Service level: the stockout risk you choose to accept
Service level is the share of replenishment cycles in which you expect to have enough stock to cover demand without running out. It is the one inventory number nobody can work out for you, because it is a choice about risk, and every point you add has a price in cash sitting on a shelf.
What service level means
Service level, in inventory, is the share of replenishment cycles in which you expect to meet demand without running out of stock. It is a target you set, not a figure your store reports back to you. A 95% service level means you accept running short in roughly one cycle in twenty, and you size your stock to make that true.
The word trips merchants up because it sounds like a quality score. It is closer to an insurance decision. You are saying how much cover you will pay to hold, and therefore how often you are willing to be caught out. Set it high and the stock sits there, tying up cash and shelf space. Set it low and you free the cash, but you will meet the sold-out page more often on the lines you least want it on.
A worked example
A candle SKU sells 4 units a day. Your supplier takes 14 days from purchase order to delivery. Lead time demand is 4 times 14, so 56 units. Reorder at 56 and you have covered an average fortnight exactly, with nothing spare, and roughly half of your fortnights are busier than average. That is a service level near 50%, which is why no one orders at bare lead time demand.
Push the target up and you add cover on top, sized from how much the daily number bounces around. On a SKU that swings between 2 and 7 units a day, that cover might be 20 units, so the reorder point becomes 76. You now carry about five extra days of stock on that line, all year, in exchange for far fewer cycles that end in an empty bin.
What each target asks you to hold
| Service level target | Safety stock multiplier | What you are accepting |
|---|---|---|
| 90% | 1.28 | Roughly one cycle in ten runs short before the delivery lands |
| 95% | 1.65 | Roughly one cycle in twenty runs short |
| 98% | 2.05 | Roughly one cycle in fifty runs short |
| 99% | 2.33 | Roughly one cycle in a hundred runs short |
The multipliers are not linear, and that is the whole trade-off
Those multipliers come from the normal distribution, and they are the reason service level is a judgement call rather than a slider you push to the top. Moving from 95% to 99% does not cost you four percentage points of stock. It multiplies your safety stock by 2.33 over 1.65, about 41% more units on that SKU, to remove four stockout cycles in a hundred.
And 100% has no multiplier at all. There is no quantity of stock that covers every possible fortnight, because demand has no ceiling you can name in advance. A merchant who says they want to never run out is really saying they have not decided yet.
Cycle service level and fill rate are not the same number
Two different measures travel under the name service level, and suppliers and spreadsheets rarely say which one they mean.
- Cycle service level counts cycles. Out of your last twenty replenishment cycles on a SKU, how many ran to the delivery date with stock still on hand?
- Fill rate counts units. Out of the units customers tried to buy, how many could you actually ship?
They come apart because cycles that go short usually go short at the very end, by a handful of units. A store can sit at a 90% cycle service level and still fill 99% of units ordered. If somebody quotes you a service level, ask which denominator they used before you compare it to your own.
Picking a target per SKU, not per store
One target across a whole catalogue is the usual mistake. A hero SKU that carries your margin, that customers come to you specifically for, and that a competitor sells too, earns a high target: the cover costs you a few days of stock and protects the line you cannot afford to lose. A slow long-tail variant with a two-week lead time and three sales a month does not. Holding 99% cover on it means paying to protect revenue that barely exists.
Two inputs should move the number up: a long or unreliable lead time, and demand that jumps around. Both widen the range of what the next fortnight might do, so both demand more cover for the same target. Two inputs should move it down: cheap, fast resupply, and a customer who will happily wait.
Related terms
- Safety stock is service level turned into units. Pick the target, measure how much your daily sales vary, and the formula gives you the buffer to hold above lead time demand.
- Reorder point is where the target lands in practice: lead time demand plus safety stock, the level at which you place the order.
- Lead time is the input only you can supply, and the one that most changes what a given service level costs you.
- Stockout is what the target is priced against: the cycle where the cover ran out before the delivery arrived.
- Inventory carrying cost is the bill for a high target, the cash, storage and spoilage risk in every spare unit.
- Backorder is the escape hatch merchants use instead of cover, and it has its own cost in cancellations and goodwill.
- ABC analysis is how merchants decide which SKUs deserve the high target in the first place.
Questions
What is service level in inventory, in one sentence?
It is the share of replenishment cycles in which you expect to meet demand without running out, chosen by you and then turned into units of safety stock.
How is service level different from fill rate?
Cycle service level counts cycles that finished with stock on hand. Fill rate counts units you could actually ship. A store can run a 90% cycle service level and still fill 99% of units, because short cycles usually go short only at the very end.
What service level should a small store aim for?
There is no single right number, and anyone quoting one without knowing your margins is guessing. The useful approach is to set it high on the few SKUs that carry your revenue and where a customer would buy elsewhere, and lower on slow lines with cheap, fast resupply.
Why can I not just set 100%?
Because demand has no upper bound you can name in advance, so there is no finite quantity that covers every possible lead time. The multipliers climb steeply near the top: going from 95% to 99% adds roughly 41% more safety stock to remove four short cycles in a hundred.
Does a long supplier lead time change the target?
It does not change the target, but it raises what the target costs. A longer lead time means more days of uncertain demand to cover, so the same service level needs more safety stock on the same SKU.
Do I need to calculate a service level to set a reorder point?
No. Plenty of merchants run reorder points from sales velocity and lead time, then keep a cushion they have learned to trust. Naming a service level simply makes that cushion explicit, so you can see what you are paying for it.