Safety Stock & Reorder Point

Which formula this uses, and why

There are two safety stock formulas in common use, and the difference between them is not academic.

The simple one, which assumes lead time never varies:

The full one, which accounts for lead time varying too:

This calculator uses the full version. The second term under the root — d² × σLT² — is the part the simple formula throws away, and in Indian supply chains it is frequently the larger of the two. A part with steady demand but a lead time that swings between 14 and 35 days needs a buffer driven almost entirely by the lead time, not by the demand.

A worked example

Using the pre-filled figures — a fast-moving part at 120 units a day, standard deviation 34, lead time 21 days with a deviation of 5 days, at 95% service:

• Demand during lead time = 120 × 21 = 2,520 units
• Demand-variability term = 21 × 34² = 24,276
• Lead-time-variability term = 120² × 5² = 360,000
• Safety stock = 1.645 × √384,276 = 1,020 units
• Reorder point = 2,520 + 1,020 = 3,540 units

Notice that the lead-time term is roughly fifteen times the demand term. Using the simple formula here would give a safety stock of about 256 units — a quarter of what this part actually needs. That gap is what a stockout looks like three months later, and it is the single most common cause of “but the system said we had enough”.

Common mistakes

• Using the supplier's quoted lead time. Quoted is not actual. Measure from PO release to stock available for picking, using your own GRN dates. The gap between quoted and actual is usually where the variability lives.
• Calculating standard deviation on monthly data, then using daily demand. The units must match. If you have monthly figures, either convert everything to months or divide carefully — mixing the two is the most frequent arithmetic error in inventory planning.
• Applying one service level to the whole catalogue. A 99% target on C-class items ties up enormous capital for almost no service benefit. Classify first, then set service levels by class — the ABC–XYZ classifier does that in one pass.
• Never recalculating. Safety stock set in 2023 against 2023's lead times is a liability now. Review quarterly, and immediately after any supplier or transport-lane change.
• Including one-off bulk orders in the deviation. A single dealer stocking order can double your measured standard deviation, and you then carry that inflated buffer for a year. Strip the outliers first, and handle them as project demand instead.

Questions people ask

What service level should I choose?

Work backwards from the cost of a stockout. For a part that stops a customer's machine, 98–99% is easy to justify. For a slow-moving consumable with a substitute on the shelf, 90% is generous. The compare table above shows exactly what each level costs you per year for this item — that number usually settles the argument faster than any policy document.

Does this work for intermittent demand?

Not well. Items that sell on only a handful of days a month break the normal-distribution assumption behind the Z factor. For those, a Croston-style method or a simple min-max based on observed maximum demand in lead time is more honest than a formula that looks precise but isn't. Most spare-parts catalogues have a long tail of these — treat them separately rather than pretending the formula covers them.

Why is my reorder point higher than my maximum stock?

Because your order quantity is too small for the lead time. If the reorder point exceeds the max level, you will be raising a new PO before the last one arrives — which is fine, but it means you need multiple open orders in the pipeline and your system must handle that. If it doesn't, increase the order quantity (see the EOQ calculator) or shorten the lead time.