YARDPUMP · Buying Guide

Does AQL 2.5 Allow 2.5% Defective Pumps in Every Shipment?

AQL is not a guaranteed defect percentage in every shipment. Understand sampling probability, lot definition, defect classes and disposition rules.

A pump inspection order states AQL 2.5. Does that permit 2.5 percent defective products in every shipment? This typical purchasing misunderstanding needs correction before inspection. AQL is a quality-level parameter within an acceptance-sampling system. It is not a guarantee of each lot's actual defective fraction. Nor does the number alone specify the sample size or acceptance and rejection counts. The complete agreed plan determines how the inspection operates.

One number is not a sampling plan

Define the lot, quantity, inspection approach, applicable plan or rules, sample selection, defect categories and decision criteria. The application of normal, tightened or other inspection states and their switching rules cannot be inferred from the AQL value alone. Where a standard is specified, confirm the applicable edition and complete requirements rather than extract a convenient number from an unrelated table.

AQL also does not turn safety-critical issues into an ordinary percentage allowance. Dedicated tests, defect classification and required action must follow the relevant product requirements and approved procedure. Sampling figures used for another consumer product cannot simply be transferred to pump electrical, sealing and performance acceptance. The purchasing document needs to connect each inspection requirement to the characteristic it is intended to control.

Lot acceptance does not measure every unit

Sampling carries statistical risks, including missing a problem or rejecting a satisfactory lot. Acceptance means the sample meets the decision conditions of the agreed plan. It does not provide individual proof for every uninspected pump. Likewise, the observed defective proportion in a sample is not automatically the exact proportion in the entire shipment. Interpretation must retain the distinction between observation and population inference.

For an illustrative independent-sampling model, suppose each unit has a 5 percent probability of being defective. A sample of 20 then has roughly a 36 percent probability of finding none. This is neither a pump-industry defect statistic nor a standard sampling table. It simply shows why zero findings must be interpreted alongside sample size and detection capability. The actual sampling arrangement may require a different statistical model.

StatementReasonable interpretationUnsupported conclusion
A specified AQLOne parameter within the sampling systemEvery lot may contain that percentage defective
Sample acceptedThe agreed acceptance conditions were metEvery unit has passed inspection
Zero observed defectsNone found in the inspected sampleThe entire lot contains none

Align defect definitions before selecting numbers

If inspectors assign the same observation to different categories, results become difficult to compare. Provide practical descriptions and decision evidence for appearance, accessory completeness, label consistency and relevant tests. Characteristics that cannot be evaluated visually need suitable measurements. AQL does not supply the missing test method, instrument capability or product specification. Those requirements must be established separately before the inspector begins work.

Samples also need random selection from a defined inspection lot, separated from previously reworked, specially selected or different-version units. Performance, electrical and sealing characteristics may require different validation arrangements. That is not permission to omit necessary checks; it is a reason to define what each check establishes. Retain sample identities and original findings even when several different inspection groups all reach an acceptable result.

Agree the response to rejection

Specify confirmation of an abnormality, containment, investigation, rework or screening, resubmission and reinspection. Avoid the improvised practice of drawing another group until one happens to pass. A second inspection must not erase the first finding, and selected good units must not be relabelled as the original random sample. The response should follow the actual risk and the agreed procedure.

Acceptance sampling can be useful within a purchasing system, but it cannot replace clear specifications, controlled production and necessary product tests. The important question is not whether the contract contains three familiar letters. It is whether both parties share an executable understanding of the samples, characteristics, decision rules and response to nonconformity, with records sufficient to demonstrate what occurred.

Questions about AQL

Does AQL 2.5 permit 2.5 defective pumps in each hundred?

Not as a direct lot-by-lot interpretation. It is not a simple permission or guarantee for the actual defective quantity.

Can zero observed defects replace electrical tests?

No. Follow the applicable product requirements and approved test and release arrangements for each characteristic.

Can a failed lot pass through an informal second sample?

Do not change the rules after seeing results. Preserve the first findings and follow the agreed resubmission procedure.

Related Reading