CNC Inspection Sampling Plan: AQL, Sample Size and 100% Inspection

Inspecting every feature on every CNC machined part may be unnecessary, expensive, or impractical. Inspecting too little creates a different risk: a nonconforming lot may reach assembly before anyone discovers the problem.

A CNC inspection sampling plan turns that tradeoff into a defined decision process. It states how a lot is formed, how samples are selected, what characteristics are inspected, how defects are classified, and what results lead to acceptance or rejection.

This guide explains acceptance sampling, AQL, sample-size selection, acceptance and rejection numbers, random sampling, defect classification, switching rules, and when 100% inspection is the better choice.

What Is a CNC Inspection Sampling Plan?

A CNC inspection sampling plan is a documented method for deciding whether to accept a production lot after inspecting fewer than all units. The plan should identify:

• The product and lot covered by the inspection

• The applicable drawing and revision

• The number of units in the lot

• The sampling standard and revision

• The inspection level or plan type

• The selected sample size

• The characteristics or defects to inspect

• The acceptance and rejection criteria

• The method used to choose units randomly

• The action required if the sample does not meet the plan

Acceptance sampling makes a lot-level decision. It is not the same as process control, and it does not prove that every uninspected part conforms.

The NIST Engineering Statistics Handbook describes lot acceptance sampling as a middle approach between no inspection and 100% inspection: a random sample provides information used to decide the disposition of the lot.

Why CNC Parts Need a Feature-Based Inspection Strategy

A single machined part may contain dozens or hundreds of characteristics, but their risks are not equal. A cosmetic tool mark on a hidden face does not have the same consequence as an undersized bearing bore, a misplaced locating hole, a blocked fluid passage, or a burr that can detach inside a valve.

An effective inspection plan separates characteristics by function and risk. It may combine:

• First-off checks used to release a machining setup

• In-process checks used to control tool wear or drift

• Final sampling for stable, lower-risk characteristics

• 100% inspection for selected critical characteristics

• Periodic material, finish, cleanliness, or functional verification

• A first article inspection for a new or changed production process

See the related guide to CNC machining quality control for measurement equipment, reports, calibration, and traceability. For a new production configuration, the CNC first article inspection guide explains the difference between FAI, first-off inspection, and routine lot inspection.

What Is Acceptance Sampling?

Acceptance sampling examines a defined number of units selected from a lot. The results are compared with predetermined criteria to decide whether the lot is accepted or not accepted.

A simple attribute plan works like this:

1. Define the lot size.

2. Select the applicable sampling plan.

3. Determine the required sample size.

4. Choose sample units randomly from the lot.

5. Inspect the specified characteristics.

6. Count nonconforming units or nonconformities as defined by the plan.

7. Compare the count with the acceptance and rejection numbers.

8. Accept or reject the lot according to the stated rule.

The sample result is not permission to ignore known problems. If an inspector finds evidence of mixed revisions, wrong material, damaged packaging, an unstable process, or a safety-related defect outside the original sample definition, the issue should be contained and reviewed.

What Does AQL Mean?

AQL means acceptance quality limit in ISO sampling terminology. It is an indexing parameter used to select a sampling plan for a continuing series of lots. The AQL value works with the lot size, inspection level, plan type, and inspection severity to determine the sample and decision criteria.

AQL is often misunderstood. It does not mean:

• The supplier is allowed to ship that exact percentage of defective parts

• Every accepted lot contains defects at the AQL value

• The uninspected units are guaranteed to conform

• A sample result precisely measures the lot’s defect percentage

• Sampling replaces responsibility for conforming product

The current ISO 2859-1:2026 covers sampling schemes indexed by AQL for lot-by-lot inspection by attributes. When a buyer invokes the standard, the purchase order or quality plan should identify the edition and any customer-specific rules rather than relying on the phrase “standard AQL inspection.”

Attribute Sampling vs Variable Sampling

The data type changes how a plan works.

Attribute sampling

Attribute inspection classifies a unit or characteristic into categories such as conforming/nonconforming, present/absent, or defect/no defect. Examples include:

• Thread gauge passes or fails

• Required marking is present or missing

• Coating damage is visible or not visible

• A burr exceeds the defined acceptance limit or does not

• A unit contains one or more specified nonconformities

The result is a count. Attribute sampling is flexible and can cover many different defect types, but it does not show how close a measured value is to a specification limit.

Variable sampling

Variable inspection records a numerical measurement, such as bore diameter, flatness, surface roughness, or coating thickness. A variables plan uses the distribution of measured values and requires appropriate statistical assumptions and controls.

The American Society for Quality notes that variables plans can provide information about how close a process is to nominal or to a specification limit and may use smaller samples than a comparable attribute plan. They are also more complex and generally require a separate evaluation for each characteristic.

Do not convert measurement data into pass/fail results only because an attribute table is familiar. Where process behavior matters, actual values and capability analysis may provide more useful information.

Lot Size, Sample Size, and Inspection Level

These terms are related but not interchangeable.

Lot size

Lot size is the number of units grouped for the acceptance decision. A lot should be clearly identified and should represent a reasonably consistent production condition. Mixing different part revisions, materials, machines, suppliers, or process routes can make a sampling result misleading.

Sample size

Sample size is the number of units selected and inspected under the plan. It is not normally chosen as an arbitrary fixed percentage of the lot. Depending on the selected standard, the sample size is determined using the lot size, inspection level, plan, and other parameters.

Inspection level

The inspection level affects the relationship between lot size and sample size. A higher level generally provides more information through a larger sample, while a lower or special level may reduce the inspection burden. The correct level should come from the contract, quality plan, or risk decision—not from whichever option gives the smallest sample.

A statement such as “inspect 10%” is incomplete. It does not define the statistical plan, random-selection method, acceptance criteria, or response to failure. The same percentage can provide very different protection for different lot sizes and process risks.

Acceptance Number and Rejection Number

An attribute sampling plan normally provides an acceptance number, often shown as Ac, and a rejection number, often shown as Re.

• If the observed count is at or below Ac, the lot is accepted under the plan.

• If the observed count reaches Re, the lot is not accepted under the plan.

For example, a plan may require a defined sample and allow no more than its stated acceptance count. The actual sample size and Ac/Re values must come from the invoked standard or approved customer plan. They should not be copied from an online calculator without confirming the standard edition, inspection level, severity, lot type, and switching status.

An acceptance decision does not change a nonconforming sample into conforming product. Identified nonconforming units should still be controlled according to the agreed quality procedure.

How to Classify Defects for CNC Parts

Sampling plans often apply different criteria to different defect classes. The buyer should define the classes according to product function and risk.

Critical defects

A critical defect can create a safety hazard, violate a regulatory requirement, or cause another severe consequence defined by the buyer. Examples might include the wrong material in a load-bearing application, a blocked safety passage, or a feature that can cause hazardous assembly failure.

Critical characteristics often need stronger controls than ordinary acceptance sampling, such as error-proofing, verified material identity, process monitoring, traceability, functional testing, or 100% inspection. The exact control must follow the application and customer requirements.

Major defects

A major defect is likely to cause functional failure, prevent assembly, reduce expected performance, or create significant rework. Examples can include:

• A precision bore outside its tolerance

• Incorrect thread form or depth

• Hole position that prevents assembly

• Flatness or parallelism that affects alignment

• Missing coating or heat treatment

• Burrs on a sealing or fluid-control feature

Minor defects

A minor defect does not usually prevent intended use but departs from an agreed requirement. Examples may include limited cosmetic inconsistency on a non-functional surface or a marking issue that does not affect traceability.

Do not use generic examples as the final classification. The drawing, end use, and customer determine severity. A scratch can be minor on an internal bracket but major on an optical sealing face.

How to Select a Random Sample

Random selection is essential. Choosing only the easiest parts to reach can bias the result and hide variation.

Practical methods include:

• Assigning sequential numbers and using a random-number generator

• Selecting from multiple boxes, trays, or pallet positions

• Sampling across the beginning, middle, and end of the production run

• Including different cavities, fixtures, spindles, or machine positions where applicable

• Preventing operators from substituting specially selected “good” units

The lot should remain identifiable while samples are removed and returned. If parts have serial numbers or cavity identifiers, record them in the inspection report.

Convenience sampling is not random sampling. Inspecting the top layer of one box may miss a dimensional shift that occurred later in the run or a packaging problem confined to another container.

Normal, Tightened, and Reduced Inspection

Sampling systems for a continuing stream of lots may include switching rules. These rules change inspection severity in response to quality history.

Normal inspection

Normal inspection is typically the starting condition under an applicable switching system. It provides the baseline plan for the selected AQL and inspection level.

Tightened inspection

Repeated lot failures or evidence of deteriorating quality can trigger tightened inspection. The plan provides stronger protection until the required recovery conditions are met.

Reduced inspection

Sustained acceptable quality, stable production, and other specified conditions may permit reduced inspection. Reduced inspection is not simply a purchasing decision to check fewer parts; the applicable switching rules and authority should be defined.

Ignoring switching rules changes the protection provided by the sampling system. If a buyer selects only one convenient table for every lot, the procedure may no longer represent the cited standard.

When 100% Inspection Is Appropriate

One-hundred-percent inspection means checking every unit for specified characteristics. It may be appropriate when:

• A characteristic has serious safety or regulatory consequences

• The process is new, unstable, or recovering from a failure

• The lot is small and inspection is quick

• A customer contract explicitly requires full inspection

• A defect can stop an assembly line or cause expensive field failure

• Automated gauging makes full screening practical

• The lot has been contained after a known process escape

However, “100% inspected” does not mean “zero risk.” Human inspectors can miss defects, gauges can be misused, automated systems can be incorrectly programmed, and repeated inspection can damage delicate parts. Full inspection should have a defined method, calibrated equipment, acceptance criteria, records, and a reaction plan.

Inspection also does not create quality. If a process continuously makes nonconforming parts, sorting is a costly containment action rather than a permanent solution.

Zero-Acceptance Plans and “Zero Defects”

A zero-acceptance plan uses an acceptance number of zero for the specified sample: finding one relevant nonconforming result causes the lot not to be accepted under that plan. This can provide stricter lot acceptance criteria than a plan with a positive acceptance number.

It does not prove that the lot contains zero defects. Unsampled units remain unknown, and inspection errors remain possible. “Ac = 0,” “zero defects,” and “100% inspection” describe different concepts and should not be used as synonyms.

If the requirement is that no known critical defect is permissible, define the product requirement, prevention controls, verification method, lot action, and records explicitly.

When Sampling Is Not Suitable

Acceptance sampling may be a poor fit when:

• The lot is not homogeneous or traceable

• The process is changing rapidly

• A single failure has unacceptable consequences

• The characteristic requires destructive testing and the sample plan is undefined

• Measurement uncertainty is too large relative to the tolerance

• The defect is extremely rare but severe

• Every unit has a unique configuration

• The customer requires individual serial-number results

• Known nonconforming product has already been mixed into the lot

Alternatives can include 100% inspection, automated screening, process validation, statistical process control, proof testing, source inspection, or redesigning the process to prevent the defect.

Building a CNC Inspection Plan Step by Step

1. Define the lot

Identify the part number, revision, quantity, production dates, work order, machine or cell, material lot, and any other information needed to keep the population consistent.

2. Identify characteristics and risks

Review dimensions, geometric tolerances, threads, materials, finishes, notes, cosmetic zones, cleanliness, and functional tests. Link each important characteristic to its failure consequence.

3. Assign the control method

For each characteristic, choose first-off, in-process, sampled final inspection, 100% inspection, certificate review, or functional testing. Use actual measurements where trend and margin matter.

4. Select the governing sampling procedure

State the standard, edition, inspection level, AQL or other indexing parameter, plan type, and current severity. Define who has authority to change the plan.

5. Define defect classes and acceptance criteria

Use objective limits. Replace phrases such as “good finish” or “no sharp edges” with measurable or agreed criteria where possible. The CNC deburring guide explains how edge requirements and burr acceptance can be clarified.

6. Plan random selection

Document how samples will represent the complete lot, including all containers and production stages.

7. Choose suitable inspection equipment

Match the gauge, fixture, datum simulation, resolution, and measurement method to the tolerance. See the CNC machining tolerances guide for feature-level tolerance and inspection considerations.

8. Define the reaction to failure

State whether the lot will be held, returned, sorted, reworked, re-sampled when permitted, or submitted for customer review. Define how corrective action and future tightened inspection will be triggered.

9. Record and review results

The report should identify the lot, sampled units, drawing revision, instruments, inspector, results, nonconformities, disposition, and final lot decision.

Example CNC Inspection Matrix

The following example illustrates how different controls can be combined. It is not a universal sampling plan.

The control plan should reflect the actual drawing and application. Avoid applying one AQL value to every possible failure mode without considering severity.

Common Sampling Mistakes

Treating AQL as an allowed defect percentage

AQL indexes a sampling plan; it is not a shipment permission for a fixed percentage of defective parts.

Using a sample-size calculator without naming the standard

Different assumptions can produce different sample sizes and decision rules. Record the governing procedure and its revision.

Selecting only convenient parts

Top-of-box samples may not represent the lot. Use a documented random method across containers and production time.

Combining unlike production into one lot

Mixed revisions, machines, fixtures, material lots, or suppliers can hide the source of variation and weaken the decision.

Applying one plan to all defect classes

Critical, major, and minor defects may require different controls. A low-risk cosmetic issue should not drive the same response as a safety-related feature.

Re-sampling until the lot passes

Repeatedly drawing new samples after failure undermines the original plan unless a defined standard or approved procedure allows the action. Follow the stated non-acceptance and disposition process.

Forgetting identified defective units

Lot acceptance does not make a known nonconforming sample acceptable. Segregate and control it.

Using inspection instead of process control

Sampling can detect evidence of a problem, but stable tooling, controlled offsets, preventive maintenance, capable measurement, and reaction to trends are what prevent repeated defects.

How Sampling Affects CNC Cost and Lead Time

Inspection cost depends on more than the number of sampled parts. Important drivers include:

• Number of characteristics per sampled unit

• CMM or vision-system programming

• Tight tolerances and complex datum setups

• Inspection fixture or gauge requirements

• Frequency of in-process checks

• Report format and actual-value recording

• Defect classification and separate sampling plans

• Traceability and serial-number requirements

• Customer source inspection or approval holds

• Sorting, rework, and reinspection after failure

Provide inspection requirements with the RFQ. Adding a sampling standard or 100% dimensional report after production begins can require new inspection programs, additional handling, and schedule changes.

What Buyers Should Put in the RFQ

Include:

1. Part number and revision

2. Lot quantity and delivery schedule

3. Characteristics requiring 100% inspection

4. Characteristics covered by sampling

5. Sampling standard and exact edition

6. Inspection level, AQL, plan type, and severity where applicable

7. Critical, major, and minor defect definitions

8. Cosmetic acceptance standard or approved sample

9. Required actual values, reports, and certificates

10. Random-sample identification requirements

11. Lot rejection, sorting, and reinspection rules

12. Customer approval or source-inspection hold points

13. Requirements for future tightened or reduced inspection

14. Packaging and traceability expectations

Use the broader CNC machining RFQ checklist to provide CAD, drawings, materials, tolerances, finishes, quantities, and delivery information.

CNC Inspection Sampling RFQ Template

Please quote CNC machining for part [part number], revision [revision], quantity [lot quantity]. Final inspection shall follow [standard and edition], using [inspection level/plan] and [AQL or other parameter] for the characteristics identified in the attached control plan. Characteristics [list] require 100% inspection and recorded results. Defects shall be classified according to [attached criteria], and samples shall be selected randomly across all containers and production stages. If a lot is not accepted, hold shipment and notify [contact] before sorting, rework, or reinspection. Provide [dimensional report, CMM report, certificates, traceability, and other records] with the shipment.

The buyer and supplier should review the template together. A sampling plan works only when both parties interpret the lot, defect definitions, and reaction rules consistently.

Frequently Asked Questions

Does AQL guarantee the quality of every CNC part?

No. It supports a statistical lot-acceptance decision using a sample. It does not verify every uninspected unit.

Is AQL the percentage of defects a buyer agrees to receive?

No. AQL is an indexing parameter in an acceptance-sampling system, not a contractual allowance to ship a stated defect percentage.

Is a 10% inspection plan statistically valid?

Not by itself. A percentage does not define random selection, acceptance criteria, consumer and producer risks, switching rules, or the governing standard.

Should critical dimensions always receive 100% inspection?

Not automatically. The correct control depends on failure consequence, process capability, error-proofing, measurement method, and contractual requirements. Some critical features need full verification; others may be better controlled through validated processes and monitoring.

Does 100% inspection guarantee zero defects?

No. Measurement error, inspector fatigue, incorrect criteria, programming mistakes, and handling can still create escapes. Full inspection requires a capable system.

Can accepted samples be returned to the lot?

That depends on handling, cleanliness, traceability, and customer rules. Sample identity and any known nonconforming units must remain controlled.

What happens when a sample fails?

Follow the defined plan. Common actions include holding the lot, notifying the customer, investigating the process, sorting or reworking under approval, and applying tightened inspection to future lots where required.

Can the supplier choose a smaller sample to reduce cost?

Only if the contract or approved quality plan permits it. Sample size should follow the selected procedure and risk decision.

Is first article inspection part of routine sampling?

FAI and routine lot sampling serve different purposes. FAI verifies a defined production process and configuration, while routine sampling supports ongoing lot acceptance.

Should measurement results be recorded or only pass/fail?

Record actual values when the drawing, customer, control plan, or process analysis requires them. Numerical data can reveal drift that pass/fail results hide.

Request a CNC Machining Quote with a Clear Inspection Plan

A useful CNC inspection sampling plan defines the lot, random-selection method, characteristics, defect classes, sample size, decision criteria, records, and response to failure. It also separates sampled characteristics from features that require full inspection or stronger process controls.

Send your CAD files, controlled drawing, quantity, material, finish, tolerance, and inspection requirements through the Aluprototype request-a-quote page. The team can review manufacturability and help align the quotation with your required inspection scope.