The stores report coverage measured in weeks for a pylon fitting; the supplier quotes a lead time measured in quarters. Between the two sits a line that consumes the part every working day. The question is no longer whether just-in-time was a good idea: it is where the point now lies beyond which holding stock costs less than the gap it prevents.
What just-in-time was really buying
Just-in-time was never a moral doctrine; it is a trade-off between two costs. On one side, the carrying cost, the permanent rent paid on a part that sits idle. On the other, the cost of a shortage, the sudden one-off bill for a part that is missing. Reducing stock amounts to exchanging a certain, regular expense for an uncertain, concentrated risk. The deal pays off as long as that risk stays small, and it stayed small for a long time.
Carrying cost breaks down into line items that can be priced: the capital tied up and its cost, the store floor area and its storage conditions — temperature, humidity, electrostatic protection, ventilated cabinets for shelf-life items —, materials handling, cycle counting, insurance and, above all, obsolescence. Not wear: the risk that a change of definition renders a perfectly new batch unusable. The usual orders of magnitude in industrial management put the whole at a far from negligible fraction of the value tied up each year, with appreciable differences from one sector to another; in aerospace, the share linked to storage conditions and documentary tracking pushes that total upwards. What matters for the trade-off, moreover, is not the rate chosen, but the fact that it is applied in the same way to every part number being compared.
This cost has one formidable flaw: it is visible. It appears in a profit and loss account, it can be compared from one site to another, it can be cut by annual target. The cost of a shortage, by contrast, is only seen after the fact and appears on no dedicated line: it dissolves into overtime, rescheduling, express freight and late-delivery penalties. Two decades of working capital reduction were therefore played out against an opponent nobody was measuring.
Just-in-time is a balance between the cost of storing parts and the cost of running out. Storing costs are easy to see, but the cost of shortages (like delays and extra work) is hidden. When shortages become more expensive, the balance shifts, and keeping more stock makes sense.
An Airbus A320 assembly line stops for 2 hours because a critical pylon fitting is missing. The cost isn’t just the part (€5,000) but also the idled workforce (€20,000), rescheduling (€10,000), and delayed delivery penalties (€30,000). Total: €65,000. Keeping one extra fitting in stock (€5,000 + storage) would have avoided this.
The cost of a stoppage, item by item
A line stoppage does not cost the price of the missing part. It costs, first, the idled workforce, which remains paid and qualified while the station waits. It then costs out-of-sequence work: the skipped operation will have to be picked up later, in a less accessible position, with improvised access means, often after neighbouring items have been fitted. Execution time multiplies, the probability of non-conformity rises, and every departure from the sequence has to be recorded, closed out and re-inspected.
Finally, it costs downstream: a test bench that is not fed at its scheduled slot, an acceptance delayed, a delivery date that slips. For the operator waiting for the aircraft, the bill changes nature again — a schedule to be rebuilt, crews to be reassigned, capacity leased at the day’s market price. That is why the cost curve of a shortage is not a straight line: the very first days can be absorbed with a little flexibility, then the rescheduling margin runs out, and beyond that the effect propagates to the following stations, where it is paid for several times over.
Three terms moved in the same direction
An equilibrium point does not move on its own. Here, three quantities changed at the same time, and two of them push in the same direction.

Lead time lengthens in cascade, not in steps
The lead time quoted by a tier-one supplier is only the last link. Upstream, the material stacks up — billets, forgings, investment castings —, along with qualified surface treatments and non-destructive testing, each with its own queue and its own capacity. When a process is qualified at a single source, that source’s queue becomes everyone’s lead time. It is the cumulative total, and not the tier-one order book, that has to be covered.
Dispersion weighs more than the mean
Safety stock is not sized on the mean lead time: it is sized on the dispersion of that lead time and on the target service level. A supplier whose mean has not moved, but whose distribution tail has stretched, mechanically imposes thicker coverage. This is the trap of dashboards that track only the contractual lead time: they show a stable situation while variability, the only quantity that matters for sizing, has changed by an order of magnitude.
The price of a lost hour has risen
When a plant runs at the ceiling of its capacity, an hour lost is not made up on Saturday: it is subtracted from the annual total. Full order books have removed the catch-up slot that used to absorb supply hiccups, and have therefore made every interruption more expensive. Carrying cost, for its part, has not fallen in the same proportion.
Where to place the buffer, and in what form
Deciding that stock is needed says nothing about where to place it. The further a part advances along the flow, the more value it has absorbed and the more specific to one definition it becomes. A buffer of generic material costs little to carry, converts into several part numbers and survives a change of definition; a buffer of finished sub-assemblies is immediately usable, but frozen on one standard, and is scrapped as a block if that standard changes. Between the two lies the decoupling point, chosen item by item and not by site-wide policy. The operational criterion is the delay the downstream stage is prepared to accept: the buffer must be placed far enough downstream for the residual conversion time to fit within that accepted delay, and no further, failing which added value is carried for protection nobody needed. Placing the decoupling point one notch too far upstream costs time on the day the part is missing; placing it one notch too far downstream costs capital and exposure to scrap.
Stock is not, in any case, the only form of buffer. Capacity held in reserve — a machine, a spare set of tooling, an operator qualified on two stations — is a stock of hours rather than of parts: it does not expire, it undergoes no change of definition, but it only protects against internal contingencies. On the operator’s side, the rotable pool and standard exchange play the same role by sharing the capital tied up between several holders, at the price of a return logistics chain that itself becomes critical.
| Form | Capital tied up | Exposure to a change of definition | Time to convert into a serviceable part |
|---|---|---|---|
| Material or rough part upstream | Low per unit, high volume | Low: the material stays generic | Long: the whole cycle remains to be done |
| Sub-assembly or finished component | High per unit | High: frozen on one standard | Short: available immediately |
| Reserved capacity (machine, tooling, qualification) | Asset tied up, no material | None | Variable, and of no effect on a material shortage |
| Rotable pool and standard exchange | Shared between holders | Medium: depends on modification status | Short if the return logistics keep up |
Non-conforming stock is not stock
Coverage has value only if the part leaves the stores usable on the day it is called for. That presupposes a receipt in due form with the accompanying release documentation, batch traceability that survives splitting, storage conditions that are respected and recorded, management of calendar limits for everything that ages while standing still — elastomers, adhesives, sealants, batteries —, first-in first-out rotation, strict segregation of items awaiting a decision, and a modification status known on issue as on receipt.
The trap. Thickening coverage without reviewing calendar limits amounts to manufacturing scrap in advance. A batch bought for consumption that runs beyond the remaining shelf life is paid for twice: in carrying cost, then in scrap. The same mechanism applies to a part stocked just before an announced change of definition, which becomes a non-conformity warehoused at its owner’s expense.
The right reflex. Before increasing coverage, cross-check three pieces of data: the item’s remaining shelf life, its modification status and the actual history of receipt dates. It is the dispersion observed on those dates, and not the lead time stated in the contract, that must set the safety level.
Redoing the calculation in your own organisation
The useful work is not to choose between two doctrines, but to reclassify one’s own portfolio of items. The approach comes down to a few steps, and it can be carried out with data the organisation already holds.
- Classify every part number by the consequence of a shortage: immediate stoppage of a station, a workaround possible with out-of-sequence work, or postponement with no effect.
- Extract from the receipt history the lead time actually observed and its dispersion, supplier by supplier, rather than the contractual lead time.
- Separate regular consumption, which can be forecast, from erratic consumption, which belongs to a service-level logic rather than to forecasting.
- Size item by item, with no uniform coverage rate, and revise as soon as a definition changes or a supplier’s workload shifts.
- Track the indicator that matters: the number of interruptions avoided, and not only stock turnover.
| Criterion | Strict just-in-time | Sized coverage |
|---|---|---|
| Nature of the expense | Low and regular, then brutal in the event of a shortage | Regular, predictable, budgetable |
| Sensitivity to lead-time dispersion | Very high: every delay travels through to the station | Damped as long as the delay stays within the coverage |
| Exposure to obsolescence and calendar limits | Almost none | Real, to be managed actively |
| Effect on supplier quality | Makes defects immediately visible | May hide them if the buffer is not monitored |
| Reaction to a change of definition | Immediate, with no stock to run down | Costly: a frozen batch to deal with |
What remains to be settled
Three grey areas remain, and none of them is settled in a spreadsheet. The first is the question of who carries: asking the supplier to hold stock off-site or on consignment does not remove the carrying cost, it moves it into the unit price, usually with a margin attached. The second is the bullwhip effect: buffers built up in a hurry, all at the same time, send upstream an apparent demand higher than actual consumption and further degrade the very lead times one was trying to cover. The third is the most insidious: a comfortable buffer masks a recurring quality defect at a supplier, until the day it empties all at once.
For anyone preparing to take up a post or to be qualified in stores, in production scheduling or in technical purchasing, the shift is clear: the skill sought is no longer stock reduction, but the ability to justify a coverage level part number by part number, with receipt data and a reasoned reading of criticality. It is a line of reasoning that is documented and defended in an audit, and it is worth preparing as such; shop-floor feedback on the subject circulates mainly among peers, notably on the FL330 forum.
Nothing above replaces a regulation or a procedure: the applicable rules, the requirements of the authority and the approved documentation of the operator or the organisation prevail over this article, which describes only the economic mechanism behind them.
Questions fréquentes
How do you size coverage when there is no reliable lead-time history?
You rebuild the history from the receipt dates recorded in stores, which exist even where no supplier indicator has been maintained. Failing that, you start from the quoted lead time plus an acknowledged margin, you write that margin down as an assumption in the file, and you revise it at every actual delivery. What matters is to date the decision and make it revisable, not to find the right figure first time.
Is it better to stock the material or the finished part?
Material costs less to carry, serves several part numbers and survives a change of definition, but it leaves the whole manufacturing cycle still to be done. The finished part is immediately available and genuinely protects the assembly station, at the price of heavier capital and a risk of scrapping the whole batch. The choice is made part number by part number, according to the residual conversion time and the stability of the definition.
Does stock held off-site at the supplier remove the carrying cost?
No, it moves it. The supplier ties up the capital, occupies the floor space and bears the obsolescence risk, then passes the whole on in the unit price with its own margin. The arrangement remains attractive when that stock is shared between several customers, much less so when the part is specific to a single customer.
How do you stop a buffer stock turning into scrap?
By tying every coverage decision to two pieces of data: the item’s remaining shelf life and its modification status. Coverage that exceeds the remaining shelf life schedules scrap, and coverage built up just before an announced change of definition amounts to warehousing a non-conformity. This review must be a step in replenishment, not an after-the-fact check.
Does buffer stock mask supplier quality problems?
It can, and that is the strongest historical objection to the return of stock. A buffer absorbs delays and rejects without anyone noticing, until the day it empties all at once. The counter-measure is to monitor the buffer level as an indicator in its own right: coverage that erodes slowly without any rise in consumption signals an upstream problem, not a need to top up.
What does this reasoning change for a storekeeper or a planner preparing for a qualification?
The expected skill shifts from stock reduction towards justifying a coverage level. You have to be able to extract a lead time actually observed and its dispersion, classify part numbers by the consequence of a shortage, then defend the sizing adopted in front of an auditor. This reasoning is documented, and it is prepared with the same rigour as a stores procedure.

Leave a Reply
You must be logged in to post a comment.