Safety stock is an inventory buffer for uncertainty in demand or replenishment. It sits above the expected requirement for the policy's coverage period. It can reduce shortage risk, but it cannot guarantee availability or compensate for every supplier disruption.
Safety stock formula and assumptions
For independent daily demand with stable variance and a fixed lead time, a normal approximation gives:
Safety stock = z × σd × √L
Here σd is the standard deviation of daily demand, L is lead time in days, and z corresponds to a selected cycle service level. Use consistent time units. The coverage period changes for a periodic-review policy; do not silently substitute lead time for the review interval plus lead time.
Cycle service level is the probability of completing a replenishment cycle without a stockout under the model. It is different from the share of demanded units fulfilled. MIT's probabilistic inventory lecture explains this distinction and the difference between continuous and periodic review.
| Selected cycle service target | Approximate z-score |
|---|---|
| 90% | 1.28 |
| 95% | 1.65 |
| 97.5% | 1.96 |
| 99% | 2.33 |
These are model inputs, not recommended targets for every business. A 95% target does not mean exactly one stockout will occur in the next 20 cycles.
Worked coffee-shop example
Assume a fictional shop uses an average of 18 lb of beans per day, daily-demand standard deviation is 4 lb, and replenishment takes a fixed three days. The buyer selects a 95% cycle service target for this example.
Expected lead-time demand = 18 × 3 = 54 lb
Safety stock = 1.65 × 4 × √3 ≈ 11.4 lb
Reorder point = 54 + 11.4 ≈ 65.4 lb
A whole-pound threshold could be rounded up to 66 lb. That is a reorder trigger, not the quantity of the next purchase. A separate order policy determines how much to buy, with incoming orders, commitments, supplier packs and capacity considered.
Use the safety stock calculator to inspect the calculation, then read the reorder-point example for the inventory-position comparison.
Choose the service target from the consequences
Consider what happens when the item is missing: a lost sale, a production stop, an acceptable substitute or a short customer wait. Compare that consequence with storage, cash, expiry and handling costs. A cheap but essential component may warrant more protection than an expensive item with a ready substitute.
ABC analysis can prioritize review by annual usage value. It does not determine a universal z-score for each tier. Slow sales alone do not justify a low service target, and volatile demand alone does not establish the correct buffer.
Review the target against actual results. Keep the definition of a stockout and the measurement period consistent when comparing policy changes.
When the simple formula is not enough
- Variable lead time: include supply uncertainty using a suitable model; a fixed-lead-time formula omits it.
- Sparse or lumpy demand: a normal approximation may describe the demand distribution poorly. A point forecast alone does not determine shortage probability.
- Stockout-censored sales: recorded sales may understate demand when the item was unavailable.
- Promotions or bookings: known future events may need an explicit adjustment rather than a larger historical standard deviation.
- Perishability: additional stock can become unusable before it is needed. Track waste without counting the same expected loss twice.
Zero estimated variation from too little history is not proof that a buffer is unnecessary. Record the fallback assumption and the person responsible for reviewing it.
Safety stock, PAR, JIT and VMI
PAR is an order-up-to target; safety stock can be one component. JIT is an approach to coordinated flow and replenishment, not simply a safety-stock setting of zero. VMI changes replenishment responsibilities; the parties can still agree and inspect stock limits and service measures.
These concepts answer different questions. A supplier-managed item can have a buffer, and a buyer-managed item can use frequent small replenishments.
Apply the calculation to purchasing
For a standalone estimate, start with the calculator. For a recurring operating problem, evaluate LineNow's retail replenishment workflow using one SKU, its demand history and an outstanding PO.
Inspect the selected sensitivity, calculation horizon, configured buffer and insufficient-data fallback. Confirm how incoming stock affects the recommendation and where the buyer reviews the supplier order. The educational lead-time formula above does not establish that every product setting uses the same horizon or service interpretation.
Review assumptions when demand, supplier timing, pack sizes or shortage consequences change. Supplier confirmations and physical receipts in the living PO provide evidence for that review; they do not make an uncertain forecast certain.