AQL Sampling vs Microbiological Sampling in Frozen Food

AQL sampling and microbiological sampling answer different purchasing questions, but they are not mutually exclusive kinds of statistics. An AQL-based quality inspection can support a lot decision about defined defects. A microbiological plan connects a particular organism or indicator, food, method and sampling design to a stated decision. Passing the first does not satisfy the second. A negative laboratory result also does not establish that every part of the lot is free from contamination.

For a frozen-food order, agree the purpose before choosing a sample count. Identify the lot, define the unit being selected, state what the laboratory will actually analyze, and record the acceptance rule and decision owner. Those details make two reports comparable without making them interchangeable.

Start with the decision the sample must support

A request to “sample the container” leaves several questions unanswered. A container can hold multiple production lots, products or packing runs. The buyer may want to verify carrot cut quality, check retail pack seals, investigate an indicator-organism trend or assess conformity with a microbiological criterion. Each purpose needs a recorded link between the selected goods and the resulting decision.

Write that decision in ordinary language first. For example: “Determine whether the identified carrot lot meets the agreed visual defect specification,” or “Assess the identified product against the applicable microbiological criterion before release.” These are proposed instruction examples, not complete sampling plans. The relevant inspection or technical team must add the selection procedure, analytical details and acceptance requirements.

NIST’s introduction to acceptance sampling explains that a random sample supports acceptance or rejection of a lot. Its main purpose is disposition, rather than estimating the lot’s exact quality. That distinction matters when someone presents an inspection “pass” as a complete description of everything in the shipment.

There is also an important overlap in terminology. Codex guidance on microbiological criteria says most microbiological lot-acceptance plans are attributes plans. Calling one report “AQL” and another “microbiological” therefore does not establish how either sample was selected or how its results should be interpreted.

Use AQL as part of a complete acceptance plan

AQL alone is not an inspection instruction. NIST calls it the acceptable quality level: a reference used to design or select an acceptance scheme. NIST’s sampling-plan explanation connects that reference with acceptance probabilities and the risks of rejecting relatively good lots or accepting relatively poor ones. The sample size and acceptance number work together. An AQL percentage is not a measured defect percentage for the lot in front of you.

For a proposed appearance inspection, ask which published scheme and version, or separately agreed plan, governs the work. Record the lot size basis, inspection level where relevant, sample size, acceptance and rejection numbers, and any applicable switching rules. A report that states only “AQL passed” leaves the buyer unable to reconstruct the actual decision.

The inspected unit also needs a definition. Is a nonconforming unit a whole retail bag, one individual piece or a specified quantity of product? Are defects counted as affected units or as separate occurrences? A bag with several damaged pieces can produce very different counts under those alternatives. Do not change the denominator between the specification and the report.

With blueberries, a buyer could define a visible quality field around form or an agreed defect description. The reference needs enough detail for two inspectors to classify the same observation consistently. A photograph can help explain the appearance being discussed, but it cannot supply the lot selection method or convert visible appearance into a microbiological finding.

Keep the defect classes explicit. A routine commercial appearance tolerance should not silently become the rule for a food-safety hazard. If the requested check includes a critical condition, the technical agreement must state its own treatment rather than leaving the inspector to infer it from a general AQL line.

Read the microbiological criterion before ordering a test

“Microbiological testing required” is equally incomplete. Start with the actual food and intended use, the organism or indicator, the purpose of testing and the point in the supply chain. Identify the applicable destination requirement and the buyer’s agreed specification. A criterion used to investigate process hygiene may support a different action from one used for lot acceptance.

For example, the current FSANZ microbiological compendium states that its general ready-to-eat guideline limits are intended to evaluate food handling controls, rather than act as lot acceptance or rejection sampling plans. Copying a number from a reference table without its purpose can change what the result appears to authorize.

Read the notation with the criterion. In a two-class plan, results fall into two categories. In a three-class plan, m and M distinguish conforming, marginal and nonconforming result bands. The number n specifies how many sample units are required, while c governs the permitted number in the specified nonconforming or marginal category, depending on the plan. In a three-class plan, a result above M is unacceptable. The precise boundary treatment and values belong to the applicable criterion.

Ask the laboratory to confirm the method, food matrix, analytical portion, reporting units and required reporting capability against that criterion. Keep the criterion’s title, version and relevant product category with the purchase specification. This prevents a later reviewer from finding a matching organism name but a different analytical basis.

A frozen vegetable described as ready-to-eat and a product supplied for a defined cooking step should not be assigned the same microbiological brief merely because both are frozen. Record the intended use as part of product identity, then have the responsible technical team confirm which requirement applies.

Separate the collected unit from the analytical portion

The amount dispatched to a laboratory is not automatically the amount analyzed. A collected unit might be an unopened bag or a separately collected quantity from an identified bulk lot. The laboratory then follows the specified method to prepare and analyze a defined portion. The chain between those levels must remain visible.

FDA’s BAM chapter on food sampling and sample preparation illustrates why representative selection, sample integrity and preparation details matter. Its sampling and compositing instructions sit within particular analytical and regulatory contexts. They should not be lifted out as a universal sample mass for every frozen food or microorganism.

When comparing quotations, ask what “one sample” means on each price line. One courier parcel, one collected unit, one laboratory preparation and one test result are different counting levels. A low price per sample is difficult to assess until the laboratory states which level it is pricing and what is included.

Use a short identity chain in the request: purchase reference, production lot, selected package or collection position, sample ID and requested analysis. Ask the laboratory to preserve those references in its receipt and report records. If its internal laboratory number differs from your sample ID, retain the cross-reference rather than replacing one identifier with the other.

Also confirm the material needed for the full requested test set, any retained portion and possible confirmation work under the method. The practical shipping quantity is a planning question for the laboratory. Increasing that quantity alone does not necessarily increase the number of independently selected units represented in the result.

Decide whether compositing is permitted

Compositing combines material from multiple units for an analysis. It can change the information available from a report, so it needs agreement before collection and preparation. “Several bags sent” does not tell you whether the laboratory analyzed them separately or pooled portions into one preparation.

Codex notes that pooling can affect concentration and is not appropriate for enumeration or three-class plans. For suitable presence-or-absence testing, pooling requires assurance that the result is not affected compared with testing the units individually. The exact permission and procedure must come from the relevant criterion and validated analytical approach.

Ask for the proposed composite structure in writing. Which original units contribute? What portion comes from each? Which method covers that combined quantity and food matrix? How will the result be reported? These questions should be settled with the laboratory before anyone combines the material, because the original separation cannot be recovered from a blended preparation.

A composite report also needs a traceable link to its contributors. If the report identifies only “mixed sample,” the buyer may be unable to establish which selected units it represents. Conversely, do not interpret one composite result as a set of separately observed negative results merely because several original sample IDs appear on the submission form.

Compare the proposed arrangement with the purpose. Separate results may be important for a decision that depends on how many units fall into particular categories. A permitted pooled analysis may answer a different defined question. Convenience or freight cost should not make that choice implicitly.

Understand why no findings do not prove absence

A simple probability example shows the limitation. Suppose independent selections from a very large lot each have a fixed 1% chance of being a nonconforming unit, and classification is perfect. Using the binomial model described by NIST, the chance of finding zero such units is (1 − 0.01)ⁿ. It is about 90.4% for ten selections and 60.5% for fifty.

These are hypothetical calculations, not recommended sample counts or estimates for a particular food. They show that a sample can contain no findings even when nonconforming units exist. The calculation does not give the probability that a real lot is clean. Actual contamination patterns, selection and analytical performance require their own assessment.

Read a negative report at the level it describes: the reported analyte, method, sample identity and analytical quantity. Keep any laboratory qualification or reporting limitation attached. Rewriting that result as “pathogen-free shipment” removes those boundaries and extends the statement beyond the evidence.

Before using a report for release, reconcile it with the other relevant information already available. A sample result should not make a documented process deviation, identity discrepancy or unresolved handling event disappear from the review. The decision owner needs the complete set of relevant evidence, including the reason a test was requested.

This is especially important when a test is commissioned after a concern has arisen. An investigation sample and a routine lot-acceptance sample may have different selection logic. Label the purpose honestly so that a reader does not mistake targeted investigative evidence for a random survey of the shipment.

Keep sample selection and handling connected to the lot

A technically suitable test can still answer the wrong question if the selected material is poorly identified. Before collection, make the lot breakdown available to the person selecting samples. Record which goods were accessible and any departure from the agreed selection procedure. A convenient handful from an open carton should not be described as random lot sampling without a basis.

The collection record should explain where and when each unit came from, who selected it and which lot marking was observed. Photographs can support that identity record when they show the relevant code and surrounding package. They should complement the written sample list, because a photograph alone may not explain how the unit was chosen.

For microbiological work, follow the laboratory’s agreed instructions for aseptic collection, packaging, temperature protection and dispatch. FDA’s BAM guidance emphasizes preserving sample condition and maintaining the integrity of collected material. Ask the laboratory to record receipt condition and any deviation that affects suitability, rather than assuming courier delivery establishes an acceptable sample.

Keep the visual-inspection sample and laboratory sample distinguishable when their handling differs. Product spread on a tray for appearance inspection should not casually become a microbiological sample. Plan the required material and containers in advance so that one inspection activity does not compromise another.

The inspection scope for the order should identify the responsible collection party, laboratory, report recipient and approval owner. Those roles make it easier to resolve a missing code or unsuitable sample while the selected goods can still be identified.

Compare two reports without merging their decisions

Consider an illustrative purchase review called Q-27 for frozen carrot dice. The buyer receives an appearance-inspection report and a microbiological report. Both carry the purchase reference. That is a useful starting link, but the reviewer still needs to establish whether they cover the same product form and production lot, and whether each required assessment is complete.

Suppose the appearance report identifies the lot and provides its defect counts, but the laboratory submission omits whether the units were composited. Do not fill that gap from the inspection report. Ask the laboratory for the preparation and reporting basis, then compare it with the criterion. The missing detail concerns analytical evidence, even if physical quality has already been accepted.

Reverse the situation and the logic remains useful. A complete laboratory report cannot resolve an appearance report that uses an undefined defect category. Request the classification reference and decision basis for that inspection. Keep each open item assigned to the person who can supply the missing evidence.

The final review can record separate statuses for physical quality, microbiological assessment and overall release. This allows a buyer to acknowledge completed work while keeping an unresolved requirement visible. Neither a green inspection box nor a laboratory report should silently close every other condition attached to the order.

Agree what happens when evidence is incomplete

Plan the response to missing, unsuitable or nonconforming evidence before the shipment is waiting for approval. Identify who can place the relevant goods on hold, investigate the result, obtain clarification and authorize the next step. Record which lot or quantity each action covers.

Repeat sampling is not a general way to replace an unwelcome result. Codex states that repeat lot testing should follow provisions in the applicable sampling plan. Keep the original result, reason for any further work and subsequent decision in one review record. A later negative finding should not erase the earlier finding or its investigation.

For an incomplete report, the next step may simply be obtaining the missing method reference, sample mapping or receipt qualification. For an unsuitable sample, it may require a new collection under an agreed procedure. The responsible technical team should distinguish those situations instead of treating every uncertainty as a request for another certificate.

Before approving repeat orders, check whether the product, intended use, criterion, laboratory method or production route has changed. Carry forward the approved reasoning only where it remains applicable. The product specification and sampling record should make the current version easy to identify.