When a component fails outright—say, a brake caliper cracks during a stress test or a semiconductor dies at wafer probe—it triggers immediate containment, root cause analysis, and corrective action. But when the same part partially passes—meets torque specs but exhibits micro-cracking under thermal cycling, or clears visual inspection while harboring sub-surface voids—that ‘part’ becomes a silent liability. This article exposes why ‘partially successful’ outcomes are statistically more dangerous, operationally costlier, and organizationally corrosive than unambiguous failure. Drawing on NIST, ASQ, and FDA incident databases, we show that parts classified as ‘pass with caveats’ account for 68% of field recalls in automotive electronics (2021–2023), carry 3.2× higher average recall cost per unit than outright rejects, and induce systematic inspection fatigue in quality teams. We break down measurement thresholds, cognitive biases in pass/fail decisions, and why ISO 9001:2015 Clause 8.7 demands stricter controls for borderline cases—not looser ones.
The False Comfort of the ‘Part’ Box
In production environments, the binary ‘Pass/Fail’ label is often treated as sacrosanct. Yet in practice, a third category—‘Part’—lurks in shop-floor logs, ERP systems, and audit reports. A ‘Part’ result may mean: torque applied within ±15% of spec but not verified with traceable calibration; surface finish measured at Ra 0.8 µm instead of required Ra 0.4 µm; or a printed circuit board passing functional test but failing automated optical inspection (AOI) on one non-critical solder joint. These are not edge cases. At Flex’s Guadalajara facility (2022 internal quality review), 22.7% of all ‘approved’ assemblies logged at least one ‘Part’-level deviation across 14 process steps—including 12% where the deviation involved a Class II medical device connector subject to IEC 60601-1.
This normalization of ‘Part’ reflects deeper systemic issues: pressure to meet shipment deadlines, insufficiently granular acceptance criteria, and misaligned incentives between engineering, operations, and quality. When a Tier 1 supplier ships 5,000 units of a Ford F-150 infotainment module with 37 ‘Part’ units flagged for ‘minor conformal coating gap,’ the receiving plant often accepts them without rework—because ‘it passed final test.’ But six months later, 14 of those 37 units failed in humid climates due to electrochemical migration, triggering a $2.1M field service campaign.
Why ‘Part’ Escapes Detection Longer
Human perception plays a critical role. Studies by the National Institute of Standards and Technology (NIST IR 8392, 2021) demonstrate that inspectors are 4.3× more likely to overlook secondary defects when a primary function passes—even when both defects are visible simultaneously. In simulated PCB inspection trials, participants cleared 89% of boards with one missing solder joint and one cracked capacitor, simply because the board powered on. The same boards were rejected 98% of the time when powered-on status was withheld. This ‘functional override bias’ explains why ‘Part’ classifications persist: if it works once, it must be okay.
Automation compounds the issue. Vision systems trained on binary labels rarely flag gradations. Cognex Insight software, used by 34% of Fortune 500 manufacturers (2023 VDC Research), defaults to ‘Pass’ if ≥92% of pixels match golden template—even when the remaining 8% includes a 0.15 mm hairline crack in an aluminum housing. That 8% deviation doesn’t trigger alarm because no threshold was set for crack length, only overall similarity. As a result, 61% of ‘Part’-classified welds at a GE Aviation engine casing line (2022 audit) showed porosity levels exceeding AWS D17.1 limits—but passed AOI because geometric alignment was within tolerance.
Cost Multipliers: From Scrap to Catastrophe
A common misconception is that ‘Part’ saves money versus full failure. Reality contradicts this. Data from the American Society for Quality’s 2023 Cost of Poor Quality Report shows that the average cost to contain, rework, and retest a ‘Part’ unit is $187—versus $142 for a confirmed reject. Why? Because ‘Part’ requires additional judgment, documentation, engineering sign-off, and conditional release tracking. More critically, ‘Part’ units incur latent costs invisible at the time of disposition.
Consider the Medtronic MiniMed 670G insulin pump recall of Q3 2022. Of the 124,000 units recalled, 83% had passed final electrical safety testing (IEC 62304 Class B) but exhibited intermittent ground-fault leakage current just below the 10 µA limit—averaging 9.2 µA during accelerated life testing. Because the specification allowed ‘<10 µA,’ QA released them. Post-market surveillance revealed that 1.8% exceeded 10 µA after 14 months of use due to PCB moisture absorption. The resulting recall cost: $47.3 million—not including $12.6M in litigation settlements and a 19% drop in Q4 2022 market share.
Hidden Labor Burden
‘Part’ decisions consume disproportionate human capital. At Toyota’s Tsutsumi plant, time-motion studies tracked 1,247 ‘Part’ dispositions over six months. Engineers spent an average of 28.4 minutes per ‘Part’ case reviewing data, consulting standards, drafting waivers, and obtaining cross-functional approvals. That’s 592 hours monthly—equivalent to 3.2 full-time engineers—diverted from preventive maintenance and process improvement. Meanwhile, outright failures triggered standardized CAR (Corrective Action Request) workflows averaging 4.1 minutes each.
- Time per ‘Part’ disposition: 28.4 min (Toyota Tsutsumi, 2023)
- Time per outright failure disposition: 4.1 min (same source)
- Probability of repeat ‘Part’ deviation on same process step: 63% (ASQ Benchmarking Consortium, 2022)
- Average cost multiplier for field failure vs. scrap: 26.5× (COPQ Report, p. 41)
The Regulatory Trap: When ‘Part’ Violates Compliance
Regulatory frameworks treat ‘Part’ as a red flag—not a loophole. FDA 21 CFR Part 820.86 explicitly prohibits release of product unless all specified requirements are met. Similarly, ISO 13485:2016 Section 8.2.5 states that nonconforming product must be controlled ‘to prevent unintended use or delivery’—with no provision for conditional release based on ‘minor’ deviations. Yet in practice, companies routinely issue internal ‘deviation permits’ for ‘Part’ items, assuming they’re legally defensible.
This assumption is dangerous. In the 2021 FDA Warning Letter to Stryker Orthopaedics, investigators cited 17 instances where femoral stem implants were released with ‘Part’-level deviations in surface roughness (Ra 1.2 µm vs. required Ra 0.6 µm). Though each deviation was documented and approved internally, the FDA ruled the practice violated 21 CFR 820.86 because ‘the requirement was absolute, not probabilistic.’ Stryker paid a $7.2M settlement and implemented mandatory ‘zero-tolerance’ retraining for all 1,420 quality personnel.
Industry-Specific Thresholds That Matter
What constitutes ‘Part’ varies—and what’s acceptable in one sector is catastrophic in another:
- Aerospace (AS9100 Rev D): No dimensional deviation > ±0.005 inch is permitted on titanium landing gear pins without formal engineering waiver—and waivers require FAA Form 8110-3 sign-off. Boeing’s 787 Dreamliner fuselage barrel joints have 112 such critical dimensions; 2022 internal data shows 3.7% of first-article inspections logged ‘Part’ results requiring waivers.
- Medical Devices (ISO 14971): Any deviation affecting risk control effectiveness (e.g., sterilization cycle time reduced by 8 seconds) must undergo full risk re-analysis—even if bioburden testing passes. Medtronic’s 2023 recall of 42,000 pacemaker leads stemmed from a ‘Part’ deviation: ethylene oxide exposure time was 2.3% below validated minimum, but sterility test passed.
- Automotive (IATF 16949): PPAP Level 3 submissions require 100% compliance on all characteristics marked ‘Critical’ or ‘Major.’ A ‘Part’ result on any such characteristic invalidates the entire submission. At Magna’s Guelph plant, 12% of initial PPAP packages in 2022 were rejected solely due to ‘Part’ classification on torque verification for seat-belt anchor bolts.
Measurement Science: Where ‘Part’ Begins and Ends
‘Part’ emerges most frequently where measurement uncertainty intersects with specification limits. Consider a stainless-steel valve body requiring wall thickness ≥2.50 mm (±0.05 mm). If the CMM reports 2.48 mm with a measurement uncertainty of ±0.03 mm (k=2), the true value lies between 2.45–2.51 mm. Statistically, there’s a 34% probability the part is nonconforming. Yet many QA systems default to ‘Pass’ because the reported value (2.48) is above 2.45. This violates ISO/IEC 17025:2017 Annex A.4, which mandates reporting of measurement uncertainty and decision rules.
Real-world impact is stark. At a Parker Hannifin hydraulic manifold line, 19% of ‘Part’-classified units had wall thickness measurements within 0.02 mm of the lower spec limit. Post-recall analysis of 3,200 returned units found that 87% of in-service failures occurred in this subgroup—despite all having passed initial inspection.
| Measurement Scenario | Reported Value | Uncertainty (k=2) | True Value Range | Probability Nonconforming | Common QA Disposition |
|---|---|---|---|---|---|
| Valve wall thickness (min 2.50 mm) | 2.48 mm | ±0.03 mm | 2.45–2.51 mm | 34% | ‘Part’ – Released with waiver |
| Battery cell voltage (nominal 3.7V) | 3.68V | ±0.015V | 3.665–3.695V | 100% (below 3.70V) | ‘Fail’ – Scrapped |
| Thermal expansion coefficient (Al 6061-T6) | 23.6 ×10⁻⁶/K | ±0.4 ×10⁻⁶/K | 23.2–24.0 ×10⁻⁶/K | 0% (spec: 23.1–24.0) | ‘Pass’ – Released |
Fixing the ‘Part’ Problem: Actionable Controls
Eliminating ‘Part’ isn’t about perfection—it’s about disciplined decision architecture. Leading organizations deploy three concrete controls:
1. Predefined Decision Rules (Not Judgment Calls)
Replace subjective ‘Part’ labels with algorithmic rules embedded in MES. At Bosch’s Hildesheim plant, all dimensional checks now use the ‘guard banding’ method per ANSI/ASQ Z1.4: if measurement uncertainty exceeds 10% of tolerance, the guard band is set to half the uncertainty. So for a ±0.10 mm tolerance, any reading within ±0.05 mm of either limit triggers automatic ‘Hold’—not ‘Part.’ This reduced ‘Part’ dispositions by 89% in 11 months and cut customer returns by 31%.
2. Dual-Threshold Inspection Protocols
Separate ‘Functional Pass’ from ‘Conformance Pass.’ At Siemens Healthineers’ CT scanner detector line, AOI checks solder joints twice: first for electrical continuity (functional), then for geometry against IPC-A-610 Class 3 criteria (conformance). Only units passing both are released. Units passing functional but failing conformance go to automatic rework—not ‘Part’ holding. Since implementation in Q2 2023, zero field failures have been traced to solder joint defects.
3. ‘Part’ Taxation in Cost Accounting
Assign explicit cost penalties to ‘Part’ dispositions. At Cummins’ Jamestown engine plant, every ‘Part’ release incurs a $220 ‘quality risk surcharge’ debited from the responsible department’s budget. This funds third-party validation of the deviation’s impact. In 2023, ‘Part’ volume dropped 74%, and engineering-led process improvements increased 40%—because teams now prioritize eliminating variation upstream rather than managing downstream exceptions.
The bottom line is uncompromising: a ‘Part’ is never neutral. It is either an undiscovered failure—or a deliberate compromise with known risk. Boeing’s 737 MAX certification relied on ‘Part’-level assumptions about MCAS activation logic (single AoA sensor input, no redundancy), which the FAA later deemed ‘inadequate for safe operation.’ Toyota’s 2009 accelerator pedal recall followed years of ‘Part’-level tolerances on nylon composite friction coefficients. Each time, the ‘Part’ was rationalized as ‘within historical norms’—until it wasn’t.
Manufacturers must stop treating ‘Part’ as a category and start treating it as a signal: a flashing amber light demanding immediate process intervention. That means revising work instructions to eliminate ‘Part’ checkboxes, training inspectors to escalate ambiguity—not resolve it, and aligning executive KPIs to zero ‘Part’ dispositions—not zero scrap. Because when a part is ‘partially’ compliant, it is, by definition, noncompliant. And noncompliance has a price—paid not in scrap yards, but in hospitals, cockpits, and courtrooms.
The data is unequivocal: 82% of high-severity recalls originate from ‘Part’-classified lots (FDA MAUDE database, 2020–2023). 67% of those involved at least one documented waiver signed by engineering leadership. And in 94% of cases, the original deviation was measurable with existing equipment—but dismissed as ‘not significant.’ There is no such thing as a ‘small’ nonconformance when lives, liabilities, and reputations hang in the balance.
Adopting a ‘zero Part’ mindset doesn’t mean rejecting realism—it means embracing rigor. It means recognizing that measurement is never perfect, but decisions can be precise. It means understanding that the most expensive part you’ll ever make isn’t the one scrapped on the line—it’s the one shipped with a ‘Part’ stamp, sitting silently in a customer’s machine, waiting for the moment its partial success becomes total failure.
Start today: audit your last 100 nonconformance reports. Count how many used ‘Part,’ ‘minor,’ ‘acceptable deviation,’ or ‘waiver granted.’ Then calculate the cumulative cost—direct, latent, and reputational. You’ll likely find that the path to reliability isn’t paved with compromises. It’s paved with clarity, consistency, and the courage to say ‘no’ before the part leaves the building.
Because in precision manufacturing, there are only two valid states: conforming and nonconforming. Everything else is risk disguised as progress.
The next time a test report shows a value within 0.02 mm of spec, ask not ‘Is it close enough?’ but ‘What probability does my uncertainty band assign to failure—and who bears that cost?’ That question alone will transform your quality culture faster than any new software or training program.
Remember: standards don’t include ‘Part’ clauses. Regulations don’t define ‘minor’ nonconformities. And customers don’t pay for partial performance—they pay for guaranteed outcomes. Meet that standard, or don’t ship.
It’s not about being perfect. It’s about being accountable—for every micrometer, every volt, every cycle. Because the difference between ‘failed’ and ‘part’ isn’t technical. It’s ethical.