On this page
A product specification states the requirements you agree to buy against; a certificate of analysis, or COA, reports results associated with an identified batch or sample. Compare them by matching the product, lot, specification version, test method, unit, acceptance criterion and reported result. A document headed “COA” does not complete that comparison for you, and a result marked “Conforms” needs to be interpreted according to the type of test.
For frozen-food orders, we treat this as a practical handoff between procurement, quality and the proposed supply route. Procurement needs to know what was promised. QA needs to know what was checked and whether the evidence supports the agreed decision. The warehouse needs records that match the cartons arriving. A useful document review connects those questions without turning one laboratory report into a guarantee about everything in the shipment.
Use each document for the decision it supports
The specification is the continuing reference for a defined product until an approved change replaces it. It can include visible form, size, ingredients, sensory characteristics, defect tolerances, analytical requirements, packaging, storage, intended use and required records. Some of those fields are tested in a laboratory; others are checked by inspection, document review or an application trial.
Illustrative document roles: the approved specification supplies requirements, while the COA reports results linked to identified material.
A COA normally presents analytical or quality results relevant to the identified material. It may be issued by a production facility using its own results, by a testing laboratory, or as a supplier summary linked to supporting reports. The issuer and underlying evidence need to be clear. A supplier-issued COA is not automatically invalid, just as an independent laboratory logo does not establish that the report covers every requirement in your order.
A typical analysis has another purpose. It can help evaluate a product during sourcing, but representative values do not become results for an allocated production lot merely because they look suitable. Keep an example COA used during quotation visibly separate from the report later supplied for the shipment. A generic product sheet with expected values belongs in the qualification file, with its scope correctly described.
We also separate the product specification from a certificate covering a management system or facility. A facility certificate may support supplier qualification within its stated scope. It does not tell you the measured Brix, defect level or microbiological result of the particular product being allocated. Request the record that answers the outstanding question rather than collecting several differently titled documents that repeat the same limited information.
The Canadian Food Inspection Agency’s supplier assurance guidance places product requirements, supplier agreements and verification within a broader programme. That is useful context for document review: a COA contributes evidence to an agreed control system. It does not replace the product requirement or the other checks needed for the buying decision.
Match the product, lot and specification version
Confirm identity before comparing results: product name, form, treatment and intended use. Whole frozen strawberries, irregular strawberry pieces and a prepared strawberry ingredient are not interchangeable descriptions. A laboratory method may cover more than one form, but the purchasing specification and the sample identity still need to describe the material being supplied.
The named fruit and presentation must match the order and the identified material described by the report.
Then connect the report’s batch or sample identifier to the supplier’s lot record, carton coding and shipment documents. A laboratory sample number and a production lot code may legitimately be different strings. When they differ, obtain the record that links them. Guessing that two similar numbers refer to the same material leaves a gap that becomes harder to resolve after cartons have been split between deliveries.
A shipment can contain several production lots. Build a simple lot-to-document list showing each lot, its quantity and the report or reports that cover it. One combined document can be usable if it clearly identifies the separate lots and the evidence applicable to each. One unqualified result for an unidentified mixture cannot silently stand in for separate lot results where your agreement requires them.
Check which specification revision supplied the limits printed on the COA. A report can accurately show compliance with an older, less demanding version while leaving the current order unassessed. Keep the approved revision or date beside the purchase order and the supplier’s acknowledgement. If the report uses a supplier’s internal specification, compare that reference against the buyer’s agreed requirements instead of assuming that identical product names imply identical limits.
Dates have different jobs. Production, sampling, receipt at the laboratory, testing and report issue dates describe different events. A newly issued copy of an old report does not mean the material was tested again. Likewise, an amendment may correct an identifier without changing the analytical work. Ask what changed and retain the amendment history when it affects the evidence being reviewed.
At this stage, an unresolved mismatch is a document question requiring clarification. It is not enough evidence to diagnose poor product quality or accuse anyone of fabricating a result. Describe the exact missing connection: for example, “Please link laboratory sample S-04 to the lot code on the allocated cartons.” That gives the supplier a concrete issue to resolve.
Compare complete rows, not isolated numbers
Review one parameter at a time across the approved specification and the COA. A complete comparison names what was measured, how the sample was prepared, the method or reference, the reporting unit, the limit and the result. Where a sampling or decision rule is part of the requirement, keep it connected to that row rather than treating it as an optional footnote.
| Row element | Question to resolve | Example of a misleading shortcut |
|---|---|---|
| Parameter and basis | Are both documents describing the same characteristic on the same material basis? | Comparing drained-fruit results with fruit plus retained liquid. |
| Method and preparation | Is the named procedure the agreed one or an accepted equivalent? | Treating two different sample preparations as identical because both say Brix. |
| Unit | Are units identical or correctly converted? | Comparing a value in micrograms per kilogram directly with a limit in milligrams per kilogram. |
| Acceptance criterion | Is this a maximum, minimum, range or qualitative requirement? | Reading a minimum as though a lower value were better. |
| Result and qualifier | Is it measured, below a reporting threshold, qualitative, provisional or missing? | Replacing a less-than result with an exact zero. |
For a unit check, 0.020 mg/kg equals 20 micrograms/kg. The conversion changes the written number, not the concentration. Retain the original reported value and show any conversion in the review record. Avoid converting away a qualifier: “less than” remains “less than” after the unit changes. These numbers illustrate unit handling and are not legal limits for any food or substance.
A method name also needs its relevant scope. An in-house procedure is not automatically unsuitable, and a familiar standard number does not automatically make two results comparable. Ask the laboratory or qualified reviewer whether the method is appropriate for the product matrix, parameter and required sensitivity. If the supplier proposes an alternative, resolve its acceptability before using it for the agreed decision.
Some physical results need a distribution or a defined fraction, not just an average. A mean broccoli crown width within the nominal band does not tell you how many pieces were outside it. Similarly, “broken fruit” needs the agreed defect definition and count or weight basis. Match the reported statistic to the specification; a number can be accurate while still being the wrong answer.
Our guide to comparing Brix in IQF fruit develops one example of this method problem. The document-review principle is the same: verify what went into the measurement before using the reported value to rank samples or accept a lot.
Read less-than results and conformity statements carefully
“Not detected,” “below quantification” and a numerical less-than result do not all communicate the same information. The laboratory should explain the qualifier, the relevant detection or reporting boundary and the unit. A result below a method’s boundary is not an exact measurement of zero, and a detection limit should not be casually relabelled as a quantification limit.
Consider an entirely hypothetical chemical parameter with an agreed maximum of 0.010 mg/kg. The report says only “less than 0.050 mg/kg.” That statement leaves open values on both sides of the agreed maximum. It therefore does not, by itself, establish compliance with the lower limit. Ask for a method and reporting capability suitable for the requirement, or have the qualified reviewer resolve what additional evidence is needed.
Hypothetical example: a less-than result can leave the actual value unresolved on either side of the agreed maximum. These are not regulatory limits.
The example is about the information available in the report. It does not say that the product actually contains 0.050 mg/kg, that it exceeds the maximum, or that the named numbers are regulatory limits. Both assuming a failure and assuming a pass would go beyond that result. Keep the conclusion proportionate to the evidence: the current statement is insufficient for the requested comparison.
“Conforms” requires a different reading. For a qualitative sensory requirement, such as characteristic colour against an agreed reference, a documented conformity outcome may be appropriate. Replacing it with an invented numerical score would add no value. For a quantitative criterion where your agreement requires the actual result, the same word may omit information needed to review the margin or compare lots.
Ask whether the conformity statement identifies the requirement, the method and the decision rule used. Results close to a numerical limit can require consideration of measurement uncertainty under the applicable rule. Do not create your own uncertainty allowance after seeing the result. The laboratory and responsible QA reviewer should establish how the decision is made and whether the reported information supports it.
Current Global ACI guidance addresses decision rules and measurement uncertainty in testing. Use that framework through the laboratory or accreditation requirements relevant to the order. The buyer’s immediate task is to obtain a clear basis for the statement, rather than treating every printed “Pass” as the same kind of decision.
Check the test panel, sample and laboratory scope
Compare the required test list with the actual panel. A COA might report routine physical checks while a separate laboratory report covers selected residues or microbiology. Another parameter may be verified periodically under an agreed programme. Identify the evidence and frequency for each requirement. A blank row is not a passing result, but the explanation may be a separate record rather than an omitted test.
Distinguish an agreed periodic verification result from a test performed on the allocated lot. Both can have a defined role, but they should not be presented as the same evidence. Ask which parameters are tested per lot, which use another frequency and what triggers additional testing. Confirm that the programme is suitable for the product, intended use, destination and customer requirements; do not borrow a test frequency from an unrelated product specification.
Illustrative sample path: evidence tied to Lot A cannot silently be transferred to Lot B. The diagram does not prescribe a sampling plan.
Sampling determines which material reached the test. The record should identify who sampled it, the source lot, the sampling stage and the relevant plan or procedure. A result on raw material may help review an upstream control, while the specification asks for a result on packed finished product. The two records need a justified relationship; similar numbers do not supply that missing connection.
Laboratory accreditation also has a defined scope. ILAC’s guidance describes accreditation in terms of specific conformity assessment activities. Check the laboratory’s current status and relevant scope, including the test, product matrix and location where applicable. With a flexible scope, request the current detailed coverage needed to confirm the activity. An accreditation symbol alone is not a complete scope check.
A sample result cannot establish that every untested piece is identical. CDC explains this limitation in food-outbreak investigations: only part of the food can be tested, so a negative result cannot prove that contamination was absent. For procurement, the practical consequence is to review sampling, process controls and other evidence alongside the COA. A negative result does not turn frozen produce into a product suitable for an unapproved use.
If a report contains a suspected or confirmed food-safety failure, route it promptly to the responsible food-safety team under the applicable control procedure. Repeated testing is not a general route to clearing the concern. FDA’s Salmonella guidance, for example, says a confirmed positive remains valid despite later negative results unless circumstances clearly demonstrate that the first result was inaccurate.
Work through a frozen-fruit document review
Imagine an order for whole IQF strawberries supplied against a named buyer specification. The allocated cartons carry Lot A, and the supplier provides a COA with the same lot and the correct specification revision. Identity alignment is a useful first step. Now review what the individual rows actually establish. The following entries are invented examples, not XMG specifications, laboratory findings or legal limits.
| Agreed requirement | Example reported information | Document-review conclusion |
|---|---|---|
| Whole IQF strawberries, identified Lot A | Product and lot match the allocated-carton record | Identity is aligned on the documents; receiving checks still apply. |
| Broken fruit maximum 5% by weight under a named frozen-state sorting method; actual percentage required | Conforms | Request the measured percentage and confirm the method before completing this row. |
| Characteristic red colour against the agreed sensory reference | Conforms, with the stated reference and inspection method | A qualitative outcome can address this field; a numerical result is not inherently necessary. |
| Hypothetical parameter maximum 0.010 mg/kg | Less than 0.050 mg/kg | The stated reporting boundary does not establish compliance with the lower maximum. |
Visible form needs its own agreed description; an analytical result cannot settle whether the supplied presentation matches the order.
The broken-fruit entry illustrates a missing quantitative record. Ask for the original observation or an appropriately issued clarification. It would be premature to describe the lot as exceeding 5%, because that is not what the document says. It would also be premature to accept the required numerical record as complete when no percentage has been provided.
The colour entry illustrates the opposite mistake. A meaningful qualitative assessment should not be rejected merely because its outcome is expressed in words. Confirm the reference, method, assessor responsibility and scope required by the programme. The goal is evidence that fits the field, not identical formatting for every row.
Visible presentation remains a separate check. A photograph can show whole pieces, exposed flesh or mixed shapes; it cannot establish an analytical value from appearance. Conversely, a suitable analytical result does not settle a disagreement about the ordered cut. Keep photographs, defect observations and laboratory results connected to the same material while preserving what each can demonstrate.
For the hypothetical chemical row, send the exact limit and reported boundary to the qualified reviewer. The question is whether the current evidence can resolve the acceptance criterion. That review may identify a more appropriate report, a missing underlying value, a suitable additional test or an error requiring amendment. The sales team should not convert “less than” to zero to make the comparison fit.
Resolve discrepancies before the acceptance decision
A useful clarification request is short and specific. Name the order, product, lot, specification revision, report number and affected parameter. State the discrepancy in neutral language and identify the record or explanation needed. This gives the supplier and laboratory enough context to investigate without asking them to reinterpret the entire order from an isolated screenshot.
For example: “For Lot A against specification revision 3, the broken-fruit row requires the actual weight percentage under method X. The COA reports Conforms. Please provide the underlying percentage, confirm the sample basis and identify any supporting inspection record.” This is a fictional wording example. Substitute the real identifiers and agreed method rather than creating them after the product arrives.
Assign an owner to each unresolved item and record the response. A coding mismatch, missing result and confirmed out-of-specification finding need different actions. The authorised acceptance decision should reflect the specific issue, available evidence and applicable procedure. Procurement can coordinate the commercial consequences, but an unresolved safety concern needs the responsible qualified review.
When a report is amended, obtain the corrected version from its issuer and retain the superseded copy with the change explanation. Carry the original units and result qualifiers into any summary you prepare. If the supplier provides a consolidated COA, keep the supporting report references so the consolidation remains traceable.
Carry the approved document set into receiving. Check that the delivered lot codes match the reviewed allocation and that quantities split across lots remain identifiable. Review product condition, packaging and any required temperature or transport records separately. A correct pre-shipment COA does not describe an unrecorded event after its sample was taken.
For repeat orders, keep the controlled specification current and compare recurring document gaps. Repeated missing methods may call for a clearer supplier agreement; recurring near-limit results may need a quality discussion. Make those changes through the agreed approval process so the next production lot is assessed against a known requirement.
Coordinate the specification and document brief
We supply frozen fruits, vegetables and mushrooms through long-term partner factories. Our specification and sample coordination connects the required product fields with the proposed supply option, while our quality and supply coordination helps clarify the records, scope and timing needed for the order.
Send the product and form, intended use, destination, current specification, required tests, packing and quantity. We will review the document requirements with the proposed supply route and coordinate clarification of available records and outstanding fields.
References
- CFIA: Supplier Food Safety Assurance Program — product requirements, supplier agreements and verification context.
- ILAC G18:01/2024: Guideline for Describing Scopes of Accreditation — the scope of accredited activities and flexible-scope considerations.
- Global ACI technical documents — current guidance on decision rules and measurement uncertainty in testing.
- CDC: Multistate Foodborne Outbreaks — Investigation Steps — the limitations of testing only a fraction of food.
- FDA: Testing for Salmonella Species in Human Foods and Direct-Human-Contact Animal Foods — guidance on interpreting confirmed positive and subsequent negative results.
