Safety stock: definition, formula and how much to hold

Safety stock is the part of your inventory you do not expect to need before the next delivery arrives. It is a target level, not a set of units held back: every unit sells normally, and the buffer is what gets drawn down when demand runs hot or a shipment runs late. Amazon's seller guidance calls the same idea buffer stock.

How do you calculate safety stock?

Start from the reorder point: reorder when available stock, meaning on hand plus inbound, falls to forecast demand during the lead time plus safety stock. Oracle's planning documentation and Amazon's inventory guide both use this structure.

For the safety stock term itself, the textbook formula with a fixed lead time is:

Safety stock = Z × σD × √L

• Z is the service level factor, set by how often you want to get through a replenishment cycle without running out.
• σD is the standard deviation of daily demand.
• L is the lead time in days.

When lead time also varies, add its standard deviation (σL) and average daily demand (D):

Safety stock = Z × √(L × σD² + D² × σL²)

Oracle Inventory also offers a version built on forecast error: Z × 1.25 × the mean absolute deviation of past forecasts from actual demand. An accurate forecast earns a small buffer.

The Z-based formulas assume each day's demand is an independent, normally distributed draw. Real demand cannot go below zero, so when the standard deviation is large next to the average, as with slow or lumpy SKUs, the formula tends to overstate the buffer. Demand that runs in streaks breaks the independence assumption too. For those SKUs, check the result against actual lead-time demand in your sales history.

What does a worked example look like?

A SKU sells 40 units a day with a standard deviation of 12. Total lead time, from purchase order to units available at FBA, is 49 days, so lead-time demand is 40 × 49 = 1,960 units.

Service level	Z	Safety stock
90%	1.28	108 units
95%	1.65	139 units
98%	2.05	172 units
99%	2.33	196 units

At 95%, the reorder point is 1,960 + 139 = 2,099 units. Going from 95% to 99% adds about 41% more safety stock, which is why not every SKU deserves the top service level.

Now let the lead time swing with a standard deviation of 7 days. Same SKU, same 95%: safety stock rises to about 482 units, roughly 3.5 times the fixed-lead-time figure. Here, lead-time reliability moves the buffer far more than demand noise does.

Why does safety stock matter for operations and growth?

Too little and best sellers run out; the stockouts page covers what that costs on Amazon and Walmart. Too much and cash sits on shelves. Amazon treats more than 90 days of supply as a sign of excess inventory and suggests keeping about two months of supply to avoid storage fees.

Amazon also charges for the thin end. In the US, eligible products pay a low-inventory-level fee on each shipped unit when both the 30-day and 90-day historical days of supply, calculated per FNSKU, are under 28 days. New Professional sellers for 365 days after their first inventory is received, products that sold fewer than 20 units in the past 7 days, and SKUs with 70% or more of units auto-replenished through AWD over the prior 90 days are among the exemptions. So where the buffer sits matters as much as its size. A deep buffer in AWD, for eligible products, or in a 3PL covers long ocean lead times without using FBA capacity, while FBA holds enough days of cover to sell until the next transfer lands. With AWD auto-replenishment on, Amazon manages the FBA top-ups itself. A 3PL buffer has to keep FBA above the 28-day line on your own schedule.

What are the common safety stock mistakes?

• One rule for every SKU. A flat two weeks or a fixed percentage ignores how noisy each SKU really is.
• Stacking worst cases. The shortcut of maximum daily sales × maximum lead time, minus average × average, assumes the worst demand and the worst delay hit together, so it can oversize the buffer.
• Using the supplier's quoted lead time. The real lead time includes production, freight, customs, prep and FBA receiving, and its variability can dominate, as the example shows.
• Adding up channel buffers. Unless Amazon, Walmart and Shopify demand move in lockstep, one pooled buffer for shared stock needs fewer units than separate channel buffers added together.
• Setting it once. Velocity, seasonality and suppliers change, and the buffer should change with them.

How does Growth OS set safety stock?

Replenishment in Ecomsellertool Growth OS runs on your own accounts and is built to size safety stock per SKU from your data:

1. Demand spread per SKU and channel from your sales history, pooled where channels share stock.
2. Lead time mean and spread from your actual PO and receipt dates, per supplier and route, rather than a supplier's quote.
3. A target service level you choose per SKU group, so best sellers can carry more cover than the long tail.
4. Placement: the deep buffer upstream in AWD or a 3PL, with FBA kept above the 28-day low-inventory line on SKUs where the fee applies, and below 90 days of supply.
5. Recalculation as new sales and receipts land.

The results feed reorder points, draft POs and shipment plans. Custom modules we build and your data are yours; the Growth OS base is licensed to you.

Where should you start?

If best sellers keep running out while slow SKUs pile up, start with the free Quick Scan in the Ops Gap Diagnostic. It works from standard Amazon reports you upload and returns your top three operations gaps. To see the base system first, look at Growth OS.