How to Find Safety Stock (With the Formula and a Worked Example)

The Formula

Safety stock is the buffer you hold above forecasted demand to cover the gap between an average sales day and a bad one, for the length of time it takes to get more stock. The standard statistical formula is:

Safety Stock = Z × σd × √L

• Z — the z-score for the service level you're targeting. Higher service level, higher Z, more buffer.
• σd — the standard deviation of your daily demand, not the average. This is the number most sellers skip and just guess.
• L — your lead time in days, measured from the day you place the PO to the day the stock is sellable, not the day it leaves the factory.

Every input here is measurable from data you already have: order history for demand, and your last several purchase orders for lead time. The formula doesn't need a forecast tool. It needs accurate history.

A Worked Example

Take a SKU that averages 40 units a day with a daily demand standard deviation of 12 units, and a supplier lead time of 30 days. You're targeting a 95% service level, which maps to a Z-score of 1.65.

Safety Stock = 1.65 × 12 × √30 = 1.65 × 12 × 5.48 = 108.4, round to 108 units.

That 108 sits on top of average lead-time demand of 1,200 units (40 × 30), which gives you a reorder point of roughly 1,308 units. When your on-hand inventory hits that number, you reorder — safety stock is one input into that trigger, not the trigger itself.

Across the $500M+ in managed revenue Full Circle has managed for 100+ brands, the single most common way this number gets wrong isn't the math. It's the lead time input: sellers plug in the contracted lead time instead of the lead time they actually experienced on the last five or six POs, which is usually longer and always more variable.

No Demand History? Use the Max-Min Method

If you don't have enough sales history to calculate a real standard deviation — a new launch, a recent stockout that wiped out clean data — use the cruder version:

Safety Stock = (Max daily demand − Average daily demand) × Lead time

Same SKU, using a peak day of 70 units against the 40-unit average and the same 30-day lead time: (70 − 40) × 30 = 900 units. That's more than eight times the statistical answer, because it treats your single worst day as the baseline instead of a rare event. Use this method as a temporary placeholder while you accumulate enough clean demand data to switch to the standard formula — not as a permanent policy.

The Mistakes That Break This Number

The formula is simple. The inputs are where it goes wrong, repeatedly, in predictable ways:

• Using contracted lead time instead of actual lead time. Suppliers quote 21 days and deliver in 35. Pull your real numbers from the last several POs.
• Ignoring lead time variability entirely. The formula above assumes lead time is fixed and only demand varies. If your lead time itself swings — a supplier that's sometimes 25 days and sometimes 45 — you need the fuller version that accounts for lead time variance too, or you'll consistently under-buffer.
• One service level for every SKU. A hero product during a launch and a long-tail accessory don't deserve the same 99% target. That's how cash gets trapped in slow movers.
• Treating a promo spike as normal variability. We've made this mistake ourselves: a demand spike from a coupon push got folded into a SKU's standard deviation and inflated its safety stock for months after the promotion ended. Strip promotional weeks out of the calculation, or flag them separately.
• Confusing safety stock with reorder point. Safety stock is the buffer. Reorder point is average lead-time demand plus that buffer. They answer different questions.

When the Number Looks Wrong

If your safety stock comes out larger than your average lead-time demand, something upstream is off before you trust the output. Check three things in order:

• Outliers in the demand data. One viral day or one stockout-driven zero-sales week can distort a standard deviation badly. Pull them out and recalculate.
• The lead time you actually used. Confirm it's PO-to-sellable, not PO-to-shipped or the number on the supplier's quote sheet.
• The service level target itself. 99% on a slow-moving SKU is often the wrong goal, not a math error.

If you fix all three and you're still stocking out, the problem usually isn't the safety stock formula — it's that the reorder trigger isn't being acted on in time, or lead time has genuinely lengthened and nobody updated the input. Recalculate whenever a lead time shifts by more than a few days, and revisit every SKU's inputs at least quarterly.

Where This Fits Into the Bigger Problem

The formula is the easy part. Doing it correctly at real Amazon scale means pulling actual lead time off your last twenty purchase orders instead of the contract, catching the SKU whose demand history is still skewed by a promotion three months ago, and recalculating the moment a supplier's lead time drifts — for every SKU, continuously, not once a quarter in a spreadsheet. Dr. Stock, the Amazon inventory and supply chain product from Fable 5, runs that recalculation on an ongoing basis alongside the reorder timing and storage-fee tradeoffs that sit next to it, with inventory purchasing decisions always routed to a human for approval. If the leak you're chasing is in the ad account rather than the warehouse, that's a question for Dr. PPC, not this one.