Pet Bag AQL: Acceptance Quality Limit for Bulk Orders

Wholesale pet bag sourcing desk · Updated 2026-10-06 · 14 min read

Pet bag retail buyers and pet bag wholesale business owners look at the same unit cost from opposite ends, and Market & Business Strategy is where those numbers meet. Pet bag procurement teams normally fix quality first and negotiate price second, which is why pet bag supplier selection is really a documentation exercise.

What AQL Is and What It Is Not

The acceptance quality limit is a threshold used to derive a sampling plan. It is expressed as a percentage or as defects per hundred units, and it describes the quality level that the plan will accept most of the time. It is a property of the plan, not a property of the lot.

That distinction is the whole subject. A lot accepted under an AQL 2.5 plan does not contain 2.5 percent defects. It may contain none, it may contain more, and the plan's job is to make the second outcome unlikely rather than impossible.

AQL is also not a process target. A production line aiming at 2.5 percent major defects is a line that has given up; a competent line runs well below its acceptance limit precisely because it needs headroom for the sampling risk.

Nor is AQL a specification. It says nothing about what a defect is, and a sampling plan applied to an undefined requirement produces an authoritative-looking number about nothing. Definition comes first, sampling second.

The three common readings of the number, in ascending order of usefulness, are: a vague quality promise, a shorthand for a sampling plan, and a negotiated allocation of statistical risk. Only the third is commercially meaningful.

The risk allocation is two-sided and worth naming. There is a producer's risk, that a lot meeting the limit is rejected by chance, and a consumer's risk, that a lot worse than the limit is accepted by chance. Both are built into the plan and neither is zero.

Buyers should therefore treat the AQL as a sampling decision with a cost. A tighter limit means a larger sample and more inspection time; a looser limit means less inspection and more residual risk. There is no free position.

The vocabulary a buyer needs is short: lot, sample size, inspection level, code letter, accept number, reject number and switching rules. Seven terms, and each one appears on a real inspection report.

ISO 2859-1 and the Sampling Standard Behind the Number

The plan behind most consumer goods inspection is ISO 2859-1, published as an international standard for sampling by attributes and known in North America in an equivalent form as ANSI/ASQ Z1.4. The two are technically aligned for practical purposes, which is why a supplier and a buyer in different regions usually agree without realising they are citing different documents.

Sampling by attributes means each unit is classified as conforming or non-conforming against each characteristic, rather than measured on a continuous scale. A zip works or it does not; a panel is aligned or it is not. This suits soft goods inspection, where most criteria are pass or fail.

The standard provides tables linking lot size to sample size through a code letter, and linking sample size and AQL to accept and reject numbers. Using the published tables is what makes an inspection reproducible: two inspectors following the same plan check the same number of units.

The framework is maintained and published by ISO, and a buyer who wants to understand rather than trust the number should read the structure of the standard once. It is not long and it resolves most of the arguments that otherwise recur every season.

There is a family of related standards worth knowing about. Plans for isolated lots, plans for continuous production and skip-lot plans all exist, and using the right one depends on whether the supplier relationship is ongoing or one-off.

For an ongoing wholesale programme, the standard switching rules matter more than the individual plan, because they make the inspection intensity respond to recent history. A supplier with a good record is inspected less; one with a poor record is inspected more.

Third-party inspectors work to the same tables, which is why a report from an inspection provider such as SGS is comparable with an in-house report. Both state the standard, the level, the sample size and the findings, and both can be checked against the same table.

The practical instruction for a buyer is short: state ISO 2859-1 or its equivalent, general inspection level II, and the per-class AQL values. That sentence contains everything needed to reproduce the inspection.

Lot Definition, Inspection Level and the Code Letter

A lot is the group of units the sample represents, and how it is defined changes the sample size and the meaning of the result. A lot should be a homogeneous group: the same style, the same colourway, the same material lot and the same production period.

Mixing styles or colours into one lot is the most common procedural error, and it invalidates the result. A sample drawn across three colourways tells a buyer nothing about any of them, and a defect clustered in one colour is diluted by the other two.

Lot size drives sample size, but not proportionally. Sample size grows far more slowly than lot size, which is the economic insight behind sampling: inspecting 10,000 units costs little more than inspecting 3,000, and neither approaches the cost of inspecting everything.

Inspection level sets the relationship between lot size and sample size. Level II is the general default and is what a buyer should specify unless there is a reason not to. Level I gives a smaller sample and is used where inspection is expensive or the risk is low; level III gives a larger sample and is used where discrimination matters more.

Special inspection levels exist for destructive or very costly tests, where a small sample is unavoidable. They are rarely relevant to a visual and functional inspection of soft goods, but they appear in testing programmes.

The code letter is the intermediate step in the table. Lot size plus inspection level produce a code letter, and the code letter plus the AQL produce the sample size and the accept and reject numbers. It is a lookup, not a calculation.

The table above shows the shape of the system, and the pattern is instructive. At AQL 2.5 the accept number grows with the sample, but the tolerated rate implied by the plan actually tightens as the lot grows. Larger lots are held to a stricter effective standard.

Defining the lot in the order terms is therefore worth doing. One sentence stating that a lot is a single style and colourway from a single production period, and that mixed lots will not be presented for inspection, removes a whole category of dispute.

Choosing AQL Values by Defect Class

A single AQL for everything is the most common specification error. Different classes of defect carry different consequences and should carry different limits, which is why a serious inspection states three numbers rather than one.

Critical defects carry a limit of zero, and this is expressed differently from the others. In practice a critical finding fails the lot regardless of the accept number, and many inspectors record it as an automatic rejection rather than as a count against a limit.

Major defects normally carry AQL 2.5 for consumer soft goods. This is the industry default for a general merchandise programme and is the number most wholesale pet bag orders should state. It tolerates a small number of findings in a sample while still failing a lot with a systemic problem.

Minor defects normally carry AQL 4.0. Minor findings are numerous and low-consequence, and holding them to a tight limit produces rejections over issues that no customer would notice, at a cost that is not justified.

Tighter values are available and cost more. A programme specifying AQL 1.0 for major defects is asking for a larger effective scrutiny at the same sample size, which means more borderline findings are classified as failures, and the rejection rate rises accordingly. That may be correct for a safety-critical or premium placement.

Looser values suit low-margin, high-volume commodity goods where the cost of inspection exceeds the cost of the defects it would catch. They are a legitimate commercial choice when made deliberately.

The classification scheme must be agreed before the limits mean anything. Two parties using AQL 2.5 with different definitions of major will produce different results on the same goods, which is why the limit sample and the classification list belong in the same document as the AQL.

The final point is consistency across the programme. Using one set of values for the first order and a different set for the second, without a stated reason, makes the trend data meaningless and looks like a negotiating tactic rather than a quality decision.

Reading the Result: Accept, Reject and the Borderline Case

The outcome of an inspection is a count compared against two numbers. At or below the accept number the lot passes; at or above the reject number the lot fails. Between them there is nothing, because the reject number is one greater than the accept number by construction.

A borderline result is where commercial judgement enters and where disputes start. A sample of 125 at AQL 2.5 accepts 7 and rejects 8, so a lot with exactly 7 major findings passes by a single unit. A buyer who believes such a lot is good quality has misread the plan.

The correct reading of a marginal pass is that the lot is probably near the limit. Sampling error means the true rate may lie on either side, and a prudent buyer responds by tightening the inspection on the next lot rather than by celebrating.

Re-inspection after rework has its own rules and should not be assumed to reset the risk. A reworked lot is commonly inspected at a tightened level, because a lot that has already failed once carries an elevated probability of a second issue.

Sorting is the alternative to rework and has a different risk profile. A 100 percent sort removes the specific defect it was looking for and nothing else, and a sorted lot frequently carries a residual rate in unchecked characteristics that is worse than the original.

The report should state the plan, the sample size, the findings by class and the decision, and a buyer should read all four. A report that gives only a pass or fail conclusion without the counts is not verifiable.

Photographic evidence of findings is standard practice and cheap to require. A buyer receiving photographs of the defects along with the report can form an independent view on severity, which is the basis of any argument about classification.

Independent verification is available where the result is disputed, and inspection providers such as SGS will re-inspect a held lot. The cost is modest against a container at risk, and it converts a standoff into a number both parties accept.

Switching Rules: Normal, Tightened and Reduced Inspection

A sampling plan is not applied in isolation over time. The standard defines switching rules that move inspection intensity up or down based on recent results, and these rules are the mechanism that makes an ongoing programme self-correcting.

Normal inspection is the starting point for every new programme and the default after any disruption. It uses the standard sample size for the code letter.

Tightened inspection is triggered by a poor record, conventionally two out of five consecutive lots failing. It uses a larger sample with the same accept number, which makes rejection more likely and is deliberately uncomfortable. Its purpose is to protect the buyer during a known problem period.

Reduced inspection is triggered by a good record, conventionally ten consecutive lots passing with production at a steady rate and no other concerns. It uses a smaller sample and saves inspection cost where the risk has been demonstrated to be low.

Return to normal is the rule buyers forget. Tightened inspection ends after five consecutive accepted lots under tightened, and reduced inspection ends immediately on a failure or on any sign of process instability. The system is designed to return to the middle.

Discontinuation is the final step and the one that ends relationships. Where tightened inspection persists without improvement, the standard contemplates discontinuing acceptance until the supplier demonstrates corrective action. It is a strong signal and a fair one.

Buyers should ask which state their programme is in. A supplier reporting AQL 2.5 results without stating the switching state is withholding information that materially changes the meaning of those results.

The commercial use of this is constructive rather than punitive. A programme that has earned reduced inspection has demonstrated quality and should be rewarded with less inspection cost; one under tightened inspection should be paying for the extra scrutiny it created.

Our production team reports the switching state alongside every result, because a pass under tightened inspection and a pass under reduced inspection are not the same statement about quality.

What AQL Does Not Tell You

The value of knowing AQL lies as much in its limits as in its use. Four things it does not do account for most of the disappointment buyers experience with an inspection regime that appeared to be working.

It does not guarantee the lot. A passed lot carries a residual risk that rises with the acceptance number, and a buyer shipping straight to a retail distribution centre on the strength of a marginal pass is accepting a known probability of a downstream find.

It does not control the process. Sampling inspects output; it does not change how output is made. A line producing a steady 2 percent major rate will produce it indefinitely under any sampling plan, because nothing in the plan feeds back into the process.

It does not cover characteristics not checked. An inspection checks a defined list against a defined specification, and a defect outside that list is invisible to it. A buyer who assumes an inspection covers everything has written an incomplete checklist.

It does not address late failures. A zip that fails after a thousand cycles passes every pre-shipment inspection, and the failure appears as a warranty claim months later. Component qualification and cycle testing are the controls that address durability.

It does not substitute for a specification. The most rigorous sampling plan applied to an undefined requirement produces a number with no meaning, which is why the limit sample and the written specification come first.

It does not measure appearance consistently without help. Cosmetic judgement varies between inspectors, and without photographic limit samples the classification drift can exceed the effect of the sampling plan itself.

The layered answer is the one worth repeating: component qualification before bulk, defined specification with limit samples, in-process control during production, and a final sampling inspection to AQL 2.5 before release. Each layer covers a gap the others leave.

Buyers who want a single sentence to remember should use this: AQL decides how much to check and how much to tolerate, and everything else in quality has to be done by other means.

Single Sampling, Double Sampling and Multiple Sampling Plans

Within the standard framework there is more than one plan type, and the choice affects how many units are checked before a decision is reached. Most consumer goods programmes use single sampling and never consider the alternatives, which is usually but not always the right call.

Single sampling draws one sample of the stated size and decides from it. It is simple, it is predictable in cost, and it is what a supplier and an inspector can agree over email without further definition. For most wholesale pet bag orders it is the correct choice.

Double sampling draws a first smaller sample and decides to accept, reject, or draw a second sample where the result is borderline. Its advantage is economic: clearly good and clearly bad lots are decided on a smaller first sample, and only the ambiguous cases consume the full inspection.

The disadvantage of double sampling is administrative. It requires an inspector to remain available for a second draw, it extends the calendar time of the inspection, and it complicates a schedule that is already tight against a container booking.

Multiple sampling extends the same logic to up to seven draws and is used where inspection is very expensive or destructive. It is rarely appropriate for soft goods, where inspection is neither.

The average sample number is the comparison metric and it rewards the right choice. For a lot of genuinely good quality, double sampling inspects materially fewer units on average than single sampling, and the saving is real where a programme inspects continuously.

Where the saving is not real is a programme with a variable defect rate. Lots that fall in the borderline band require the second draw, and a programme whose quality is inconsistent will frequently use the full sample anyway while carrying the extra administrative burden.

The practical recommendation is to use single sampling for a normal wholesale programme and reserve double sampling for a high-volume, continuous relationship where the inspection volume justifies the administration. Either way, name the plan type in the order terms alongside the standard, the level and the per-class limits.

The choice has no effect on the meaning of the AQL itself. A lot accepted under double sampling at AQL 2.5 carries the same class of residual risk as one accepted under single sampling at AQL 2.5, because the plans are designed to comparable operating characteristics.

Inspection Cost, Risk Allocation and Order Wording

Inspection cost is driven by sample size and by the number of characteristics checked. A standard general inspection at level II on a mid-size lot is a part-day assignment for one inspector, and the cost is small relative to a bulk order but not relative to a marginal product.

Buyers should price the inspection into the order. A programme quoting a unit price without an inspection line is either absorbing the cost silently or not inspecting, and both are worse than knowing the number.

The cost of a failure dwarfs the cost of the inspection. A failed lot means rework or replacement, a second inspection, a delayed container and potentially a missed retail window, and the least of these is usually larger than the inspection fee by an order of magnitude.

Risk allocation is the real content of the AQL decision. A tighter limit shifts risk to the supplier and is priced into the unit cost; a looser limit leaves risk with the buyer and is paid for in returns and claims. Neither is free, and the choice should be explicit.

The order wording should therefore be a short, complete clause. It names the standard, the inspection level, the lot definition, the per-class AQL values and the consequence of a failure. Five elements, one paragraph, no ambiguity.

Re-inspection cost after a failure should be allocated in the same clause, along with responsibility for freight on any replacement. These are the two items most often disputed after a failed lot.

Timing must be built into the schedule. Samples take 6-10 working days and bulk production runs 35-50 days after approval at MOQ 500 pieces per colourway, and a failed lot adds a rework and re-inspection cycle inside that window. Buyers should not plan a container date on the assumption of a first-time pass.

Final practical note: require the report, the counts and the photographs as a condition of release, and file them by lot. The accumulated reports are the evidence base for every specification decision in the following season.