Atlas/A.02 Inventory & replenishment/Service level (z factor)

Service level

Definition

The service level is the quantified promise you set against uncertainty: the probability of not stocking out. It translates into a z factor, drawn from the normal distribution, which directly drives the thickness of safety stock. Choosing a service level is therefore not a declarative exercise: it is a trade-off, figures in hand, between the risk of disappointing a customer and the capital you accept to tie up.

Page signature
The tail you accept
z low demand high demand stockout
Service level
·
Z factor
·
Stockout risk
·
Slide: the golden area is the stockout risk you accept.

Why it matters

The one parameter you truly choose.

In inventory sizing, almost everything is imposed from outside: demand varies, lead times drift, suppliers keep their promises or do not. A single parameter belongs entirely to the company’s decision, the service level. It states how much risk you accept to carry, and it alone turns measured uncertainty into a quantity of stock to finance.

That special place makes it a leadership topic, not a configuration setting. Setting 95% or 99% on a product family is not a technical adjustment: it decides how many customers you accept to disappoint, and how much capital you are willing to freeze to avoid it. When that choice is made by default inside a tool, the company has delegated a major financial decision to a factory value.

You also need to know which service you are talking about. The cycle service level measures the probability of getting through a cycle without a stockout; the fill rate measures the share of demand actually served. The two blur together in everyday language and diverge sharply in the numbers. The same headline 95% can cover two very different commercial realities.

Business impact

Translating service into stock is not linear: it is convex. The z factor grows slowly in the common zone, then runs away as it approaches 100%. Going from 90% to 95% takes a moderate effort; targeting 99.9% takes a buffer that can double relative to 99%, for a commercial gain that is often marginal.

The expert lesson: service levels are differentiated, never uniformised. You set them by looking at what a stockout really costs, immediate lost margin, contractual penalty, risk of losing the customer, and weigh that against the cost of tied-up capital. Critical references justify high levels; the long tail settles for far less. In practice, this differentiation is the lever that frees the most cash without degrading the commercial promise.

The mechanism

From promised percentage to financed stock.

The reasoning assumes demand during the lead time is distributed around its mean. Covering the mean alone means being wrong one time in two. To hold a more ambitious promise, you must sit higher in the distribution, and that is exactly what the z factor measures: the number of standard deviations you move away from the mean to cover the target service level.

The signature above makes this visible. The curve is the distribution of demand; the dark area is what you cover, the golden area the tail you accept to leave exposed. Raising the service level moves the line to the right: the tail closes, but each additional step costs more than the last, because you must reach ever further into the distribution.

One final point matters for interpretation. What is guaranteed here is a service per cycle, over the exposure window. It is not the share of demand served over the year. Steering with one indicator and communicating with the other is a classic source of misunderstanding between supply chain, sales and finance.

SS = z · σLT     z = Φ−1( service level )
Φ−1 the inverse normal, which converts a service percentage into a number of standard deviations; σLT the variability of demand during the lead time. Useful markers: 1.28 for 90%, 1.65 for 95%, 2.33 for 99%, and about 3.09 for 99.9%. Direct reading: moving from 95% to 99% adds roughly 40% of safety stock, and from 99% to 99.9% adds another third.

The traps

Three service-level errors.

01

Confusing the two services

Cycle service level and fill rate do not measure the same thing and do not give the same figures. Setting one and communicating the other creates commercial expectations that the computed stock cannot meet.

02

Inheriting a default value

Many configurations carry a 95% that nobody ever discussed. A decision that commits the tied-up capital of a whole product family deserves an explicit trade-off, informed by what a stockout costs.

03

Chasing perfection everywhere

Announcing 99.9% across the whole portfolio looks virtuous and amounts to financing maximum insurance on references where a stockout costs almost nothing. The cost explodes long before the customer perceives the difference.

The rollout

Four steps to an arbitrated service.

Choose the definition and stick to it

Settle between cycle service level and fill rate, write it down, and align steering and commercial communication on the same definition.

Price a stockout

Estimate, by segment, what a stockout really costs: lost margin, penalty, deferral or loss of the customer. Without that figure, the service level remains an opinion.

Differentiate by segment

Assign service levels by crossing value and regularity (ABC-XYZ logic): high on critical references, moderate on the long tail. This is where capital is released.

Confront promise with reality

Periodically compare the service actually observed with the target. A lasting gap signals mis-estimated variability or a promise the chain cannot keep.

Neighboring concepts

Read next.

From knowledge to action

Who decided your service levels?

A percentage inherited from a configuration commits considerable capital every year. Our Inventory & distribution file re-arbitrates your service levels segment by segment, against the real cost of a stockout.