October 24, 2026 · Quality Systems Guide
Every connector plant produces parts that miss a requirement. The expensive part is almost never the metal — it is the four weeks of arguments that follow, because nobody agreed in advance who is allowed to decide what happens to those parts.
Disposition is the formal answer to one question: what happens to this specific quantity of this specific lot, and on whose authority? There are only five answers, and two of them need permission from outside the factory. Everything else in this guide is about getting the right answer quickly, with a record that survives an audit three years later.
A connector is nonconforming when it fails a requirement, and the requirement is never a single document. Three of them define it, and which one you broke decides who may dispose of the part.
| Reference document | What it controls | Who owns a departure | What the departure actually is |
|---|---|---|---|
| Drawing (with revision) | Dimensions, materials, finish callout, marking, keying, contact arrangement | Design authority / OEM engineering | An engineering change, or a deviation if granted for a limited quantity |
| Product specification / slash sheet | Performance: contact retention, DWV, insulation resistance, durability, salt spray, temperature | The qualification authority (the specification owner, a QPL listing, or the customer’s own spec) | A part built outside its qualified envelope — and on a QPL part, outside its listing |
| Purchase order | Flow-downs: FAI, source inspection, lot traceability, approved sub-tier processors, packaging, record retention | The buyer | A contractual nonconformance even when the part works perfectly |
The last two rows are where good hardware gets grounded. A lot can pass every electrical test and still be nonconforming because the plating certificate names a class you did not order, because a special process ran at an unapproved sub-tier, or because the finish letter on the label does not match the finish letter in the part number. Those are identity and evidence failures, not performance failures — and nobody catches them by testing the part.
| Disposition | What it means | Conforming afterwards? | Typical authority | Record produced |
|---|---|---|---|---|
| Rework | Restore full conformity to the drawing and specification using a process approved before the event | Yes — the part is conforming again | Supplier quality / process engineering | Rework record plus full re-inspection of the affected characteristics |
| Repair | Restore the part to a condition acceptable for use, in a way that does not conform to the drawing | No — permanently nonconforming but accepted | Design authority; customer where the contract requires | Repair procedure, repair record, and usually a concession |
| Use-as-is (concession / waiver) | Accept the part exactly as built, with no change | No — permanently nonconforming but accepted | Design authority plus customer approval | Concession tied to a named lot and quantity |
| Scrap | Destroy so the part cannot enter service by any route | Not applicable | Supplier quality | Scrap record with quantity and method of mutilation |
| Return to supplier | Reject back to the maker or sub-tier | Not applicable | Buyer quality / purchasing | Nonconformance report to the supplier, and a SCAR |
Sort-and-screen is not a disposition. Inspecting 100 % and shipping the good ones is a containment action. The bad ones still need one of the five answers above, and the screened population still needs a statement of what was measured and against which limit.
The practical line between rework and repair: rework is a process, repair is a decision. If the procedure already exists in your routing and has been qualified, it is rework. If you are inventing the procedure while holding the part, you are in repair territory and you need a second signature.
On a machined bracket, a repair is often genuinely harmless. On a qualified connector it usually is not, because what was qualified is not the shape of the metal — it is the process chain that produced the surface. The salt-spray hours, the mating-cycle count and the altitude dielectric figures on the datasheet are claims about a specified alloy, plated by a qualified process, at an approved line, with an insert moulded or bonded to a specified recipe.
So a “small local fix” can quietly remove the only thing that made the part qualified:
The honest framing: on a connector, most repairs are re-qualification questions wearing a maintenance hat. Price them that way.
| Nonconformity | How it is normally found | Realistic dispositions | Why the alternatives are closed |
|---|---|---|---|
| Plating thickness below the drawing minimum | XRF or coulometric on a coupon or first article | Scrap, or concession with customer approval | Re-plating is a new process cycle needing a full re-qualification; touch-up is not the qualified finish |
| Plating broken through to base metal on a sealing land, thread or EMI ground path | Visual at 10× | Scrap, unless the drawing names a touch-up process and a permitted zone | Sealing and grounding are functional, not cosmetic |
| Cosmetic scratch outside sealing lands, threads and grounding paths | Visual | Use-as-is with a drawing note | Nothing functional is affected — this is a documentation exercise, not an engineering one |
| Contact retention below the specified load | Axial pull test per the product specification | Replace the contact and re-test; scrap if the cavity or clip is not serviceable | Retention is a drawing characteristic; a “tighter” clip is a different part |
| Bent contact, or splayed socket tines | Visual, go/no-go gauge | Replace the contact; scrap if the contact is not removable | Bending back work-hardens the spring and re-forms the mating geometry |
| Cracked or chipped insert / dielectric | Visual, DWV, IR | Scrap | No approved repair exists; a crack is a creepage path that grows with thermal cycling |
| Damaged key or keyway / master key | Visual, mating trial | Scrap | Keying is a cross-mating safety function; a dressed key defeats the only barrier separating otherwise identical connectors |
| Shell thread damage | Go/no-go thread gauge | Dress inside the minor-diameter limit then re-gauge; otherwise scrap | The thread carries the coupling preload and the seal compression |
| DWV failure at sea level | Hipot at sea level and altitude | Rework: clean, dry, re-seal, retest; scrap if it recurs after bake-out | Most sea-level DWV failures are moisture; a failure that survives dry-out is a material or void problem |
| Insulation resistance below limit at 25 °C | Megohmmeter at 500 VDC | Dry-out and retest, then the DWV rule above applies | Same physics as DWV — the film is usually water, not a flaw |
| Low-level contact resistance above limit | 4-wire Kelvin at 100 mA maximum | Clean and re-measure; if it persists, re-terminate or replace the contact | Contact resistance is an asperity-scale property — cleaning can restore it, crimping never does |
| Shell-to-shell resistance above the EMI budget (typically <2.5 mΩ) | Milliohm meter per MIL-STD-1344 Method 3007 | Usually the accessory: replace or re-torque the backshell, then re-measure | The ground path runs through the accessory threads — the connector itself is often innocent |
| Hermetic leak rate above limit | Helium mass spectrometer | Scrap | A glass-to-metal seal is a furnace process; there is no bench repair |
| Mixed lots, or traceability that cannot be reconstructed | Document audit | Sort and re-identify to lot where the records exist; otherwise scrap or return | Traceability is a deliverable, not an attribute of the metal |
Two rows are worth memorising because they are misread constantly. Thread damage looks cosmetic and is not — the thread sets the coupling torque envelope and therefore the interfacial seal load. High shell-to-shell resistance looks like a connector problem and usually is not — the shell plating is fine and the accessory interface, its torque or its conical ferrule is the fault.
The material review board is a decision body with named members and a defined scope. Its authority is narrower than most people assume.
| Decision | Typical authority | Customer approval? | Notes |
|---|---|---|---|
| Rework using an already-approved, already-qualified process | Supplier quality / process engineering | Only if the contract requires notification | This is why the process must exist before you need it |
| Repair — any non-standard restoration | Design authority | Yes on flight, safety-critical and most aerospace contracts | The procedure must be documented and repeatable, not improvised on the bench |
| Use-as-is / concession | Design authority plus customer | Yes | Tied to a named lot and quantity — it expires with them |
| Scrap | Supplier quality | No | Physical mutilation, not a red tag in a locked cage |
| Return to supplier or sub-tier | Buyer quality / purchasing | No | Followed by a formal corrective action request, not an email |
| Change a specification limit | Nobody at the MRB | Qualification authority | The MRB has no authority here at all — see below |
An MRB cannot vote a specification limit away. A drawing dimension and a product-specification limit belong to the design and qualification authority; the board can only decide what happens to the parts. If the fix you want is “accept a few tenths of a thousandth under the cadmium minimum”, that is not a disposition — it is a design change or a deviation request, and it goes to people who are not in the room.
QPL-listed parts deserve their own sentence. A qualified products list qualifies a product to a specification. A part built outside that specification is outside its listing until it is re-qualified, whatever the concession document says. A buyer accepting a concession on a QPL part should record, in writing, that the accepted part sits outside the listing — otherwise the acceptance is invisible to the next programme that inherits the hardware.
| Term | Timing | What it covers | Granted by | Lifetime |
|---|---|---|---|---|
| Deviation | Before production | “We intend to build it this way” — a defined departure for a defined quantity or period | Design authority / customer | Expires at the stated quantity or date |
| Waiver | After the fact | “This already-built lot is acceptable” | Design authority / customer | One lot, one quantity |
| Concession | After the fact | Acceptance of nonconforming material; often the same act as a waiver, and the usual term in European practice | Design authority / customer | One lot |
| Engineering change | Permanent | Changes the drawing or specification itself | Design authority | Until superseded |
The rule that keeps these honest: never let a deviation become the de facto drawing. If the same deviation is requested for the third time, the drawing is wrong and the deviation is hiding a design problem. That is a finding, not a convenience.
Before anyone argues about root cause, the escape has to be stopped. Containment has three levels, and most suppliers only do the first:
For a connector, containment is fundamentally a traceability query. Which lot codes are affected? Which date-code range? Which plating bath, which heat, which operator, which test record, which shipment? A plant that can answer that in an hour contains a problem in a day; a plant that cannot answer it is exposed to the whole year’s output. This is the operational reason traceability is specified as a deliverable in the first place — the lot code has to resolve to raw material heats, plating chemistry logs, operator sign-offs and test results, and those records have to outlive the warranty.
The 24-hour expectation is real: an interim containment action within one working day of a confirmed escape, before the root cause is known. For connectors the credible interim action is 100 % inspection of the affected characteristic with actual measured values attached to each lot. “We will be more careful” is not a containment action and should be rejected as one.
The most common defect in a supplier’s corrective action is that it fixes only half the problem. Every nonconformity has two causes:
| Occurrence root cause | Escape root cause | |
|---|---|---|
| The question | Why did the process produce this defect? | Why did our detection system let it leave? |
| Typical answers | Tool wear, plating bath chemistry drift, a setup change, material lot variation, an operator step performed in the wrong order | Gauge not capable at that resolution, sample too small, acceptance criterion transcribed from the wrong revision, characteristic absent from the inspection plan, inspector measured a different feature |
| Who feels it | The supplier | The customer |
| Fix without the other | The same defect returns from a different cause | This cause is fixed, but every future defect of any kind still escapes |
A worked example. Low-level contact resistance drifts above limit. The occurrence cause is plating bath chemistry moving outside its control window. The escape cause is that lot acceptance measured contact resistance on five pieces with a handheld multimeter, which cannot resolve a few milliohms at all — the instrument’s own lead resistance swamps the reading. Fixing the bath closes this event. Fixing the measurement closes the class of events, because the acceptance method itself was never capable of the specification.
Corrective action formats differ mainly in how hard they force that second path. A five-why chain is fast and routinely stops at the occurrence cause. A full 8D disciplines the structure and adds containment, but still needs the reviewer to demand the escape branch explicitly. A SCAR is the buyer-side wrapper around either — its value is the closure criterion, not the template. Whichever format is used, the demand should be the same: two causes, two corrective actions, two verifications.
Corrective actions are not closed by actions. They are closed by evidence that the actions worked. The following are activities, not proof: operator retraining, a toolbox meeting, a revised work instruction, an added inspection step with no data behind it.
What does count:
Verification is a finite test with an agreed pass criterion. A verification with no end date is not verification; it is an open action item that everyone has learned to ignore.
| Metric | What it reveals | How it is usually gamed |
|---|---|---|
| Escape rate (defects found by the customer) | Whether detection works at all | Counting units instead of events — one escape of 4,000 parts is one event and 4,000 units |
| DPPM to the customer | The customer’s actual experience | Reclassifying escapes as “documentation issues” so they leave the numerator |
| MRB rate per 1,000 units | Process stability and specification realism | Pushing disposition back onto the operator so it never reaches the board |
| Scrap rate and rework hours | Cost of poor quality, and where it is generated | Rework hours booked to production instead of quality |
| Repeat nonconformance rate | Whether corrective actions actually closed | Opening a new nonconformance number for a recurrence |
| Corrective-action cycle time | Responsiveness — and whether verifications ever finish | Closing on “action taken” rather than “effective” |
The metric that lies most often is the MRB count. A supplier with a falling MRB rate and a rising escape rate is not improving — it is moving the discovery downstream, where it costs the customer instead. A rising MRB rate alongside flat escapes can be the opposite: better detection catching more before shipment. Read the pair, never either one alone.
If an auditor, a customer or a lawyer picks up one nonconformance three years from now, the record needs to answer these without help: part number including the finish and keying letters and drawing revision; lot, date code or serial range; quantity affected and quantity inspected; the requirement violated, cited to a drawing zone, a specification paragraph or a PO clause; the measured value against the limit; where and how it was detected, with the gauge or instrument identified; the containment actions and their scope; the disposition and its justification; the authorising person, role and date, plus the customer approval reference where one was needed; the re-inspection or re-test results after rework; the corrective action reference and its verification result; and the effectivity statement — which lots before and after this one are clean.
Retention is set by contract, not by preference. Aerospace flow-downs commonly run to seven or ten years and longer for shelf-life-limited items; records that cannot be produced are treated as records that never existed.
Production is 100 % electrically tested rather than sampled, so acceptance data exists for every unit — continuity, insulation resistance, dielectric withstanding voltage and shell-to-shell resistance — and each connector carries a lot code that resolves to raw material heats, plating bath chemistry logs, the assembly operator, the test results and the shipping date. Those records are retained for well over a decade, which is what makes a containment query answerable in hours rather than weeks.
The families most often specified with a disposition and corrective-action flow-down are the J599 Series III equivalents of MIL-DTL-38999, the ZH83723 bayonet and threaded range, MIL-DTL-26500-class aerospace circulars, and the M85049 accessory and backshell range, alongside M39029 contact spares and the wider circular connector and US military standard catalogue. The acceptance basis those dispositions are measured against is published on our quality page.
Need a nonconformance dispositioned, a corrective action closed, or the acceptance basis for a lot you already hold?
Send the part number with its finish and keying letters, the lot or date code, the requirement you believe was missed and the measurement you took. We will answer with the disposition route, the authority required and the records that support it.
Email: info@airoadcon.com | WeChat: 18991927716