Safety stock formulas compared: four methods with examples

Safety stock protects against two things going wrong during a replenishment lead time: demand running higher than average, and the delivery arriving later than planned. Different formulas cover one risk, both, or neither in a statistical sense. Comparing them on one item makes the differences concrete.

The example item

The z-score comes from the normal distribution: it's the number of standard deviations that covers the chosen probability of not running out during a replenishment cycle (the cycle service level). Cycle service level is not the same as fill rate, the share of demand filled straight from stock. Fill rate also depends on the order quantity and replenishment policy, so if fill rate is your target, calculate it separately.

All four formulas assume continuous review (you can reorder any day), demand that's independent from day to day and from the lead time, and lead-time demand that's roughly normal. With periodic review, protect across the review period plus the lead time. Calculations use unrounded z values; the numbers shown are rounded.

Method 1: demand variability only

SS = z × σd × √L = 1.6449 × 12 × √9 = 59.2 units

Use it when lead times are genuinely reliable. It assumes daily demand varies independently from day to day and the lead time never varies. When suppliers are late, this method is too low.

Method 2: demand and lead time variability combined

SS = z × √(L × σd² + d² × σL²) = 1.6449 × √(1,296 + 10,000) = 174.8 units

This is the standard choice when both demand and lead time vary, assuming the two are independent. Look at the two terms under the square root: demand variability contributes 1,296 and lead time variability contributes 10,000, about 89% of the total. For this item, a late supplier is a much bigger risk than a demand spike.

That points to a lever besides holding more stock: if the supplier became reliable enough to cut σL from 2 days to 1, safety stock at 95% would fall from 174.8 to 101.3 units, about 42% less. Whether that's worth pursuing depends on what better reliability would cost compared with carrying the extra stock.

Method 3: lead time variability only

SS = z × d × σL = 1.6449 × 50 × 2 = 164.5 units

Use it when demand is very steady but deliveries are not. It's close to method 2 here only because lead time variability dominates this item.

Method 4: max minus average

SS = (max daily demand × max lead time) − (d × L) = 80 × 12 − 50 × 9 = 510 units

A simple rule of thumb that needs no statistics, but it stacks the worst observed demand on the worst observed lead time. Here it gives nearly three times the combined method. Treat it as an uncalibrated worst-case scenario, not a guarantee: past maximums don't bound future ones, and it doesn't correspond to any particular service level.

Side by side

Service level is expensive at the top end: moving from 95% to 99% adds 72.4 units with method 2 (computed before rounding), a 41% increase. That's why it pays to set service levels by item class (see ABC/XYZ analysis) instead of one level for everything.

From safety stock to reorder point

With continuous review, you reorder when inventory position (on hand + on order − backorders) falls to the reorder point:

Reorder point = d × L + SS = 50 × 9 + 174.8 = 624.8 units

Common mistakes

• Mixing time units. A monthly standard deviation with a lead time in days gives nonsense. To convert, daily average = monthly ÷ 30.44 and daily standard deviation ≈ monthly standard deviation ÷ √30.44 (about 66 per month ≈ 12 per day). The square-root step assumes day-to-day demand is independent and stable; 30.44 is the average calendar month, so use working days instead if that's how you count lead time.
• Ignoring lead time variability. Method 1 alone understates risk when suppliers are late.
• Using demand variability when you have a forecast. If you plan from a forecast, the standard deviation of forecast error is usually the better input than raw demand variability.
• One service level for every item. Low-value, easily substituted items rarely need 99%.
• Intermittent demand. Items with many zero periods don't follow a normal distribution well; treat formula results for them with caution.

The safety stock calculator computes all four methods side by side for your own numbers, and the reorder point calculator checks whether it's time to order.