Four Points of Service Level Cost 41 % More Stock
Gepubliceerd op 22-9-2026 · 3 min leestijd · Zakelijke tools
Lena Hoffmann — Redacteur wetenschap en onderwijs bij OneKitly
Wiskunde · Natuurkunde
Getoetst aan 2 bronnen
Safety stock exists to absorb two kinds of surprise — demand higher than forecast, and a delivery later than promised — and the formula multiplies a z-score by the combined standard deviation of both. The z-score is the only place the service level enters, and it is not linear. For demand of 100 a day with a standard deviation of 20, a ten-day lead time with a standard deviation of 2 days, a 95 % service level gives z = 1.645 and 346 units of safety stock. Raise the target to 99 % and z becomes 2.326: the safety stock jumps to 488, forty-one per cent more inventory for four points of availability. Push toward 99.9 % and the curve steepens again. That is the whole economics of the decision — the first ninety points are nearly free and the last few are where the working capital goes.
Going from a 95 % service level to 99 % raises the z-score from 1.645 to 2.326 and the safety stock from 346 units to 488. The last few points of availability are the expensive ones.
The z-score is a normal table, and that is an assumption
Turning a service level into a z-score assumes demand is normally distributed. Real demand often is not: a product with occasional bulk orders has a long right tail, and the normal curve underestimates exactly the events safety stock exists to survive. The formula still helps — it ranks products correctly and sizes the trade-off — but a 99 % target on a skewed product will not deliver 99 % availability. Where the history shows fat tails, the honest move is to raise the target above what the theory asks for, and to say that is what you are doing.
Lead-time variability is usually the bigger half
The combined standard deviation adds the demand variation over the lead time to the demand level multiplied by the lead-time variation — and the second term is squared against a much larger number. On the figures above, two days of uncertainty on a ten-day lead multiplied by 100 units a day contributes more to the total than the demand noise does. The practical consequence is that negotiating a more reliable delivery window usually cuts safety stock faster than forecasting demand better, and it is often the cheaper of the two projects.
| Service level | z | Safety stock | Reorder point |
|---|---|---|---|
| 95 % | 1.645 | 346 | 1,346 |
| 99 % | 2.326 | 488 | 1,488 |
Uitgewerkt met onze eigen calculator
Veiligheidsvoorraadcalculator
Gegeven
- Serviceniveau (%)
- 190
- Gemiddelde vraag per dag
- 80
- Levertijd (dagen)
- 14
- Standaardafwijking van de dagvraag (σ_d)
- 24
- Standaardafwijking van de levertijd (σ_LT, dagen)
- 4
Resultaat
- Servicefactor Z
- 3,719
- Veiligheidsvoorraad (stuks)
- 1.237
- Bestelpunt (stuks)
- 2.357
Deze cijfers komen uit de calculator hieronder, ze zijn niet met de hand ingetypt — ze worden opnieuw berekend telkens als de tool verandert.
Reken het door met je eigen cijfers →Veelgestelde vragen
- What service level should I target?
- Not one number for the whole catalogue. The right target compares the cost of holding one more unit against the cost of not having it — which differs enormously between a cheap consumable a customer will wait for and an expensive part that stops a production line. Classify the range first, then set a target per class. A single 98 % across everything overstocks the cheap items and understocks the critical ones simultaneously.
- Is the service level the same as the fill rate?
- No, and confusing them flatters the numbers. The service level in this formula is the probability of not running out during a replenishment cycle — a per-cycle chance. The fill rate is the proportion of demand actually satisfied from stock, measured in units. A 95 % cycle service level usually produces a fill rate well above 95 %, because most stockouts are short and cost only a few units. Reporting one and computing the other is a common way to appear to have missed a target that was never the one being measured.
Artikelen die je misschien interesseren
Alle gidsen →Vergelijkbare tools
Bronnen
Een fout in dit artikel gezien?