The stockout rate measures how often demand arrives without being servable. It is the mirror of the service level, seen from the damage suffered rather than the promise kept. Its real value lies not in the counting, often done well, but in quantifying what a stockout genuinely costs, almost always missing, and yet the only thing able to say what service level it is rational to target.
Why it matters
Most companies know how many stockouts they suffer. Very few know what they cost. These two pieces of information have quite different uses: the first feeds a dashboard, the second allows a decision. Without it, any service target remains an opinion, defended more or less well depending on the balance of power between sales and supply chain.
The cost of a stockout is not the margin on the missed sale, and that is the most widespread error. What actually happens depends on the behaviour of the customer who was not served, and that behaviour varies enormously. Some wait, and the company loses almost nothing. Some take another reference from the catalogue, and the loss is limited to a margin difference. Some buy elsewhere this time. A few never come back.
The signature shows this split across a hundred customers. Moving the slider changes one thing only, the share of those who will not return, and the average cost of a stockout changes completely. It is those few customers, marginal in number, who carry most of the economic damage, because what is lost is not a sale but a relationship.
The second panel draws the consequence, and it is the most important point on this page. The service level it is rational to target derives directly from that cost, through newsvendor reasoning. In the example it moves from about seventy percent when nobody leaves, to over ninety when one stockout in five loses the customer.
This reading turns the usual service-level negotiation on its head. Rather than pitting a commercial demand against a financial constraint, it makes customer behaviour the input of the calculation. The question becomes factual and verifiable: what do our customers actually do when we fail to serve them?
The answer can be measured, and rarely through a heavy study. Tracking unserved orders over a few months, cross-referenced with those same customers’ later purchases, gives a sufficient estimate of the four behaviours. That modest effort turns a debate of opinions into a documented trade-off, and it often reveals that the target service was badly calibrated, in one direction or the other.
The expert lesson is that stockout cost differentiates even more strongly than service. On a commoditised, readily substituted reference, an unserved customer takes something else and the company loses little. On a signature reference, or on a component that halts a customer’s production line, a single stockout can jeopardise an entire contract. It is this heterogeneity, not a global target, that should drive buffer differentiation.
The mechanism
Measuring the rate itself calls for a rarely respected precaution. A stockout only becomes visible if the demand is recorded, yet unserved demand often leaves little trace: the customer gives up before ordering, or the salesperson spontaneously offers something else. Systems therefore measure served demand, and structurally understate stockouts.
Costing then follows the breakdown in the signature. Each behaviour carries a loss: near zero for waiting, a margin difference for internal substitution, the full margin for a lost sale, and the remaining value of the relationship for a customer who leaves. The average cost of a stockout is the weighted average of these losses by the frequency of each behaviour.
Once that cost is known, the economically justified service level derives from the ratio between what a stockout costs and the sum of that cost and the holding cost. This is exactly the critical ratio of the newsvendor model, applied here not to a one-shot decision but to the routine setting of buffers.
This chain of reasoning closes the domain on itself. Holding cost, covered on the neighbouring page, gives the denominator. Customer behaviour gives the numerator. Their ratio gives the service to target, which feeds the safety factor, which sizes the buffer, which determines the capital tied up. The whole domain sits in that sequence.
The traps
Costing a stockout as a missed sale ignores the only case that is genuinely expensive, the customer who never returns. This understatement leads to targeting too low a service on references where the relationship is at stake.
Unserved demand rarely leaves a trace: the customer gives up before ordering, or an alternative is offered spontaneously. The measured rate is therefore structurally below the real one, and decisions taken on that basis are biased.
A stockout on a commoditised, substitutable item and one on a component that halts a customer’s line have nothing in common. A single average cost over-protects the former and leaves the latter exposed.
The rollout
Put in place a record of requests that do not convert, including those refused upstream. Without that trace, the measured rate will stay structurally optimistic.
Track, over a few months, what unserved customers do: wait, substitute internally, buy elsewhere, disappear. A sample is enough to estimate the proportions.
Assign a loss to each, from a margin difference up to the remaining value of a lost relationship. It is the last line that dominates the result.
Compute the ratio of stockout cost to holding cost to obtain the justified service, segment by segment, then compare it with the levels currently configured.
Isolate references where a single stockout commits a contract or halts a customer, and protect them by risk rather than by the average.
Neighboring concepts
From knowledge to action
The service to target follows from their behaviour, not from a negotiated objective. Our Inventory & distribution file measures the four reactions, values them and resets your service levels segment by segment.