Safety Stock Formula: The Calculation, a Worked Example, and the Reorder Point It Feeds

The safety stock formula, three ways

There isn't one safety stock formula — there are three tiers of it, and most guides only give you the weakest one. Pick based on how much data you actually have, not which one looks simplest.

• Basic (days of supply): SS = Average Daily Sales × Safety Days. Fast, but the "safety days" number is a guess dressed up as math. Fine for a low-value SKU you're not going to lose sleep over.
• Average-Max: SS = (Max Lead Time × Max Sales) − (Average Lead Time × Average Sales). Better because it's built from your own history, but it locks onto your worst historical outlier and keeps protecting against it forever — including the one time a supplier had a customs delay that will never repeat.
• Statistical (Z-score × variability): SS = Z × √[(Lead Time × Demand Variance) + (Average Demand² × Lead-Time Variance)]. This is the one that actually separates demand risk from supply risk and lets you dial in a service level on purpose instead of by accident. It needs more inputs — standard deviation of daily demand, standard deviation of lead time — but those are sitting in your sales history and your PO records already.

The Z-score is your target in-stock probability translated into a multiplier: 1.28 for 90%, 1.65 for 95%, 1.96 for 97.5%, 2.33 for 99%. Higher service level, higher Z, more stock, more cash tied up. There's no free version of this trade-off.

Worked example: calculating safety stock for a real SKU

Take a SKU selling an average of 50 units a day, with a daily demand standard deviation of 15 units — some days 35, some days 80, that's normal noise. Average lead time from the supplier is 30 days, with a lead-time standard deviation of 5 days. You're targeting a 95% service level, so Z = 1.65.

SS = 1.65 × √[(30 × 15²) + (50² × 5²)]
= 1.65 × √[(30 × 225) + (2,500 × 25)]
= 1.65 × √[6,750 + 62,500]
= 1.65 × √69,250
= 1.65 × 263.2
≈ 434 units of safety stock

Notice the lead-time variance term (62,500) dwarfs the demand variance term (6,750) here. That's common for anything shipped internationally with a supplier who doesn't confirm ship dates reliably — the buffer is protecting against your supply chain, not your customers. Run the two terms separately before you commit to a number; if you don't, you can't tell whether the fix is more stock or a better supplier conversation.

Reorder point: what the safety stock number actually protects

Safety stock on its own doesn't tell you when to order. It's an input to the reorder point:

Reorder Point = Safety Stock + (Average Daily Sales × Average Lead Time)

Continuing the example above: 434 + (50 × 30) = 434 + 1,500 = 1,934 units. That's your trigger. When on-hand inventory (including anything in transit that hasn't landed yet) drops to 1,934, you place the next order — not before, or you're paying for storage on stock you didn't need yet; not after, or the 434-unit buffer is the only thing standing between you and a stockout during the next 30-day lead time.

If you're also using an EOQ (Economic Order Quantity) for how much to order each time, safety stock and reorder point don't change — they just sit underneath whatever quantity EOQ tells you to buy.

The mistake we see most often — and one we've made ourselves

The formula is rarely the problem. The problem is that brands calculate safety stock once, at launch, off the first few months of sales data, and then never touch it again. Demand shifts, lead times drift as a supplier's factory gets busier or a freight lane changes, and the number quietly stops matching reality in either direction — too low and you stock out mid-BSR-climb, too high and you're funding someone else's warehouse with aged-inventory surcharges.

Across the $500M+ in managed revenue we've managed for 100+ brands, the single most repeated fix isn't a smarter formula — it's re-running the same formula on a rolling window instead of a fixed one, and re-running it after every seasonal spike, not just before it. We've done this wrong ourselves: sized a buffer off a clean six-month stretch, then watched a Q4 demand spike blow through it because the standard deviation used was calculated on a period with none of that volatility in it.

The other common error is using the average-max method's outlier as if it were normal. One bad shipment from one bad quarter should not set your buffer for the next two years.

When the number is wrong: what to check before you raise the buffer

If you followed the formula and still stocked out, don't just push the safety stock number up and hope. Check, in this order:

• Is the lead-time variance real or guessed? Most sellers use the supplier's quoted lead time as if it were fixed, then wonder why the buffer wasn't enough. Pull your actual PO-to-receipt history and calculate the real standard deviation.
• Did demand actually change? A launch promotion, a competitor stockout, or a seasonal pull can move your average and your variance at the same time. A formula run on stale data will look precise and be wrong.
• Is the service level target still the right one? A 90% target (Z=1.28) on a hero SKU during a launch push is usually too low — a stockout there costs you rank, not just a few days of sales.

And if the buffer is fine but stock is still sitting too long and picking up storage or aged-inventory fees, that's not a safety stock problem — that's a demand-forecast or a liquidation-timing problem, and raising the buffer further only makes it worse.