How Should You Inspect a Small Batch After the Prototype Is Approved?
After a prototype is approved, inspect the small batch with a written control plan—not by assuming every copy will match the prototype. Identify the characteristics that affect fit, function, safety, assembly, and appearance; verify the first production part before the run continues; monitor the process while parts are being made; and define final acceptance evidence. Use 100% inspection for characteristics whose failure cannot be accepted or whose process is not yet proven. Use sampling only when the lot, risk, method, and acceptance rule are agreed in advance.
A prototype answers, “Can this design work?” A batch plan answers, “Can this process repeatedly produce acceptable parts?” Those are different questions. This guide explains what changes between prototype approval and a run of 10, 20, or 100 parts, without inventing one universal sample size.
1. Freeze what “prototype approved” actually means
Approval should refer to a controlled design revision and a recorded result. A physical prototype alone is not a complete specification: it may contain hand fitting, a repaired surface, a different material lot, or a one-time adjustment that was never added to the drawing. Before batch production, record the model and drawing revision, material, finish, critical dimensions, approved deviations, test method, and any assembly components used for the decision.
Keep an approved sample when visual texture, color, edge break, tactile feel, or assembly behavior is difficult to communicate numerically. The sample supports the drawing; it does not replace it. Store it so it cannot be mixed with production parts or damaged through routine handling. If the prototype was accepted with a concession, state whether that concession applies to the batch. Silence is not a reliable acceptance rule.
Also separate design approval from process approval. A CNC prototype made from billet may not predict a later molded part. A 3D-printed prototype may confirm envelope and assembly but not the stiffness, surface, or dimensional behavior of another material or process. When the production route changes, repeat the characteristics affected by that change instead of treating the prototype as universal evidence.
2. Turn product requirements into an inspection plan
Start with function, then classify characteristics by consequence. A critical sealing diameter, connector location, load-bearing wall, or interference risk deserves more control than a hidden nonfunctional surface. ASME Y14.5 provides a common language for dimensions, tolerances, datum reference frames, and geometric controls so design intent can be interpreted consistently. The drawing still needs sensible requirements: inspection cannot rescue an ambiguous model or an unnecessarily tight blanket tolerance.
| Characteristic | Possible evidence | Typical control decision |
|---|---|---|
| Safety or irreversible failure risk | Defined test or measurement with traceable result | Often 100% until a capable process and formal plan justify otherwise |
| Assembly interface or critical fit | Dimension, geometric control, or functional gauge | First-off plus in-process monitoring; final frequency based on risk |
| Functional performance | Leak, load, movement, electrical, or fit test | Use a repeatable fixture and a clear pass/fail threshold |
| Cosmetic surface | Approved sample and controlled viewing conditions | Define defect classes and visible zones before production |
| Noncritical hidden feature | Visual or sampled dimensional check | Avoid inspecting more than the requirement and risk justify |
For each controlled characteristic, state the method, instrument or fixture, inspection stage, frequency, responsibility, record, and reaction when a result fails. “Check dimensions” is not enough. “Measure the mounting-hole pattern from datums A, B, and C on the first part and every setup change; stop production if any position result exceeds the drawing limit” is actionable.
3. Use first-off, in-process, and final checks for different purposes
First-off inspection confirms that the released file, material, tool offsets, setup, program, and fixture can produce the intended part. It should happen before the complete batch is made. Check the characteristics most likely to be affected by the setup, including datum relationships, tool-dependent features, wall thickness, holes, threads, and assembly interfaces. If the first part fails, quarantine it, correct the process, and inspect a new first-off part.
In-process inspection detects drift before it affects the rest of the lot. The frequency should reflect cycle count, tool wear, material behavior, setup stability, and consequence of failure. It may be triggered by a tool change, spool or resin change, machine restart, refixture, operator change, or an unusual event—not only by a fixed count. Record enough context to connect a result to the production sequence.
Final inspection confirms lot acceptance and completion of secondary operations. It should verify that deburring, cleaning, coating, marking, packaging, and assembly did not damage or change controlled features. Final inspection is not the ideal place to discover a systematic error that began with the first part; that is why first-off and in-process controls matter.
A functional gauge can be valuable when it represents the real mating condition and is itself controlled. A go/no-go fixture may check a hole pattern quickly, but it should not hide which requirement it represents. Define fixture revision, maintenance, verification, insertion force, orientation, and pass/fail method so operators do not “make the part fit.”
4. Choose 100% inspection or sampling from risk—not convenience
Inspecting every unit can be appropriate for critical characteristics, unstable processes, very small lots, automated tests, or inexpensive checks. It is not automatically perfect: a weak measurement method can produce repeated wrong decisions, and human visual inspection can lose consistency with fatigue. Improve the method as well as the frequency.
Acceptance sampling evaluates a lot using a defined sample and acceptance rule. ISO 2859-1:2026 provides AQL-indexed schemes for lot-by-lot inspection by attributes, including switching rules. It should be applied as a complete agreed system, not reduced to an arbitrary statement such as “inspect three pieces.” AQL is not a promise that a particular lot contains no defects, and it does not replace control of the production process.
Before using a sampling plan, define the lot boundary, random-selection method, defect classes, inspection level or other governing inputs, acceptance and rejection numbers, treatment of rejected lots, and whether reinspection is allowed. Customer and supplier should agree which standard and edition govern. If the batch has only 10 parts and one failed critical feature would stop assembly, 100% inspection of that feature may be clearer and cheaper than constructing a formal sampling argument.
5. Make sure the measurement system is fit for the decision
A result is useful only when the method can distinguish acceptable from unacceptable parts with adequate confidence. ISO 10012:2026 addresses measurement management systems intended to support valid and reliable results. NIST’s measurement-process guidance discusses repeatability, reproducibility, stability, calibration, and uncertainty. These concepts matter even in a small workshop: two operators should not reach opposite decisions simply because the fixture, force, environment, or method was uncontrolled.
Select equipment for the feature and tolerance, not for convenience. A functional gauge may be best for assembly acceptance; a height gauge may suit accessible location measurements; a CMM may be appropriate for geometric relationships; and a dedicated test fixture may be needed for load, leak, or motion. Record instrument or fixture identity, calibration or verification status, environmental limits where relevant, method revision, result, operator, date, part or lot identity, and disposition.
Plan the reaction before a failure occurs. Stop or contain production, identify the last known acceptable point, segregate suspect parts, investigate the cause, correct the process, and define reinspection. Do not quietly adjust a result, mix inspected and uninspected units, or repair parts without authorization. A concise nonconformance record is more valuable than a perfect-looking spreadsheet that cannot trace what happened.
Frequently asked questions
If the prototype passed, why inspect the first batch?
The prototype proves one build or one configuration met the approval criteria. A batch introduces repeated cycles, tool wear, material variation, multiple setups, secondary processing, and handling. First-off and in-process checks confirm that the production process repeats the approved intent.
How many parts should be inspected in a small batch?
There is no responsible universal number. The decision depends on lot size, feature risk, process history, measurement capability, customer requirements, and the sampling system being used. Critical or unstable characteristics may need 100% inspection even when other features are sampled.
Is AQL the same as an allowed defect percentage in each lot?
No. AQL indexes an acceptance-sampling system and its operating rules; it is not a guarantee that each accepted lot contains that exact defect percentage. Use the governing standard, inputs, random sampling, acceptance numbers, and rejected-lot procedure as an agreed system.
What should I send with a small-batch manufacturing request?
Send the released STEP model and drawing, revision, material, quantity, critical characteristics, mating parts or functional test description, cosmetic standard, inspection frequency, report requirements, and the action required when a result fails.
For a part that may use different manufacturing routes, MALIEV can review the production package and inspection expectations through its custom manufacturing service. Start with the controlled files and the few characteristics that determine whether the part truly works.