Best before vs use by in South Africa: what a production line should check

How do best-before and use-by dates differ?

Best before refers to minimum durability: the period during which the product retains its expected qualities under stated storage conditions. R146 defines use by as the end of the estimated period under stated storage conditions after which food should not be regarded as marketable. Neither expression should be selected by the coder operator or vision recipe without an approved product specification.

This article concerns label and production control. A date alone is not an assessment of the safety of a particular pack; storage history, product condition and the producer’s instructions also matter.

What should the print specification contain?

Use the current applicable regulation and approved artwork for the particular product. The R146 source below is the gazetted text used in our evidence ledger; this article does not turn a draft replacement regulation into a rule in force.

How should an expiry-date camera decide pass or fail?

Read the printed characters and compare them with the expected string for the running SKU. Include the date prefix where it is variable, the batch and the agreed format. Obtain the expected value from controlled recipe, PLC or production data. Reading “some date” successfully is insufficient: yesterday’s date may be perfectly legible.

At a day, shift or product change, update the coder and inspection reference under one controlled procedure. Confirm the new values before releasing product. If both systems receive the same incorrect source data, their agreement does not reveal the upstream mistake.

What should the challenge test include?

• A missing code, faint print and a deliberately wrong character.
• A readable date with the wrong value or wrong format.
• A correct date on the wrong SKU.
• A changeover, midnight transition or other date-rule boundary relevant to the product.
• A rejected pack followed through to physical removal and confirmation.

Follow OCR/OCV date and batch inspection for station design and pouch inspection for flexible packaging. Marks rubbed off after the camera require packaging and durability checks beyond the image at the station.

Why a readable date can still be the wrong date

A date-code inspection has several independent layers: the approved date statement, the rule that calculates its value, the value delivered to the printer, the physical mark and the comparison performed by the camera. A failure can occur before ink reaches the package. If the printer and camera both receive an obsolete recipe, their agreement merely confirms that the same mistake reached both systems. Keep an independent first-off approval against the released product specification.

Treat the shelf-life rule as product master data. The vision engineer should not choose a shelf life or substitute best-before wording for use-by wording to make a recipe easier. Record the responsible product owner, the rule revision, the production event from which the date is calculated and any relevant time-zone or shift boundary. These are controls for implementation; the product’s validated shelf life and applicable labelling requirements remain outside the camera’s judgement.

The dated R146 text cited here distinguishes minimum durability from use-by/expiry wording and addresses batch identification and date marking in regulations 11 and 12. Use that text as a source for the discussion, while the responsible food-label specialist confirms the current applicable requirements and any commodity-specific provisions before artwork approval. A new draft regulation is not automatically a rule in force. (South African food labelling regulations R146, published 1 March 2010)

A hypothetical changeover and date-boundary test

Example, not a product shelf-life recommendation: a released packaging order specifies the text “BEST BEFORE 30 11 2026” and batch “LOT-A”. The camera should compare the printed statement and values with that authorised record. A package reading “BEST BEFORE 30 11 2025” is not acceptable merely because every character is sharp. A package carrying the expected date but batch “LOT-B” also fails the defined order match.

Run the same test across the actual production boundary that changes the expected data. A line that continues past midnight may follow a production-order rule rather than the workstation’s local calendar. Restart after a communication loss and verify which approved record is restored. Reprinting a label must preserve the original batch association unless the authorised procedure explicitly changes it. None of these decisions should be improvised by a camera operator during a fault.

Where the package also carries a GS1 symbol, keep the human-readable text and encoded dates consistent with their respective formats. GS1 General Specifications section 3 distinguishes date and batch Application Identifiers; a successful decode does not prove that the selected identifier or value is correct for this product. Do not compare the raw displayed text with a decoded data string without accounting for separators and field definitions. (GS1 General Specifications, Release 26.0)

Control legibility, comparison and release separately

Inkjet dots, thermal-transfer voids, condensation and curved film can cause character confusion. First stabilise the image: locate the print region relative to the package, manage glare and choose exposure for motion. HIKROBOT’s lighting glossary describes geometries that change reflected contrast, while its VisionMaster material describes identification and inspection tools. Those capabilities still require a trial on the actual ink and substrate. (HIKROBOT Light Source Catalog, 2025 Q3, glossary) (HIKROBOT VisionMaster product page)

Separate OCR from comparison in the result record. Save what the tool recognised, the expected value and the reason for rejection. An “unreadable” result points toward image or print quality; a “readable but wrong” result points toward data or changeover control. Combining them under one reject counter prevents maintenance and quality from identifying the source of repeated losses. Also verify that a mark accepted at the camera remains visible after downstream folding, rubbing and cooling.

After a failed challenge, identify the potentially affected interval from the last valid check, then hold and investigate that product according to the plant procedure. Do not resume simply because the next sample reads. Preserve the faulty sample and the recipe revision; determine whether the problem was wrong expected data, poor printing, unstable imaging or failed physical rejection. Restore the approved configuration and repeat the relevant challenge before release.

Acceptance should include ordinary good packs and specific wrong-date, wrong-batch, missing-print and changed-format packs. Report false accepts and false rejects separately. NIST’s proportion guidance explains why a finite successful trial cannot establish a zero future failure rate. (NIST e-Handbook, 7.2.4.1 Confidence intervals for a proportion) This article defines the date-control logic. The linked OCR/OCV application guide covers camera and reject-system design, while the producer remains responsible for shelf-life validation, storage instructions and release authority.

From the approved date to the label on the line

Once the quality team has approved the date statement and shelf-life rule, define how the printed label will be checked and how failed trays will be handled.