Six Precision Parts Purchasing Myths That Can Increase Assembly Risk and Total Cost

A shipment arrives with a dimensional report, an ISO 9001 certificate, and confirmation that every part was inspected. Purchasing releases the batch. Assembly then discovers that shafts bind, mounting holes need adjustment, or components fit only after selective matching.

The immediate question is usually whether the supplier manufactured the parts incorrectly. That question matters, but it does not cover every possible cause.

Two situations require different responses. Parts may satisfy every agreed requirement while the assembly still fails because the specifications do not fully describe the functional interface. Alternatively, inspection records may show acceptance even though the delivered parts fail the actual requirements.

Global buyers need to distinguish these situations before demanding tighter tolerances, additional inspection, or replacement production.

Six familiar purchasing claims often blur that distinction: tighter tolerances are better; CMM reports guarantee fit; ISO 9001 guarantees every part; complete inspection eliminates risk; approved samples prove production consistency; and the lowest quotation delivers the lowest cost.

Each claim contains something useful. Problems begin when buyers treat one piece of evidence as proof of the entire manufacturing and assembly process.

The numerical examples below are illustrative calculations from the supplied research, rather than industry averages or documented supplier results.

1. Functional Tolerances: Why Tighter Dimensions Do Not Automatically Improve Fit

Functional tolerancing refers to defining the dimensional and geometric limits required for components to assemble and operate under specified conditions.

A tolerance becomes valuable when it controls something that matters to function. A smaller number alone does not establish that connection.

ASME Y14.5 provides rules for communicating dimensions, geometric relationships, and datum references. Its purpose includes expressing the design intent needed for fit, function, and interchangeability. A drawing therefore needs more than a collection of narrow plus-or-minus limits. [ASME Y14.5] ASME

Consider a shaft-and-hole interface requiring diametral clearance between 0.010 mm and 0.030 mm. The initial dimensions are:

The minimum clearance is:

9.990 − 10.010 = −0.020 mm

The maximum clearance is:

10.010 − 9.990 = 0.020 mm

The permitted combinations include interference. Both components can satisfy their individual size limits while failing the intended clearance requirement. [Supplied Research—Illustrative Clearance Calculation]

Suppose purchasing asks the supplier to tighten both tolerances to ±0.005 mm, leaving the nominal diameters unchanged. Clearance now ranges from −0.010 mm to +0.010 mm. Manufacturing becomes more demanding, but the required minimum clearance remains unprotected. [Supplied Research—Illustrative Clearance Calculation]

Changing the hole to 10.020 ± 0.005 mm while keeping the shaft at 10.000 ± 0.005 mm produces the intended 0.010–0.030 mm clearance range. The improvement comes from correcting the nominal relationship and allocating tolerances appropriately. [Supplied Research—Illustrative Clearance Calculation]

When I recalculated the shaft-and-hole example supplied for this article, I observed that halving both original tolerances still left a possible interference condition. Shifting the hole’s nominal diameter changed the result: the worst-case clearance then matched the stated requirement. This was a calculation review, rather than a physical assembly experiment, but it demonstrates why I would check the required fit before requesting a smaller tolerance.

That example addresses diameter only. Real assemblies also depend on form, alignment, surface condition, temperature, and loading. A correctly sized bore can still create assembly difficulty if its axis is misaligned with another locating feature.

Surface treatment creates another purchasing question: does the drawing apply before or after the final treatment? An acceptable machined dimension may change during subsequent processing. Buyers should establish the required delivery condition before comparing quotations.

The opportunity is to spend precision where it produces a functional benefit. Some critical interfaces may need tighter control, while unrelated dimensions can remain less restrictive. This can reduce unnecessary machining and inspection effort without weakening the assembly requirement.

Action: Have engineering identify the required fit, locating features, and final surface condition before negotiating tighter dimensional limits.

2. Measurement Evidence: Why a CMM Report and Complete Inspection Still Have Limits

Measurement evidence refers to the results, methods, and decision rules used to determine whether specified product characteristics meet agreed requirements.

A CMM report can provide strong evidence. Its usefulness depends on what was measured and how the measurement relates to the assembly.

Buyers should ask whether the inspection alignment represents the way the part is located in service. A report based on a convenient manufacturing surface may answer a different question from an assembly located by a mounting face and locating holes.

Measurement coverage also matters. A reported diameter does not automatically establish adequate form throughout a bore. A hole-position result does not establish thread engagement, sealing performance, cleanliness, or acceptable insertion force.

NIST research on small circular features found that changing the angular distribution of three measurement points could produce uncertainty differences spanning four orders of magnitude under the study’s conditions. That finding is specific to the investigated measurement arrangements; it is not a general performance claim about all CMM inspection. [NIST—Small Circular Feature Measurement Uncertainty] NIST

The practical implication is straightforward: owning a capable machine does not establish that every measurement program is suitable.

Borderline results need particular care. Suppose a feature has an upper limit of 20.000 mm, a measured value of 19.998 mm, and expanded measurement uncertainty of 0.004 mm.

Under an agreed acceptance rule requiring the measurement interval to remain entirely within specification:

19.998 + 0.004 = 20.002 mm

The interval crosses the upper limit. That result does not demonstrate conformity under this rule, although it also does not independently prove that the physical feature exceeds the limit. [Supplied Research—Illustrative Measurement Decision Calculation]

ISO 14253-1 addresses decision rules for verifying conformity or nonconformity while accounting for measurement uncertainty. Buyers and suppliers should establish the applicable rule before a disputed borderline result occurs. [ISO 14253-1] iso.org

“100% inspection” introduces a related misunderstanding. It may mean every component received one check. It may mean every listed characteristic was measured. Those are substantially different inspection scopes.

Even complete part coverage cannot compensate for a method that misses the relevant defect.

Assume an illustrative batch contains 10,000 parts, including 200 defective parts, and the inspection method detects 95% of those defects. Expected escapes equal:

200 × 5% = 10 parts

That is 1,000 ppm relative to the original batch, despite every component passing through inspection. The assumed detection rate is an example, rather than a benchmark for human inspectors or measurement equipment. [Supplied Research—Illustrative Inspection Escape Calculation]

Complete inspection remains useful for containing a known problem. Functional gauges and controlled assembly tests can also strengthen verification where several features interact. However, each method needs defined operating conditions, maintenance, and acceptance criteria.

A sensible inspection package links each critical requirement to an appropriate method. It also identifies the inspected production stage, relevant datum references, report traceability, and the procedure for resolving uncertain results.

Action: Approve an inspection plan that states what is checked, how it is checked, and how borderline results are decided before accepting a promise of complete inspection.

3. Supplier Qualification: What ISO 9001 Can Establish About Your Order

Supplier qualification refers to evaluating whether an organization has the management systems, technical resources, and project controls needed to meet a defined purchasing requirement.

ISO 9001 certification contributes to that evaluation. It supports confidence in the supplier’s quality management system, but it does not certify the technical conformity of each delivered component.

The ISO 9001 Auditing Practices Group explains that the standard defines management-system requirements rather than product requirements. Its guidance also states that accredited certification does not necessarily ensure that an organization always achieves 100% product conformity. [ISO/IAF—Expected Outcomes for Accredited Certification] committee.iso.org

That distinction should shape supplier approval.

First, establish whether the certificate corresponds to the legal organization and production location involved in the order. Review its scope and status. A certificate associated with a different site or unrelated activity may provide limited evidence for the proposed work.

Then examine the project itself. The relevant questions concern drawing interpretation, manufacturing sequence, outsourced processes, measurement capability, traceability, and change control.

For example, a supplier may have an effective quality system while still lacking experience with a particular thin-wall component. The project may require a new fixture, additional process development, or verification after unclamping. Certification does not remove those engineering tasks.

A simple hypothetical calculation shows why buyers should avoid translating a certificate into a defect-rate promise. If an actual process has a 0.1% defect rate, a shipment of 10,000 parts has an expected 10 defective parts before any additional screening. These figures are assumptions; certification supplies no numerical conversion to this defect rate. [Supplied Research—Illustrative Defect Expectation]

Equally, one defective shipment does not by itself establish that the supplier’s certification is invalid. The investigation should determine what failed, whether the controls were followed, and whether corrective action prevents recurrence.

The opportunity lies in using the management system to support project discipline. Controlled revisions, retained inspection records, and effective corrective action can make production problems easier to trace and resolve.

Buyers can test that discipline with practical requests. Ask how the supplier prevents an obsolete drawing from reaching the machine. Ask how externally processed parts retain their batch identity. Ask what triggers renewed approval after a material, fixture, or process change.

The answers should connect procedures to actual records and responsibilities. A polished quality manual has limited purchasing value if nobody can explain how its controls apply to the component being ordered.

Qualification should also match consequence. A noncritical bracket and a component whose failure stops an expensive assembly line may justify different levels of project review, even when both suppliers hold the same certification.

Action: Use ISO 9001 certification as one qualification input, then approve the order against project-specific manufacturing, verification, and change-control evidence.

4. Production Validation: Why Approved Samples Do Not Prove Repeatability

Production validation refers to demonstrating that the intended manufacturing process can repeatedly deliver acceptable parts across relevant operating conditions.

A successful sample establishes that those particular components worked under the conditions tested. It does not establish how production will behave after tool wear, a setup change, another material batch, or a different operator.

The distinction becomes especially important when samples receive extra attention. Hand finishing, selective matching, and unusually slow machining may produce an excellent approval set without representing the quoted production route.

Buyers should therefore establish how samples were made and whether the same method will be used for recurring orders.

Small sample sizes can also create false confidence. Assume independent random sampling from a process with a true defect rate of 1%. The probability that 30 parts all pass is:

0.99³⁰ ≈ 73.97%

A process with a meaningful defect rate can therefore produce a completely successful small sample quite easily. [Supplied Research—Illustrative Binomial Calculation]

Under an independent, identically distributed binomial model, zero failures among 30 samples gives an exact one-sided 95% upper confidence bound of approximately 9.50% for the defect probability. With 300 samples and zero failures, the bound falls to approximately 0.994%. These are calculated statistical limits, not universal sample-size requirements. [Supplied Research—Illustrative Binomial Confidence Calculation]

Those calculations are useful only when their assumptions are reasonable. Selected samples, correlated observations, or a process changing over time can make a simple binomial interpretation inappropriate.

Process stability is a separate requirement. NIST’s guidance treats stability as a foundation for assessing process performance over time. A single successful trial cannot reveal whether the process mean or variability will shift during production. [NIST—Assessing Process Stability] itl.nist.gov

Capability analysis then compares process variation with specification limits. NIST explains capability indices in the context of an in-control process and identifies distributional assumptions relevant to commonly used indices. [NIST—Process Capability] itl.nist.gov

Consider a dimension specified as 10.000 ± 0.030 mm, assuming a centered, stable, normally distributed process:

The specification remains unchanged. Reducing variation improves the calculated result. These figures describe a mathematical normal-distribution model; they are not demonstrated long-term assembly defect rates. [Supplied Research—Illustrative Capability Calculation]

That creates a useful alternative to blanket tolerance tightening. Tool management, fixture improvement, thermal control, and consistent setups may address the actual source of variation.

For purchasing, the strongest validation evidence combines representative production conditions with traceable measurements and appropriate functional checks. Data should retain enough sequence information to reveal drift, rather than presenting only a sorted list of acceptable values.

Where interchangeability matters, testing should also consider the permitted variation in mating components. Parts that assemble only with one selected counterpart may conceal a problem that appears when production batches are mixed.

Action: Validate samples made through the intended production route and review performance across relevant process changes before approving recurring supply.

5. Total Purchasing Cost: Why the Cheapest Quotation Can Produce the Most Expensive Parts

Total purchasing cost refers to the combined expenditure required to obtain an agreed quantity of usable parts, including acquisition, verification, disruption, and recovery costs.

Unit price remains important. The mistake is comparing quotations as though every supplier creates the same downstream workload and risk.

ASQ’s cost-of-quality framework distinguishes prevention, appraisal, internal failure, and external failure costs. This provides a useful foundation for examining what happens beyond the purchase order. [ASQ—Cost of Quality] ASQ

For international sourcing, a purchasing model may also include normal freight, customs costs, additional inventory, and emergency transport. These extensions should reflect the buyer’s actual operating conditions.

The comparison must use equivalent outcomes. A supplier delivering fewer usable components cannot be compared fairly with one delivering the complete required quantity unless replacement and recovery costs are included.

The following hypothetical completed-order comparison assumes both suppliers ultimately provide 10,000 usable parts. All figures are in US dollars. Replacement-related incremental costs are listed separately. [Supplied Research—Illustrative Total Cost Comparison]

Supplier B’s purchase price is 8% higher, but its total cost is $17,000 lower, a reduction of approximately 12.1% against Supplier A. The higher purchase price is outweighed by the assumed difference in other net costs. [Supplied Research—Illustrative Total Cost Comparison]

This example does not establish that expensive suppliers perform better. It demonstrates which evidence is needed to justify paying more: credible differences in usable output, verification effort, and failure exposure.

Cost models also need discipline. If delayed production is recovered through overtime, charging both the entire missed sales value and all recovery costs may exaggerate the loss. If regular freight is already included, emergency transport should capture the relevant incremental expense.

Supplier reimbursement needs similar care. A submitted claim is not cash recovered. Completed-order analysis should deduct compensation actually received; a forecast can separately account for uncertain recovery.

Improvement spending can be evaluated against the same model. At 100,000 parts annually, an additional $0.20 per part costs $20,000 per year. If validated changes reduce expected annual failure costs by $60,000, the remaining $40,000 can contribute toward one-time implementation costs. If avoided losses total only $5,000, the proposal needs a different justification. [Supplied Research—Illustrative Improvement Investment Calculation]

The right measure depends on the failure mechanism:

These are engineering decision options, rather than a mandatory sequence. Existing defects may require immediate containment while the permanent correction is developed.

Action: Compare suppliers by cost per usable part, using verified operating data and consistent assumptions about disruption, replacement, and recovery.

Preguntas frecuentes