On this page
A frozen-food tolerance is usable only when it states what is measured, in which product state, on what sample and denominator, by which method, and what happens at the boundary. “Few broken pieces” cannot do that work. Neither can “10 mm, typical” if one person treats 10 mm as a cutting target and another treats it as the largest acceptable dimension.
Start by separating three things: the product you want, the measurement used to describe it, and the rule used to accept a lot. Put the nominal target in its own field. Give acceptance limits their own units, inclusive or exclusive boundaries, sample basis and decision rule. Attach the current method and any controlled visual reference. Then identify who may approve a deviation, for which lots, before the product moves.
We find it more useful to challenge one disputed clause with a real sample than to collect another page of adjectives. If two people can follow the written instruction and still count different objects, discard different material or inspect different states, the clause needs repair before the limit is negotiated.
A usable clause defines a test and a decision
Consider “minimal fragments.” The uncertainty begins before anyone chooses a percentage. Does a fragment mean a piece below a specified dimension, a detached portion of an otherwise whole fruit, or material passing through a defined screen? A small intact strawberry might be undersize without being broken. A large broken piece might satisfy a dimensional threshold while failing a whole-fruit requirement. Treating these categories as interchangeable changes the result.
A practical clause has a defined object, a property, a measurement instruction and an action. For fragments, name the product form and the fragment definition; state whether the result is a count fraction or a mass fraction; reference the sample selection and sorting method; give the acceptance limit; and identify the disposition of a result outside it. A document can keep these details in linked appendices, but the clause must point to the applicable versions.
Conceptual measurement sequence: identify the sample and method, record the result, then apply the agreed limit and review rule. The instruments and blank record do not show a measured lot or an approval.
The commercial discussion becomes clearer when each missing field is visible. Procurement can explain why the property matters to a filling, a retail bag or a portioning system. The producing facility can assess whether the proposed form and tolerance are feasible. QA can decide whether the measurement will distinguish acceptable product from the condition the application cannot tolerate. These are connected judgments, and none is replaced by a signature on a vague specification.
Use a method reference rather than trying to repeat a complete laboratory procedure inside every table row. The row still needs enough information to identify the correct procedure and interpretation. “Tested” is incomplete if it does not say what was tested; “conforms” is incomplete if the corresponding requirement has several possible meanings.
Name the product and the denominator
Write the product identity before the tolerance: species or product name, relevant variety, whole or cut presentation, treatment, and intended use. Where several forms can be supplied, give them different codes or clearly separated versions. A supplier’s general strawberry sheet should not silently govern both whole berries for visible decoration and pieces intended for a cooked filling.
Whole frozen strawberries show why the product form and eligible units must be identified before calculating a percentage. Appearance alone does not establish a grade or a lot result.
Then write the denominator beside every percentage. “Broken: maximum 4%” can mean broken units divided by total units, broken mass divided by total sample mass, or bags containing breakage divided by bags inspected. Those are three different statements. Define whether loose frost, thaw liquid, stems, fines or unclassifiable material remain in the denominator. Do not let the operator choose the basis after seeing the result.
In a hypothetical counting example, four classified units out of 100 units gives 4% by count. In a separate weighing example, 8 g of classified fragments in a 200 g sample gives 4% by mass. The identical percentage does not make the samples equivalent. Large pieces contribute more mass than small ones, while a count gives each eligible unit one place in the denominator. Choose the basis that represents the failure you need to control.
Named standards can demonstrate how specific that wording becomes. The USDA frozen strawberry grade description combines several characteristics and a scoring system, and expresses its whole-fruit small-size provision by count. That provision belongs to the named standard. It is not a default tolerance for every strawberry contract, and extracting a single adjective from the grade description loses its surrounding rules.
Write the sample state beside the method
Frozen, partially thawed, fully thawed, drained and cooked are not interchangeable sample descriptions. Ice can obscure a surface. Thawing can change handling behavior and release liquid. Cooking adds another preparation history. If the acceptance clause concerns shape after a particular process, an incoming frozen measurement alone does not answer the application question.
Visible crowns and stems are different measurement features. State the landmarks and sample condition before comparing floret dimensions; this photograph is not a measured tolerance result.
Specify the state required for each property, rather than assigning one state to the entire document. For a floret measurement, define the landmarks: crown width, total length, stem length or another stated dimension. Explain orientation and whether the operator measures the largest dimension or a prescribed axis. For texture after preparation, reference the actual preparation instruction and the point at which the assessment occurs.
The CFIA procedure for processed-product grade verification illustrates this distinction. Within its Canadian grading scope, it identifies frozen-state color assessment for some fruits and different preparation needs for vegetable physical and cooked sensory factors. The lesson is to state the assessment condition explicitly. The procedure’s product-specific directions should not be converted into one universal receiving or thawing method.
Keep the initial sample condition in the record even when the method changes that condition. Note the sample identity, receipt condition, preparation start and relevant handling observations. If an item arrives already partly thawed but the comparison assumes a frozen starting state, record that difference and determine whether the trial remains usable. Quietly refreezing it to resemble the reference changes the history without resolving the discrepancy.
Separate a target from an acceptance limit
A target expresses the result the process is intended to approach. An acceptance limit defines a boundary used for the agreed decision. An acceptable range has two boundaries. A typical result describes observed or expected performance under a stated basis; it does not automatically create a guarantee. Give these terms separate columns so the recipient does not have to infer which role a number plays.
For example, a hypothetical fragment specification could have a 2.0% process target and an inclusive maximum of 4.0% by mass under its agreed method. Under a simple comparison rule, 3.8% lies below the maximum, 4.0% equals it and 4.2% exceeds it. All three are above the target. A result can therefore miss the target while still meeting that particular acceptance limit. This example explains contract wording only; the numbers are not recommended tolerances or a complete lot-release plan.
Illustrative clause only: a 2.0% target and inclusive 4.0% maximum serve different purposes. The example values explain boundary wording and are not recommended product limits or a complete lot-release decision.
Write “not more than” or an unambiguous mathematical symbol when equality is intended to pass that field. Write the rounding convention and the result precision used in the decision. If a value is reported to one decimal place, clarify whether the comparison uses the underlying unrounded result or the displayed number. Otherwise a dispute can arise from presentation rather than product difference.
Do not confuse this commercial tolerance with measurement uncertainty, an instrument’s resolution, or a process-control warning threshold. They serve different purposes. Nor does a commercial concession override an applicable safety or legal requirement. Keep mandatory requirements identified separately, so an authorized quality deviation cannot be mistaken for permission to disregard them.
Make the method reproducible between locations
A method identifier is useful only if both parties can obtain and follow the same version. Include the reference and revision, sample preparation, equipment or measurement principle, calculation, result units and reporting requirements. Where an internal method is used, share enough controlled detail for the agreed inspection party to apply it. “Supplier method” without an accessible instruction leaves the buyer unable to reproduce the comparison.
Before committing to a narrow tolerance, trial the measurement itself. Ask two trained operators to classify the same type of borderline piece independently, then compare their classifications. For dimensional work, compare the chosen landmarks. For a screen-based separation, check that the opening and handling instructions are specified. When the two locations disagree, locate the differing step instead of averaging incompatible methods into a reassuring number.
NIST’s measurement terminology distinguishes repeatability under the same measurement conditions from reproducibility under stated changed conditions. In practical terms, repeated results from one operator on one instrument do not establish agreement between two sites using different preparation instructions. Describe which conditions were held fixed and which changed before claiming that a trial demonstrated consistent measurement.
Analytical criteria need the same discipline. BRCGS’s discussion of analytical testing identifies intended purpose, suitability for the sample matrix, detection or quantification capability, laboratory scope and interpretation as relevant considerations. A numerical limit should therefore be reviewed with the appropriate technical specialist alongside the method. Do not promise that every available test can quantify the property at the proposed threshold.
Five ambiguous clauses rewritten
The examples below show the information a clause needs. Set the numerical limits from the product’s requirements and select sample sizes through the agreed sampling plan. Attach your controlled methods and complete the product-specific values before approval. A working draft with an unresolved field should remain visibly unapproved.
| Ambiguous wording | More complete clause structure | Decision still to agree |
|---|---|---|
| Small amount of broken fruit | For the named whole-fruit product, classify broken units using the attached definition and reference set. Report the specified count or mass fraction under the identified sampling and sorting method. | Acceptance limit, treatment of fragments and unclassified material, sample-to-lot decision rule. |
| 10 mm dice, typical | Nominal cut target: 10 mm. Measure the specified dimensions in the stated sample condition. Record the distribution and out-of-range fraction using the approved method. | Dimensional range, permitted out-of-range fraction, measurement landmarks and rounding. |
| No obvious discoloration | Compare the defined surfaces in the specified condition and lighting against the identified current reference set. Record each defined discoloration category separately. | Borderline classification, units or area basis, applicable limit and photographic evidence. |
| Good texture after cooking | Prepare the stated sample quantity by the approved recipe and assess the named texture attributes at the defined stage against the retained reference or agreed assessment scale. | Preparation parameters, assessor agreement, acceptable outcome and reference replacement. |
| Same as approved sample | The identified sample supports approval of the listed attributes under specification revision and approval date. Analytical, packing and release requirements remain those separately stated in the specification. | Approved scope, sample storage and life, change triggers, approver and affected order. |
A photograph is valuable beside a defect definition, but its job must be explicit. An overview shows product form; a close-up can show the boundary of a defect category. Include a scale when size matters and use consistent preparation and lighting when color is compared. A single attractive bowl cannot establish the distribution throughout a lot. A photograph also cannot demonstrate a property that the camera does not measure.
When a reference set includes acceptable, borderline and unacceptable examples, label the category and the feature being judged. Do not rely on an unexplained red circle or a green tick. Retain the actual file version with the method so a later crop, compressed message attachment or altered display does not silently become the new reference.
Keep approval versions and exceptions connected
Give the specification a unique identifier, revision, issue date, approval status and effective scope. List the changed clauses rather than asking the receiving team to compare two long documents line by line. Identify whether the new revision applies to a new order, a future production date, named lots or existing stock. A revised document does not answer that question on its own.
Conceptual document-control relationship: keep the current revision, sample identity and affected lot connected. The previous version remains separate; no real customer document or approval is pictured.
Link sample approval to those same details. Record the sample identifier and stage, attributes reviewed, comments, approver and date. A visual approval of a pre-production sample should not become an implied approval of a later batch’s analytical results. The separate guide to comparing a COA with a product specification addresses the evidence needed when a lot result is checked against the approved requirements.
A deviation needs its own bounded record: the clause involved, measured difference, reason, affected quantity and lots, intended disposition, conditions, approver and expiry. State whether further processing, segregation or another approved action is required. A concession for one delivery must not quietly revise the standing specification. Conversely, a permanent accepted change belongs in a formally revised specification rather than an endless series of exceptional emails.
Agree document precedence as well. The purchase order, specification, drawing, method, sample approval and packing artwork may each contain a relevant instruction. If two conflict, identify the controlling document and resolve the inconsistency before production or release. Forwarding both files without a written resolution passes the ambiguity to the next person.
Resolve a borderline result before release
First check whether the two results describe the same property on the same basis. A buyer may be reporting drained mass while the supplier is reporting frozen mass; one operator may include a detached stem while another removes it. Repeating both tests several times will not resolve a disagreement built into the definitions.
Next check sample identity, selection, handling and method version. Separate variation between product samples from variation in measurement. Preserve the original report and observations. If a repeat is justified, state why it is being performed, which material will be used, who will witness or conduct it, and how the repeat result affects the original decision. Testing repeatedly until a favorable result appears is not a defined acceptance procedure.
Write the decision rule before such a case occurs, including the agreed treatment of uncertainty where relevant. A report close to a limit may require technical interpretation rather than a fresh tolerance invented during a shipment argument. Keep the material’s status clear while the issue is resolved, and identify who can authorize movement. A field-level result is only one part of release: the applicable identity, safety, packing and document requirements still have to be satisfied.
A final specification challenge before the order
Give the document to someone who did not write it and ask them to explain exactly what they would do with an identified sample. They should be able to identify the applicable product and revision, select the correct instruction, prepare the sample, calculate the result and find the decision rule. Every point that requires an unwritten assumption is a useful repair item.
Then challenge the edges. Which category contains a small intact piece? What happens exactly at the maximum? Does a result from one bag control the whole lot, or is a separate lot rule specified? What happens if the approved sample deteriorates? Who may replace the visual reference? The exercise tests whether the document will still work when the product is close to a boundary and the original author is unavailable.
Finally, check that each demanding field serves an application need and can be verified by the agreed method. An unnecessarily narrow cosmetic tolerance can limit supply without improving your finished product. A broad tolerance can conceal the failure that causes dosing interruptions or poor appearance. The useful balance comes from application trials and a shared measurement basis, followed by a clear written decision.
Review your specification with XMG Food
We supply frozen fruits, vegetables and mushrooms from China through long-term partner factories. We can review the product and form against your draft, coordinate samples for the attributes you need to assess, and connect the agreed inspection requirements to the proposed order. The production facility or agreed inspection party performs the checks within the confirmed scope.
Our specification and sample-approval process keeps the product, sample stage and approved fields connected. Our inspection-scope coordination helps establish who checks each requirement and what record is needed. Send the product and form, intended use, destination and the clause you are uncertain about. We will review the open fields and coordinate suitable supply and sampling options.
References
The linked USDA and CFIA resources illustrate defined grading and inspection approaches within their stated product and jurisdictional scopes. NIST supports the distinction between measurement conditions, and BRCGS discusses selecting and interpreting analytical tests. The numerical examples and clause structures above are illustrative purchasing tools, not a substitute for a product-specific approved specification, applicable legal requirements or a qualified analytical method.
