Safety Stock
Safety stock is inventory held above expected demand during replenishment to absorb variability in usage and lead time. It buys coverage when a forecast runs high or a delivery runs late, converting a would-be stockout into a drawdown that replenishment quietly refills.
Safety stock is the quantity of an item deliberately held beyond what average demand over the replenishment lead time requires. It exists because neither demand nor supply is a single number. Usage varies week to week, suppliers deliver early and late, yields differ from standard, and a plan built on averages will be short roughly half the time. Safety stock is the inventory that absorbs that spread so a normal variation does not become a line stoppage.
Mechanically it enters planning in two places. In a reorder point system it is added to lead time demand, so the trigger level sits above expected consumption and the buffer is what remains when a delivery is late. In material requirements planning it is treated as stock that must not be consumed, so netting adds it back as a requirement and the plan replenishes down to it rather than to zero. Two sizing approaches dominate. The max-minus-average method takes the worst observed usage over the worst observed lead time and subtracts the average case, which is simple and tends to be generous. The statistical method multiplies the standard deviation of demand over lead time by a service factor drawn from the chosen service level, which costs you a demand history but prices the buffer against a stated risk.
Safety stock is not cycle stock and not the reorder point. Cycle stock is the working inventory consumed between deliveries; safety stock is the part of the balance you hope never to touch. The reorder point is a trigger level that contains safety stock as one of its two terms. Safety time is the alternative lever: rather than holding extra units, the planner pulls the need date forward by a few days and holds extra time instead.
The error shops make is setting a level once and never revisiting it, so a buffer sized around a supplier who has since halved its lead time quietly carries obsolete stock for years. The opposite error is using safety stock to paper over a cause rather than a variance. A chronically late supplier, a forecast nobody owns, or a bill of materials with the wrong quantity per is a problem to fix, and padding the buffer hides it while the carrying cost compounds. Safety stock also protects against variability, not against a systematic shortfall; if usage has permanently stepped up, the buffer drains once and does not come back.
On a real shop floor the buffer is unevenly applied by design: deep on the cheap long-lead fastener nobody wants to chase, thin or absent on the expensive casting bought to order, and reviewed each quarter against what actually happened.
Safety Stock = (Maximum Daily Usage x Maximum Lead Time) - (Average Daily Usage x Average Lead Time); the statistical form is Z x standard deviation of demand over lead time.
A contract electronics assembler consumes 1,200 of a connector per week on average, with a worst week of 1,800. The supplier quotes three weeks and has taken five. Max-minus-average gives (1,800 x 5) minus (1,200 x 3), or 9,000 minus 3,600, which is 5,400 pieces. At 0.40 dollars each that buffer ties up 2,160 dollars. The statistical route, using a weekly demand deviation of 250 pieces over three weeks and a 95 percent service factor of 1.65, asks for about 715 pieces instead.
What is the difference between safety stock and reorder point?
Safety stock is a quantity; the reorder point is a trigger level that includes it. The reorder point equals expected demand over the lead time plus safety stock, so if you use 20 a day over a 5-day lead time and hold 40 of buffer, you reorder at 140. Raising safety stock raises the reorder point one for one.
Do we need safety stock on every part?
No, and spreading it evenly is the expensive mistake. Buffer the items where a stockout is costly and replenishment is slow or unreliable: long-lead purchased parts, shared components used across many parents, and anything that gates a line. Skip it on parts bought to order for a single job, on expensive items with a dependable supplier, and on anything short-shelf-life.
How do you choose a service level for safety stock?
Work backwards from what a stockout costs. A 95 percent service level means roughly one replenishment cycle in twenty runs short, and the service factor for it is about 1.65; 99 percent needs about 2.33, which is roughly forty percent more inventory for the last four points of coverage. Parts that idle a line justify the higher number.
What problems does safety stock fail to solve?
It buffers variability, not bias. If your forecast is consistently low or a supplier is consistently late, the buffer drains and stays drained, because nothing refills it faster than it empties. It also cannot help against a wrong bill of materials quantity, an inaccurate stock balance, or scrap nobody posted, since the plan never knew the part was gone.
How is safety stock maintained in an inventory system?
It is a field on the item record, so planning reads it on every run rather than a planner remembering it. Reorder point logic adds it to lead time demand; MRP netting adds it back as a requirement so the plan replenishes down to the buffer rather than to zero. Review the stored values on a schedule, not when a shortage prompts it.