A Safety Stock Example You Can Actually Follow

What Safety Stock Actually Means, in Numbers

Safety stock is the extra inventory you hold above your forecasted demand for the lead time window, sized to cover the gap between what you expect to happen and what actually happens. It exists because forecasts are wrong and lead times move, not because more stock is inherently good.

Two things drive the number: how much your demand swings day to day, and how much your lead time swings shipment to shipment. A SKU with steady sales and a reliable supplier needs almost none. A SKU with spiky sales and an unreliable supplier needs a lot, and no formula makes that go away — it just tells you how much to carry instead of guessing.

Every example below uses the same starting SKU so you can see how the answer changes as you add more information, rather than three unrelated formulas that don't connect.

Example 1: The Days-of-Cover Method

This is the method most warehouses start with because it needs no statistics, just a gut number for how many days of trouble you're covering.

The numbers: a SKU sells an average of 40 units a day. Lead time from the supplier is 14 days. Based on past shipments, you decide 4 days of buffer covers the worst delays you've actually seen.

• Safety stock = average daily sales × buffer days = 40 × 4 = 160 units
• Reorder point = safety stock + (average daily sales × lead time) = 160 + (40 × 14) = 720 units

You reorder the moment inventory hits 720 units, and the 160-unit cushion is what's left if that replenishment runs late. It's fast to set up and easy to explain to someone with no forecasting background. Its weakness is that the '4 days' is a guess dressed up as a plan — it doesn't move when your actual variability changes, so it drifts wrong in both directions without anyone noticing.

Example 2: Demand Variability With a Service Level Target

Instead of guessing a buffer, this method uses your actual sales variability and a target service level — the percentage of the time you're willing to not stock out.

Same SKU, more information: average daily demand is still 40 units, but sales actually swing enough that the standard deviation is 10 units a day. Lead time is a reliable, fixed 14 days. You want a 95% service level, which corresponds to a z-score of 1.65.

• Safety stock = z × standard deviation of demand × square root of lead time
• = 1.65 × 10 × √14 = 1.65 × 10 × 3.74 ≈ 62 units

Notice this is a third of the days-of-cover answer. That's the usual pattern: once you measure variability instead of estimating it by feel, the 'safe' number carried from habit is often padded well past what the actual demand risk requires. The 95% target is a business decision, not a law of physics — push it to 98% and the number climbs; drop it to 90% and it falls, because you're explicitly trading a few more stockouts for less cash tied up in stock.

Example 3: Demand and Lead Time Variability Combined

Real supply chains rarely have a fixed lead time. Add lead time variability to the same SKU and the number changes again — usually upward, sometimes sharply.

Same SKU, one more variable: average daily demand 40, standard deviation of demand 10. Average lead time is still 14 days, but it varies too — a standard deviation of 3 days, because the supplier occasionally ships late.

• Safety stock = z × √(lead time × demand variance + demand² × lead time variance)
• = 1.65 × √(14 × 100 + 1,600 × 9)
• = 1.65 × √(1,400 + 14,400)
• = 1.65 × √15,800 ≈ 208 units

Same SKU, same demand pattern, but adding a modest 3-day swing in lead time more than triples the required safety stock compared to demand variability alone (62 to 208 units). This is the number most sellers underestimate, because they track sell-through obsessively and barely track how much their own lead time actually moves shipment to shipment.

The Mistakes That Wreck a Safety Stock Number

The formula is rarely the problem. The inputs are.

• Using one bad quarter as the new normal. A supplier delay during a single disruption gets baked into 'average lead time' forever, and safety stock never comes back down after the disruption clears. We've made this mistake ourselves on a client account — a single congested-port quarter inflated a lead time average for two full reorder cycles before anyone rechecked it.
• Treating every SKU the same. A top-seller and a slow-mover don't carry the same risk, and a flat 'two weeks of stock' rule across a catalog overprotects the slow movers and underprotects the fast ones.
• Ignoring skew. The normal distribution assumes lead times cluster symmetrically around an average. Ocean freight delays don't — they're rarely early by much but sometimes very late, which understates the safety stock a symmetric formula recommends.
• Confusing safety stock with the reorder point. Safety stock is the cushion; the reorder point is the trigger to place an order. Mixing them up means either reordering too late or thinking you're holding double the buffer you actually have.

Full Circle manages more than $500 million in Amazon spend across 100+ brands, and the pattern that shows up account after account isn't a wrong formula — it's a right formula run once, on stale numbers, and never rerun after the supply chain that generated those numbers changed.

When the Number Comes Back Wrong

Sometimes you run the calculation and the answer is obviously bad news — either the safety stock number is far higher than you can afford to hold, or you followed it and still stocked out.

If the number is too high: check whether you're using a capped lead time, not a raw average that still includes the one-off customs delay or factory shutdown. Check whether the service level target was set for the whole catalog when it should be set per SKU tier — a 99% target on a low-margin accessory is usually the wrong trade.

If you followed the number and still stocked out: check the lead time distribution against reality first. A normal-distribution formula run on a lead time that's actually skewed — most shipments on time, a few catastrophically late — will underestimate the tail risk every time. The fix isn't a bigger safety stock across the board; it's modeling the actual shape of your delays, or shortening the lead time itself through supplier changes rather than buffering around it forever.

Where the Calculation Ends and the Work Begins

Running these formulas gets you a number. Keeping that number correct — rechecking the lead time distribution after a disruption, catching the SKU where demand quietly went from stable to spiky, deciding whether a stockout risk is worth the holding cost — is ongoing work, not a one-time calculation.

Dr. Stock, run by Fable 5 out of Full Circle, is built around that ongoing work: reorder timing, aged-inventory and storage-fee decisions, and the true cost of a stockout mid-campaign, with inventory purchasing decisions always going to a human regardless of automation level. It doesn't replace an ERP or a 3PL — it doesn't run your whole business or store your boxes — and if the leak you're chasing is in an ad account rather than a warehouse, that's a job for Dr. PPC. If it's a safety stock number gone stale, that's the part worth having someone else watch.