Every supplier approval you sign is a bet placed with someone else's process and someone else's discipline. Years later, when parts arrive late or porosity shows up in the fifth batch, nobody remembers the sales pitch. They remember who signed the approval. Across two decades in automotive and aerospace, I have learned that the moment of maximum leverage over a supplier is the week before you approve them — not the year after.
The dangerous supplier is not the obviously bad one. It is the supplier you cannot afford to leave once you have said yes. Tooling sunk into their plant, a validation campaign built around their process, a sub-tier relationship nobody else can replicate — these are the handcuffs. Qualification is the last moment you can still walk away cheaply, and most organisations squander it by treating the approval as paperwork rather than an engineering decision.
This article sets out the four disciplines that make qualification earn its name: capability assessment on the floor, sample approval that interrogates rather than admires, an honest dependency audit, and a decision structure that survives schedule pressure. Each one converts a gamble into evidence.
Capability Assessment Before Charm
Start with the question most audits dodge: can this process, as physically configured today, hold the tolerances we are asking for? A supplier with a capable process and sloppy documents will usually survive; a supplier with beautiful documents and a marginal process never will. Go and look at the machines, not the manuals. If you are buying machined components, ask what the spindle bore is, how the machine behaves thermally across a shift, and how the operator compensates for tool wear.
The ratio between the process's natural spread and your tolerance band tells you more than any certificate. Push the supplier to show machine capability studies on parts similar to yours — same material, similar cycle times, same fixturing philosophy. A five-axis cell that demonstrates superb capability on aluminium brackets has told you nothing about thin-walled steel housings. Capability is contextual, and context is exactly what sales presentations leave out.
Assess the second layer as well: maintenance discipline and staff turnover in the cells that will run your work. I have audited plants that wrecked their delivery record simply because the setter who understood the fixture left and nobody had written down why the shim was there. Ask to see the setup sheets. Ask how many people can set the job. A process that lives in one person's head is not a process; it is a single point of failure wearing a machine-guard shirt.

Sample Approval: Parts Are Interrogated, Not Admired
Sample approval is where good intentions collapse into a signature on a dimensional report. Resist that. First-time samples must be produced on production tooling, production material from the production sub-tier, at production cycle time, by production operators. Anything else is a laboratory demonstration. State in the qualification pack that samples run on prototype tooling will be rejected regardless of measurement results, because the parts you will buy for the next decade come from the other tool.
Then interrogate the parts properly. Full dimensional layout on every feature, yes — but also material certification traced to the heat or cast number, hardness traverses where case depth matters, and a proper look at the surface. For weldments and castings, section a part. Radiograph it if the integrity class demands it. Etch a machined surface to reveal grinding burn. Destructive evaluation of the first article is the cheapest insurance you will ever buy; the same defect found in the field costs orders of magnitude more and takes your name with it.
Watch how the supplier handles a failed sample, because that failure is a gift. Do they argue the drawing interpretation, or do they pull the fixture apart and find the datum scheme was ambiguous? Does a corrected batch arrive within days, or does the response come as a polished presentation with no changed process? The behaviour during sample approval previews the behaviour during every future crisis. A supplier who hides, blames or decorates during courtship will not transform under pressure.
First article interrogation sequence
- 01Provenance checkMaterial certification traced to heat or cast number, sub-tier identified.
- 02Full dimensional layoutEvery feature against the drawing, with the datum scheme made explicit.
- 03Metallurgical evaluationHardness traverses, sections, radiography and etching where the integrity class demands.
- 04Process confirmationVerify parts came from production tooling, material, cycle time and operators.
- 05Failure behaviour testHow the supplier responds to a rejected sample predicts every future crisis.
Dependency Audit: Can You Afford to Leave Them?
Here is the question that separates qualification from wishful thinking: once approved, what does it cost us to exit? Do this audit honestly, in writing, before approval. If the moulds, fixtures or dedicated gauges sit in their plant, understand the contractual mechanics of retrieving them, and whether they are even portable. Some tooling is married to a specific machine tonnage and platen size; it is retrievable but not transferable. Write that down as a known constraint.
Sub-tier exposure is the second trap. A casting house that buys its alloy from a single smelter, a coater with one approved chemistry supplier, a stamper reliant on one mill's particular coil characteristics — each extends your dependency chain a link further, and qualification is the only point where you can demand visibility into it. Ask for the sub-tier flow-down directly. Where the chain is opaque, price that opacity into your decision rather than pretending it away.
Capacity is dependency's quiet cousin. A supplier running your parts at full utilisation on one line, with no alternate routing and no surge plan, is not a partner; they are a bottleneck with a purchase order. Ask what happens when their largest customer doubles an order, when the machine breaks on a Friday night, when the annual shutdown lands in your peak month. Suppliers with genuine capacity buffers will show you the routings. Suppliers without them will show you a commitment statement. Learn to hear the difference, and treat unaddressed dependency as a disqualifying condition, not a negotiation point.
The Cost of a Wrong Yes
Nobody approves a bad supplier on purpose. The wrong yes happens through accumulation: a launch date looming, a purchasing function measured on piece price, a plant tour with good coffee. Make the cost of a wrong yes explicit before the decision. Where the part is safety-relevant or flight-critical, a systemic escape means containment, sorting, possible re-validation of assemblies — and in aerospace, potential suspension of your own approvals while you prove the supplier's parts retrospectively. That liability does not transfer with an indemnity clause; it lands on the name on the product.
The commercial cost compounds in ways piece-price analysis never captures. Every rejected lot triggers a quality engineer's week, an expediting premium, a line-side containment and often a premium freight shipment to rescue the build plan. Every concession request opens an engineering review. Every emergency source qualification repeats the work you thought you had finished, except now under duress and at whatever price the market asks. A marginal supplier is not cheap; they are a subscription service for problems.
The wrong yes is the one decision in the supplier lifecycle that cannot be quietly reversed, only expensively survived.
The reputational dimension deserves sober treatment too. Your customer does not see the supplier's name on the delivery; they see yours. When a sub-standard fastener or an out-of-spec forging reaches an end customer, the corrective action cascade, the audits of your own systems, the loss of preferred status — all of it belongs to you. I have never seen a contract clause that changed that arithmetic.
What a bottleneck shows you versus what a partner shows you
Signals of unmanaged dependency
- A commitment statement instead of alternate routings
- Sub-tier identity described as confidential or undefined
- Tooling that cannot be physically transferred to another machine
- No answer to the Friday-night breakdown question
Signals of managed capacity
- Documented routings and surge options you can inspect
- Sub-tier flow-down provided on request
- Tooling portability assessed and written down as a constraint
- Named escalation path with response times
The Approval Board and the Discipline to Say No
Structure the decision so that no single function can carry it. My preference is an approval board with engineering, quality, logistics and sourcing in the room, each holding a genuine veto, and the criteria written down before the supplier is toured. Written criteria matter because supplier selection is emotionally loaded: someone has already championed this candidate, someone's bonus is tied to the timing, and the tour was impressive. The criteria are the anchor that keeps the meeting tethered to engineering evidence.
Give the veto teeth. If quality can be overruled by a programme manager under schedule pressure, you do not have a board; you have a theatre with minutes. A workable rule is that any overridden veto must be signed by the person overruling it, with the specific residual risk named and an interim control assigned. Remarkably few overrules survive being written down in those terms. The discipline is not bureaucratic — it converts a vague gamble into an owned, visible decision.
Condition the approval where honest uncertainty remains. A limited approval covering a defined part family and process, with mandatory re-review after an initial production period, lets you capture a needed source without pretending the evidence was complete. The conditional yes keeps you honest and keeps the supplier motivated, because graduation to full approval depends on demonstrated performance, not on the passage of time and everyone's forgetfulness.
Disqualifying Conditions Worth Respecting
Some findings should end the conversation regardless of schedule pressure. Falsified or recycled inspection records — I have caught reports where the same measurement set appeared under different part numbers with serial numbers inadequately changed — is one. Refusal to grant access to the actual production site, or a tour that steers you away from the cells running your work, is another. A sub-tier that cannot be identified at all, or a financial position so fragile the qualification might outlive the company, both belong on the list.
Treat environmental and compliance posture as engineering, not decoration. A plating line operating without proper effluent control is one inspection away from shutdown, and when the environmental authority closes the line, your parts stop with it. A supplier with no control over calibration of the gauges they use to accept your parts is telling you, plainly, that their inspection results are decorative. These are not minor nonconformities to be noted and monitored; they are structural faults in the foundation you are about to build a supply chain on top of.
The hardest discipline is walking away late. You have invested audits, samples, engineering hours; the sunk cost argues for one more chance. But every month of qualification effort is trivial against a decade of containment. If the evidence says the process cannot hold the tolerance, or the dependency cannot be managed, the correct decision is no — delivered early, explained honestly, and recorded. The suppliers you decline politely today are sometimes the suppliers you re-engage in five years when they have genuinely changed. The ones you approve wrongly, you are simply stuck with.
That is the whole point of treating qualification as an engineering decision. The approval signature is permanent in a way few other decisions in the quality system are. Everything you put in front of it — capability data, interrogated samples, a written dependency audit, criteria fixed before the tour — exists to make that single signature an act of evidence rather than an act of hope.
