Replenishment lead time: the most underestimated variable in spare parts inventory
Most teams working to optimize their spare parts inventory pour almost all their energy into a single topic: forecasting demand better. They refine models, test forecasting methods, argue over past consumption. That is useful, but it is often only half of the equation, while the other half, which sometimes weighs more, gets ignored: the replenishment lead time.
On slow-moving parts, supplier lead time, its value as much as its variability, is frequently the primary driver of your stock. Far more than demand. Here is why.
Two sources of uncertainty, not one
Safety stock exists to absorb uncertainty during the replenishment lead time. That uncertainty has two origins:
Demand can vary during the lead time. You expected to consume 2 parts, you consume 5.
The lead time itself can vary. You expected delivery in 3 months, the supplier takes 6.
The classic safety stock formula, when demand and lead time vary independently, combines both sources:
Safety stock = z × √( L × σD² + D² × σL² )
Where:
z: service factor tied to the target service level (for example: 1.65 for 95%)
L: average lead time, in periods
σD: standard deviation of demand per period
D: average demand per period
σL: standard deviation of the lead time, in the same period unit
The key sits in the two terms under the square root. The first, L × σD², captures demand uncertainty. The second, D² × σL², captures lead time uncertainty. Most people know the first and forget the second. That is a costly mistake.
An example where lead time crushes demand
Take a typical slow-moving spare part in an industrial setting:
Average demand: D = 2 units per month
Standard deviation of demand: σD = 1.5 units per month (erratic demand)
Average lead time: L = 3 months
Standard deviation of lead time: σL = 2 months
Target service level: 95%, so z = 1.65
Let us compute both terms separately:
Demand term: L × σD² = 3 × (1.5)² = 3 × 2.25 = 6.75
Lead time term: D² × σL² = (2)² × (2)² = 4 × 4 = 16
Total variance = 6.75 + 16 = 22.75; Square root = 4.77; Safety stock = 1.65 × 4.77 ≈ 8 units
Look at the split. The lead time term accounts for 16 out of 22.75, roughly 70% of the total variance. On this part, lead time uncertainty explains the large majority of the safety stock. Demand, the thing everyone works on, accounts for only 30%.
Which lever is worth the investment?
This is where the example becomes instructive. Suppose you have two possible projects, each requiring the same effort (a simplification for the sake of the demonstration).
Project A: halve demand uncertainty (σD drops from 1.5 to 0.75, thanks to a much better forecast).
Demand term: 3 × (0.75)² = 1.69
Lead time term: 16 (unchanged)
Variance = 17.69; Square root = 4.21
Safety stock = 1.65 × 4.21 ≈ 7 units
Project B: halve lead time variability (σL drops from 2 to 1, by making the supplier more reliable, for example).
Demand term: 6.75 (unchanged)
Lead time term: 4 × (1)² = 4
Variance = 10.75; Square root = 3.28
Safety stock = 1.65 × 3.28 ≈ 5 units
The verdict is clear. A massive effort on demand forecasting moves stock from 8 to 7 units, about a 12% reduction. The same effort on lead time variability moves it from 8 to 5, close to a 37% reduction. Three times the impact on the variable almost nobody looks at.
The contractual lead time trap
There is a second, more insidious effect. In most systems, the lead time used to compute stock parameters (safety stock, reorder point, and so on) is the contractual one, negotiated with the supplier. And its variability is almost never entered: the system implicitly assumes σL equals zero.
Take the same part, but assume σL = 0, as a default setup does:
Demand term: 6.75
Lead time term: 0
Variance = 6.75; Square root = 2.60
Safety stock = 1.65 × 2.60 ≈ 4 units
The system recommends 4 units. Reality needs 8. You are structurally under-protected by half, and you do not know it, because the number on screen looks clean. The result: repeated stockouts on parts you thought were safe.
And lead time does not act only through its variability, its value itself shapes the reorder point:
Reorder point = D × L + Safety stock
The first term, D × L, depends directly on the lead time value. With our part (D = 2, SS = 8) and a true 3-month lead time, the reorder point is 2 × 3 + 8 = 14. If the system uses 5 months by excess, it climbs to 18, meaning 4 units of permanent overstock. If it uses 2 months by default, it drops to 12, and you cover one month less than the real lead time. In both cases, the gap between contractual and actual observed lead time is a silent source of overstock or stockout.
How to think about your real lead time
What this example illustrates is that a lead time is not a number, it is a distribution. Seeing it that way changes how you reason.
One useful approach is to compare, across your history, the order date and the actual receipt date, part by part. You care not only about the average but also about the spread: two suppliers can share the same 3-month average, one very steady, the other delivering sometimes in 1 month, sometimes in 6. The first costs you little safety stock, the second a great deal, even though on paper they look identical.
Another simple approach, when the distribution is highly skewed (which is often the case), is to use not the average lead time but a high percentile, a P80 or P90 for example, that is the lead time met in 80 or 90% of cases. This builds a margin directly into the lead time value, rather than loading everything onto safety stock. It is a different angle from the formula above.
The principle is the same: stop planning on a single fixed lead time value, and look at it as it truly is in your data. On the few parts where lead time is both long and widely spread, the ones that weigh most on your stock, several levers then exist: renegotiation, dual sourcing, consignment stock, or simply an honest measure of the real lead time so you stop relying on a false contractual value.
In short
Before investing in a better forecasting model, one question is worth asking: do you really know the distribution of your replenishment lead times? On slow-moving parts, the opportunity often sits there, in a variable nobody looks at, rather than in demand, which everyone already watches.
Wondering what your replenishment lead times and stock parameters are really hiding? We analyze your spare parts data and quantify your optimization potential, free of charge, for a limited number of industrial companies this quarter (3 spots).
Comments