You can’t just “guess” a safety buffer and call it a policy — pick a number too low and you’re still stocking out; pick one too high and you’re quietly paying to store inventory you’ll rarely touch. The formula exists to turn “how much buffer” into a number grounded in how variable your actual demand is and how long your actual lead time runs.
The Formula
Safety Stock = Z × σ(D) × √(Lead Time)Three inputs. Here’s what each one means.
Z: The Service Level Target
Z (the Z-score) is a statistical stand-in for your target service level — the probability you’re comfortable with of not stocking out during a lead time window. You don’t need to compute a Z-score yourself; it’s a standard, looked-up value for a given service level:
| Service Level | Z-Score | Stockout Risk |
|---|---|---|
| 90% | 1.28 | 10% |
| 95% | 1.65 | 5% |
| 97% | 1.88 | 3% |
| 99% | 2.33 | 1% |
| 99.9% | 3.09 | 0.1% |
Picking a service level is a business decision, not a math problem: it should reflect how costly a stockout actually is for that specific item, weighed against how much extra it costs to hold the corresponding buffer.
σ(D): Standard Deviation of Demand
This is how much your daily demand actually varies, not the average demand itself. A product that sells a steady 60 units almost every day has a small standard deviation; one that swings between 20 and 100 units has a large one. This number comes from your own historical sales data — How to Calculate Safety Stock covers exactly how to pull it out of what you already have on record.
Lead Time
This is your supplier’s average delivery time, in days. Longer lead times mean more days during which demand can drift away from the average before a new order arrives — which is why lead time appears under the square root: variability compounds with exposure time, but sub-linearly. Double the lead time and the safety stock contribution from it grows by about 41% (√2), not 100%.
An Important Limitation, Stated Plainly
This formula treats lead time as a fixed, known number — it does not include a separate term for lead time variability the way some more advanced safety stock models do. If your supplier’s lead time itself swings around unpredictably (sometimes 10 days, sometimes 20), this formula’s demand-only view of uncertainty will understate how much buffer you actually need. That’s a real, worth-knowing limitation of this exact formula, not a flaw in the concept of safety stock generally — Safety Stock Reality Check covers what to do when lead time variability, not just demand variability, is the bigger risk in your situation.
Why Service Level and Safety Stock Aren’t Proportional
Look at the Z-score table again: going from 90% to 95% service level moves Z from 1.28 to 1.65 — a jump of 0.37. Going from 95% to 99% moves it from 1.65 to 2.33 — a jump of 0.68, almost double the previous increment, for a smaller absolute improvement in service level. Since safety stock scales directly with Z, that means each additional point of protection costs more than the last. Chasing 99.9% “just to be safe” across your whole catalog usually means paying a steep, accelerating price for protection most of your items don’t need.
Common Misconceptions
“More safety stock always means better service.” True only up to the point where the extra buffer is actually protecting against real variability. Past that, you’re just paying carrying cost with no matching reduction in stockout risk — the formula already accounts for your variability; adding more on top of what it recommends is a guess, not an optimization.
“One service level fits the whole catalog.” A stockout on a critical, hard-to-substitute item costs far more than one on a commodity product with three alternatives on the shelf next to it. Applying a blanket 99% target everywhere overspends on the items that don’t need it.
“The formula accounts for everything that can go wrong.” It accounts for demand variability, assuming lead time is fixed. It’s a genuinely useful starting point, not a complete risk model — see the limitation above.
Where This Goes Next
Once the formula’s logic makes sense, the real work is pulling trustworthy demand-variability and lead-time numbers from your own operation — that’s How to Calculate Safety Stock: Step-by-Step. Or plug your own numbers straight into the Safety Stock Calculator and see the buffer size respond in real time.