Safety Stock Management: How to Calculate It (With a Worked Example)

What safety stock actually is

Cycle stock is the inventory you expect to sell between one reorder and the next — a straightforward product of average demand and order frequency. Safety stock is the buffer sitting on top of that, sized to cover the days when demand comes in higher than average, or the shipment arrives later than promised.

If demand and lead time were perfectly predictable, you wouldn't need any. They aren't, so the question isn't whether to hold safety stock — it's how much, and that number comes from a calculation, not a habit. A common failure mode is picking a round number ('two weeks of cover') because it feels safe, then never checking whether two weeks was ever the right answer for that specific SKU.

How to calculate it — a worked example

The most commonly used version accounts for both lead time variability and demand variability: Safety Stock = Z-score × standard deviation of lead time (σLT) × average daily demand (D avg). Each piece matters:

• Z-score — the service level you're targeting, expressed as a probability of not stocking out. 95% service level = Z of 1.65.
• σLT — how many days your actual lead time swings from its average, based on real delivery history, not the number on the supplier's quote sheet.
• D avg — average units sold per day over a representative period, ideally excluding one-off spikes.

Take a SKU selling 18 units a day on average, with an average lead time of 35 days and an observed lead time deviation of 6 days, targeting a 95% service level: Safety Stock = 1.65 × 6 × 18 = 178 units. The reorder point — the stock level that triggers a new order — is average demand times lead time, plus safety stock: (18 × 35) + 178 = 808 units. That's the number that should trigger the PO, not a calendar date.

When the number is wrong

Two signals tell you the safety stock figure is wrong, and they point in opposite directions. A stockout while safety stock is supposedly in place means the inputs were stale — usually the lead time deviation was smaller than reality. Rising storage costs and units that haven't moved in months mean the opposite — the number was set too high, or demand fell and nobody rebuilt the calculation.

The single most common input error is using the supplier's quoted lead time instead of the observed one. A spec sheet says 30 days; ocean freight and customs actually average 38 with a 9-day swing in Q4. Run the formula on the quoted number and the safety stock is short before the first replenishment order even lands — we've watched this happen on our own managed accounts, not just other people's.

Across the $500M-plus in managed revenue we've managed for 100-plus brands, this is the pattern that shows up over and over: the formula was fine. The number fed into it wasn't.

Setting the service level by SKU, not by catalog

Not every SKU deserves the same Z-score. A hero product driving most of the revenue justifies a higher service level and more capital tied up in buffer stock. A low-margin, easily substituted item doesn't — every extra unit of safety stock there is cash sitting still for no good reason.

The table below shows how the target service level, the Z-score it maps to, and the resulting stockout risk change together. Move a SKU from 95% to 99% and the required safety stock climbs sharply for a fairly small reduction in stockout risk — worth doing for a hero SKU, rarely worth it for a slow mover.

Where this breaks down specifically on Amazon

The formula doesn't change on Amazon. The cost of getting it wrong does. A stockout doesn't just lose the sale — it can drop organic rank mid-campaign, and if ads keep running against a listing that's out of stock, that's spend going nowhere. That specific leak — money burning in the ad account while the listing is unavailable — is a job for Dr. PPC, not an inventory calculation.

On the inventory side, the failure modes are different: long-term storage fees and aged-inventory surcharges when safety stock runs high, and a removal-versus-liquidation decision nobody wants to make when it's too late to sell through. Both are downstream of the same root cause — a safety stock or reorder point number that was calculated once and never rechecked against what actually happened.

Who should be doing this by hand, and who shouldn't

A single SKU with stable demand and a reliable supplier is a spreadsheet problem. A catalog of a few hundred SKUs with real seasonality and shifting freight lead times is a software problem — inventory management systems and ERPs exist to run this calculation continuously across every SKU, and they do that job well; nothing here replaces one.

Dr. Stock, from Full Circle, is built for the case in between: a catalog large enough that nobody's rechecking lead time deviations weekly, where stockouts, aged inventory, and FBA fee errors are leaking money that can be checked, not guessed at. It runs inventory and finance tracking (Orbit is included, covering ASIN profitability, BSR, buy box, price and fee movement) with the client choosing how much autonomy to hand over — inventory purchasing decisions always go to a human. There's no published price; it's quoted on a call, with a first-30-days-free demo. A reader who never buys anything should still leave this page able to run the formula, spot when the number is wrong, and know which team — inventory or ads — owns the fix.