordered together. Where those effects are large the calculated buffer is a starting point rather than an answer.

Demand variability is the input that is hardest to get. It has to be the standard deviation of daily demand over a comparable period, not of monthly totals, and it has to be measured on the same product rather than on a category average.

Safety stock is not the same as the reorder point. The reorder point is average demand during the lead time plus the safety stock, and it is the level at which an order is placed. The two are often confused because they are quoted together.

Lead time variability is a separate risk that the basic formula ignores. If a supplier is sometimes early and sometimes very late, the buffer has to cover the late cases as well as the demand variation, and a more elaborate formula is needed.

The cost side is holding cost times the buffer. Every unit of safety stock earns nothing while it sits, so the target service level is usually a commercial decision rather than an operational one, balancing the margin lost to a stockout against the carrying cost of the buffer.

Perishable and seasonal goods cap the buffer differently. A safety stock that would cover a thirty day lead time is useless if the product expires in ten, so the target has to be capped by what can actually be sold.

Finally, review cadence matters more than precision. A buffer recalculated monthly from fresh demand data stays close to right, while one calculated once and left alone drifts as demand and lead times change.