Safety Stock Calculator

The calculation itself

Safety stock is the buffer held against demand and lead time being worse than expected: z × σ × √lead time, where z is the service factor for your target service level and σ is the standard deviation of demand per period.

The z factor is where the service level enters. 90% service is z = 1.28, 95% is 1.65, 98% is 2.05, and 99% is 2.33. Each step up costs more stock than the one before it, and the last few points cost the most.

A concrete case

Weekly demand of 40 units with a standard deviation of 12, and a three-week lead time. σ over the lead time is 12 × √3 = 20.8 units.

At 95% service: 1.65 × 20.8 = 34 units, costing $612 of tied-up capital at $18 a unit. At 98%: 2.05 × 20.8 = 43 units, $774. At 99%: 2.33 × 20.8 = 48 units, $864.

So moving from 95% to 99% service costs 14 extra units, a 41% increase in safety stock to remove four percentage points of stockout risk. Whether that is worth it depends entirely on what a stockout costs.

What the number hides

The formula assumes demand is normally distributed, and for slow-moving or highly seasonal products it is not. A product selling zero units most weeks and forty in one has a standard deviation that describes nothing useful, and the calculated safety stock will be wrong in both directions.

It also treats service level as the thing to optimise, when the real objective is total cost. The right service level is where the marginal cost of another unit of safety stock equals the marginal saving in stockout cost, and that point is different for every product.

Where to go from here

Set the service level by product rather than as a company standard. An A-class item that customers buy specifically deserves 98%; a C-class accessory that most buyers would substitute deserves 85%, and applying one number to both overspends on one and underserves the other.

Then measure the actual stockout rate against the target. A 95% service level that produces stockouts in 15% of cycles means the demand variability input is understated, which is more common than not.

The cost of the last few percentage points

Safety stock rises non-linearly with service level because the normal distribution's tail is thin. Going from 50% to 90% costs 1.28 σ. Going from 90% to 99% costs another 1.05 σ. Going from 99% to 99.9% costs a further 0.77 σ, nearly as much again for a tenth of the improvement.

For the example product: 90% service needs 27 units, 99% needs 48, and 99.9% needs 64. The last 0.9 percentage point costs 16 units, which is more than the first forty percentage points cost.

The practical consequence is that near-perfect availability is achievable and expensive, and that the decision should be made deliberately rather than inherited. Most catalogues have a handful of products where 99% is genuinely right and a long tail where 90% is fine, and the money freed from the tail usually funds the head several times over.