Dual source or single source: what the manufacturer gains and loses by concentrating

A technician examines a critical aircraft part in a dimly lit manufacturing facility, highlighting the risks of single-source production.

A tier-2 forge shop shuts down — a fire, insolvency proceedings, an export control decision — and a few quarters later a storeman reads “no stock” against a part number he had been ordering without a second thought for years. Between the two, no technical failure whatsoever: only a purchasing trade-off made years earlier and never reopened. What follows takes that trade-off apart from both sides — what concentration really earns, what a qualified second source costs to maintain, and where the line falls.

“Single source” and “sole source” do not describe the same problem

Purchasing vocabulary blurs two situations that have nothing in common. Single source by choice describes a part that only one shop produces although others could: volume was concentrated to obtain a price, a priority ranking when supply tightens, a simple technical interface. De facto sole source describes a part that nobody else knows how to make: a live patent, a proprietary process, a press or furnace of which very few exist worldwide, a special-process accreditation held by a single shop for that family of parts. The first case is handled by contract and can be undone. The second can only be handled by stock, by owning the means of production, or by a decision to redesign.

“Dual source” likewise covers three things that should not be treated as equivalent: two shops delivering the same part number in parallel, a qualified but dormant back-up source, and a second process approved in the type design — different material, different forming route, part declared interchangeable. The three do not carry the same annual maintenance cost, the same wake-up lead time, or the same value on the day the first source stops.

One point comes up far too often in purchasing meetings as though it were a discovery: for a critical part, the approved type design does not freeze only a geometry and a material grade. It freezes a chain of processes — heat treatment, shot peening, NDT, coating — and very often the shop that performs them, because qualification was conducted on that shop’s own equipment. Changing source then amounts to touching the type design. The change is classified, processed by the design organisation, then approved at the level determined; and it is that route, not the commercial negotiation, that sets the real lead time of the operation.

What concentration actually earns, measured honestly

The first gain is the amortisation of non-recurring costs. Forging or moulding tooling, inspection fixtures, machining programmes, the industrialisation data package, the qualification campaign: this spend is committed per part number and per shop. Sharing it between two sources does not quite double it, but it comes very close, and every additional euro is paid for the same part.

The second gain is the learning curve, and it is the worst assessed of the three. On a part with complex geometry, scrap rate, cycle time and dimensional scatter improve with the number of units produced, not with the calendar. Splitting the volume in two means keeping two shops at the top of their curve, where scrap still costs dearly, and slowing the descent of both.

The third gain is commercial weight, which is paid back as allocation priority. When a material or a capacity runs short, the shop serves first the customer who accounts for a major share of its order book, and that priority is not shared pro rata: it is granted by rank. A prime contractor whose volume is split between two shops weighs less at each of them, and ends up lower in both queues on the day capacity runs short — the least quantifiable downside of redundancy, and the one that is paid at the worst possible moment.

To this must be added a less visible gain: in-service feedback. A single source produces a homogeneous population. Non-conformities, process drift and in-service technical events are analysed against a single set of causes. Two sources mean two populations, two documentary configurations, and statistics half as dense for detecting the same drift.

Concentrating production at one supplier saves money on setup costs, helps workers get better at making the part, and gives the manufacturer priority when supplies are tight. But it also means that if something goes wrong with that supplier, it could take a long time to find a replacement.

An Airbus A320 needs a special titanium fitting for the landing gear. If only one supplier makes it, and that supplier has a fire, the airline might have to wait 6 months for a new one. But if there are two suppliers, the switch can happen in weeks, saving the airline from costly delays.

Unit cost against cumulative volume, concentrated source versus split volumeTwo falling curves starting from the same point. The solid curve, volume concentrated on one source, falls faster and lower. The dashed curve, volume split between two shops, levels off higher: each shop stays at the top of its learning curve and the fixed costs are paid twice.unit costcumulative volumeconcentrated sourcesplit volumelearning gap
The gap between the two curves does not close with time, but with cumulative output. It is the number of parts produced, not the age of the contract, that brings unit cost down: splitting the volume leaves both shops in the zone where scrap is still expensive.

The cost of maintaining a second source, line by line

The “piece price” line of a second source is the least interesting item in the file. What makes it expensive is the annual spend required for it to remain a source rather than a name on a list.

What each cost item becomes depending on whether one source or two are maintained
Item Single source Genuinely maintained second source
Non-recurring costs Committed once Committed once per shop
Qualification and first article Initial, then nothing Initial, then repeated at every wake-up
Supplier surveillance One audit plan, one special-process plan Two plans, two accreditation calendars
Floor volume Not applicable Regular orders imposed to keep the skills alive
Documentary configuration One part number, one catalogue Two part numbers, interchangeability to be demonstrated and maintained
Safety stock One chain to cover Two chains, or deliberately reduced coverage
In-service feedback Homogeneous population Split population, slower detection
Recovery lead time after a stoppage Long, everything to be redone Short, ramp-up of an existing line

The item that financial models forget most readily is the floor volume. A special-process accreditation is maintained through periodic audits, and those audits are carried out on parts actually produced. A shop that has made nothing against that part number for several years has lost its operators’ dexterity, has often renewed its equipment, and has sometimes changed its own lower-tier suppliers. Maintaining a second source therefore means buying parts from it that could have been bought more cheaply elsewhere: that is the price of the option, and it is settled every year.

The dormant-source trap. A “qualified” source that no longer produces is a source qualified on paper. Expired accreditations, tooling stored without periodic inspection, the team gone, tier-3 subcontractors replaced: on the day of need all that remains is a file, and requalification is required. The question to ask in review is not “do we have a second source?” but “when did it last deliver a conforming part?”.

What the second source buys: time, not safety

A second source does not make a part better and does nothing to reduce the probability that a shop will stop. It acts on a single quantity: the recovery lead time, that is, the time between the stoppage and the first conforming part delivered. The whole decision comes down to comparing that lead time with the coverage, expressed in months of consumption held in stores and in the pipeline.

With a back-up source already loaded, recovery is counted in weeks rather than months: an existing rate is increased. Without it, the whole path has to be walked again — selection, transfer of the data package, new tooling, first article, qualification testing where the part requires it, approval of the type design change, ramp-up. That path is typically counted in quarters, and in years when long testing or a primary structure part is involved. The order of magnitude alone settles the matter: no reasonable stock covers a full requalification on a high-consumption part.

Recovery lead time after a source stopsTwo horizontal bars starting at the same instant. The upper bar, single source, stretches almost to the right-hand edge: a shop must be qualified before the first conforming part. The lower bar, second source already loaded, stops a quarter of the way along; the remaining interval is the time saved.single source: requalifyfirst conforming partsecond source: switch overtime savedtime since the stoppage
The second source does not remove the shortage, it shortens its invisible part. The left-hand segment is identical in both cases: it is detection and decision. Everything that follows is qualification lead time that redundancy has already paid for in advance.

Between the two extremes lies a third lever, systematically overlooked because it belongs squarely neither to purchasing nor to the design office: ownership of the means of production. Owning the tooling, holding the complete industrialisation data package and a contractual right to have it executed elsewhere turns a requalification into a transfer. It is not given away: tooling ownership is negotiated when the contract is signed, never at the moment of the crisis. But it costs far less than a second line permanently loaded.

The right reflex. For every critical part number, keep two figures side by side: coverage in months and recovery lead time in months. As long as the second exceeds the first, the question is no longer “do we need a second source?” but “with what do we fill the gap?” — stock, owned tooling, a loaded second source, or a redesign of the part.

When a purchasing trade-off becomes an airworthiness matter

Seen from the front line, a sourcing decision taken ten years earlier shows up in a mundane form: an announced lead time that keeps stretching, a component on back order, a minimum equipment list item deferred once more. The missing part is not a bad part, and safety is not directly affected. What is affected is the margin for manoeuvre of everyone who decides afterwards.

The shortage shifts the pressure onto three known points. Deferral tolerance first: an item extended because the part is not arriving ends up being extended for that reason rather than for a technical one, and the justification erodes as the deferral is renewed. Cannibalisation next: legitimate and documented, it multiplies removals and reinstallations, and therefore the opportunities for error, and moves the problem to another aircraft. Procurement last: it is when supply is tight that the grey market becomes attractive and that the traceability of a component deserves to be looked at twice rather than once. The authorised release certificate is never as useful as on the day the part cannot be found through the normal channel.

Hence a methodological point that often surprises buyers: the criticality that matters in a sourcing trade-off is not the safety criticality of the part, it is its ability to ground an aircraft. A part with no effect on safety but without which the aircraft does not depart deserves, in this particular file, more attention than a critical part for which three qualified sources exist.

The decision matrix, and the criteria that outrank price

The trade-off rests on two inputs: what a stoppage costs in aircraft downtime, and how easily the work can be done elsewhere. Unit price does not enter the matrix — it only arbitrates within a cell, once the cell has been chosen.

Matrix of operational criticality against substitutability of the partA square divided into four cells. Top left, grounding part with low substitutability: second source maintained and loaded. Top right, grounding but substitutable: second process approved in the type design. Bottom left: long stock and owned tooling. Bottom right: accepted single source.second sourcemaintained and loadedsecond processapprovedlong stockowned toolingsingle sourceacceptedhighcriticalitylowlowhighsubstitutability of the part
The criticality used here is operational: it measures the ability of the part to immobilise an aircraft. A cell does not say how much to spend, it says by which lever the risk is covered; price only separates two suppliers within the same cell.

Four families of criteria regularly outrank price in a serious file. They deserve to be written down in black and white, because they never come up spontaneously from a tender.

The inputs that tip a file independently of unit price
Criterion What is examined What tips the decision
Geographic concentration Single site, natural hazard, energy dependency, export control regime All the volume produced at a single site
Depth of the chain Who pours the metal, who heat treats, who inspects, who coats Two apparent sources sharing the same tier 3
Supplier health Order book, margin, ownership, dependency on one prime contractor A shop of which one is the majority customer
Programme horizon Residual consumption, last-time-buy notice, component obsolescence Announced end of production with no final buy organised
Technical substitutability Live patent, rare accreditation, tolerances beyond common practice A single shop accredited for the determining process

These criteria all come back to the same economic question: from what level of risk exposure does the expected cost of concentration exceed the recurring cost of redundancy? The crossover always exists. It is simply placed far to the right for standard fasteners, and far to the left for a part coming out of a single shop at the end of a three-link chain.

Expected annual cost against exposure to the risk of a supply breakTwo straight lines that cross. The single-source line starts low and rises steeply with exposure, because the cost of the break is borne in full. The dual-source line starts high, fixed costs paid twice, but rises little. To the left of the crossing point concentration remains the right answer, to the right it no longer is.expected annual costsingle sourcedual sourcetipping pointrisk exposure
The tipping point is not an opinion, it is an intersection. What moves it is not the price of the part but the slope of the line: the more a supply break costs in aircraft downtime, the further the crossing point moves to the left.

The source is decided in the design office, not in purchasing

A part specified by a performance requirement has sources; a part specified by a named process has a supplier. That is probably the most profitable distinction a design data package can carry, and it is settled years before the first invitation to tender.

Three practices make the difference. Describing what the part must withstand rather than how to obtain it, whenever the compliance demonstration allows. Qualifying a second process or a second material from the outset when the part will be manufactured for several decades: qualification then costs a fraction of its emergency price, because it is handled in the same file and with the same tests. And finally retaining control of the data — complete definition drawing, industrialisation data package, ownership of the tooling, right to have the work performed by a third party if the incumbent fails.

Decision sequence between single source, stock and second sourceThree questions in a row. If the stoppage grounds no aircraft, the single source is accepted. If stock covers the requalification lead time, stock is reinforced and the tooling is owned. Otherwise, and if the volume supports it, a second source is qualified and loaded.does the stoppageground an aircraft?does stock coverrequalification?does the volumesupport it?yesnonoyesyessingle sourceacceptedtargeted stock andowned toolingsecond sourcequalified
Three questions in that order, and a single exit per path. The first is operational, the second logistical, the third industrial: reversing the order leads to qualifying a second source for parts that stock already covered. A “no” to the third question releases nothing — it sends you back to stock and to ownership of the tooling, for want of the volume needed to sustain two lines.

The same logic applies to off-the-shelf equipment, where the manufacturer controls neither the design nor the upstream chain. There, the lever is no longer the second source but the clause: a support commitment over a duration aligned with the life of the programme, last-time-buy notice, escrow of the data allowing production to be resumed if the supplier disappears. Such clauses read poorly in a technical meeting and are expensive to obtain after the fact.

What to go and check in your own documentation

On the operator’s side, the industrial trade-off can be read in documents already on the shelf. The illustrated parts catalogue says whether a part number has approved alternates and since when: a part number left without an alternate after many years in service is a signal, not a coincidence. The support contract says what the manufacturer guarantees in terms of availability, at what notice it may discontinue a part number, and what becomes of the tooling on that occasion. The long-lead item list, when it is genuinely maintained, says where the part numbers capable of immobilising an aircraft are to be found.

A man in a warehouse views a 3D model of a mechanical part on a computer.
Digital models streamline manufacturing precision and efficiency.

There remains a grey area that no matrix resolves: the false dual source. Two independent shops, two contracts, two prices — and the same metal heat, the same accredited heat treatment shop, the same patent holder at tier 3. On paper the redundancy exists; on the day of the fire, both sources stop together. Verifying it means going back up the chain beyond tier 1, which few organisations do outside the most critical parts, and which operators generally discover at the same time as the shortage. Such breaks circulate first among practitioners: peer-to-peer exchanges signal a supply strain well before it appears in an official communication, and often before the stores see it.

On the operator’s side, redundancy is not qualified: it is bought. A traced used part returned to service by an approved organisation, standard exchange, a spare part approved outside the manufacturer where the framework allows — these channels exist, but their availability does not depend on the part alone. The maintenance programme, the lease agreement and the aircraft return conditions can close them one by one, and a small fleet does not have the volume that would make the effort worthwhile. That is why two operators of the same aircraft type do not suffer the same shortage: one has several channels and pooled stock, the other has a single channel and clauses that forbid the rest.

There remains the question posed by the opening of this article: when does one reopen a trade-off made ten years earlier? Not on a fixed date, because the annual panel review rarely goes down to the individual part number. The triggers are facts, and there are few of them: a change of shareholder or a collapse of the order book at the incumbent, a lasting extension of the announced lead time with no technical explanation, the takeover or relocation of a lower-tier supplier, consumption that doubles because a repair has become systematic, an announced end of production. Each of them is known to someone in the organisation on the day it happens; what is almost always missing is the route that carries it up to the person authorised to reopen the file.

For anyone preparing to take up a post in continuing airworthiness management or technical support, the useful question to ask in the first few days is therefore not “do we have dual sources?” but “who receives the information when a supplier changes hands, and what does it trigger?”. The answer says in one sentence whether the organisation endures its sourcing decisions or holds them.

The applicable regulations, the certification specifications in force and the operator’s approved documentation take precedence over the foregoing: this article sets out a decision mechanism, it replaces neither a procurement procedure nor a design data package.

Questions fréquentes

Is a second source enough to guarantee the availability of a critical part?

No, and that is the most common misconception. A second source guarantees nothing if it has been dormant for years: expired accreditations, uninspected tooling, operators gone. Nor does it guarantee anything if both suppliers share the same metal heat or the same heat treatment shop. A source only counts if it has recently delivered a conforming part using its own resources.

Does changing supplier for an identical part constitute a change to the type design?

For a critical part, most often yes. The approved type design freezes a chain of special processes and often the shop on which qualification was conducted. The change is then classified and processed by the design organisation, with approval at the level determined. It is that route, and not the commercial negotiation, that sets the real lead time of the switch-over.

How can a real dual source be told apart from an apparent one?

By going back up the chain beyond tier 1, until you identify who melts the material, who performs the heat treatments, who carries out the non-destructive testing and who holds the determining patents. If the same player appears behind both suppliers, the redundancy is fictitious: a single incident will stop both lines on the same day. This mapping is generally maintained only for the most critical parts.

What can be done when a second source is technically impossible?

Three levers remain available and can be combined. Stock, sized in months of coverage against the recovery lead time. Ownership of the means of production: title to the tooling, a complete industrialisation data package and a contractual right to have it executed elsewhere, which turns a requalification into a transfer. And, on long programmes, redesigning the part to move away from the proprietary process.

Why is a parts shortage a safety matter and not only a cost matter?

Because a missing part is not dangerous in itself, but it reduces the margin of everyone who decides afterwards. Minimum equipment list deferrals are renewed for a logistical rather than a technical reason, cannibalisation multiplies removals and reinstallations, and the grey market becomes attractive at the precise moment when traceability would deserve to be checked twice.

Which single indicator should an operator track?

The ratio between coverage in months and recovery lead time in months, critical part number by critical part number. As long as the lead time exceeds the coverage, the gap exists and must be filled by an explicit lever. This indicator has the advantage of being calculated with data the operator already holds, without requiring visibility over the manufacturer’s chain.

Commentaires

Leave a Reply

Découvrez l’écosystème FL330

FL330, c’est toute l’aviation professionnelle — explorez nos autres espaces.