Menu
Book a demo
What Is OCR in Manufacturing? OCR vs OCV, Date Codes and Read Rates

What Is OCR in Manufacturing? OCR vs OCV, Date Codes and Read Rates

How OCR and OCV check date and lot codes on a packing line, what a wrong use-by date costs, the read rate a line needs and a decision table by station.
What Is OCR in Manufacturing? OCR vs OCV, Date Codes and Read Rates

In short: OCR (optical character recognition) in manufacturing is a camera and software reading printed characters on a pack, label or part, such as a use-by date, a lot code or a serial number, and turning them into data. OCV (optical character verification) is the stricter job on a packing line: it compares each print with the string the line should have printed and rejects packs that differ. Grading is a third job, for barcodes: it measures print quality the way a customer's scanner will see it.

  • For date and lot codes, use OCV and feed it a date computed from the order and the product's shelf life. A camera that checks the print against a mistyped entry passes the mistake.
  • A use-by date printed too late is a recall, not a cosmetic fault: under UK and EU food law, food past its use-by date is deemed unsafe. Two recall notices from Food Standards Scotland, in 2024 and 2026, had the month wrong by one and by two months.
  • Example plant: a wrong month found 2.5 hours after it was keyed puts $13,860 to $27,615 of sliced ham on hold. If it reaches the shops, the run is worth $30,492 to $60,753.

Guide, updated 16 November 2026. Sources are listed at the end.

On a food or FMCG packing line, the printed date and lot code are the record the rest of the supply chain works from. Warehouses rotate stock by them, shops mark down by them, and a recall notice names product by them. This guide explains what OCR and OCV do at a coder, what each one can and cannot catch, which read rate and checks a line needs, and what a wrong code costs, with a worked example that ends in money. Code checks are an inspection job, done pack by pack. Counting output for OEE is a different camera job, and the trade-off between the two is set out in label and code vision vs an OEE camera.

What OCR, OCV and grading mean on a packing line

OCR reads. A camera takes an image of the printed area, software finds each character and returns text, such as "L2731 USE BY 23 MAR". It does not need to know the answer in advance. That makes OCR the tool when the text varies and has to be captured: logging the lot code of each case, or reading a supplier's code on an incoming pallet.

OCV checks. It starts from the string the line is meant to print and asks whether each pack carries it, complete and legible, in the right place. It passes or rejects the pack. Because it only has to confirm a known string, it is the check to use for date and lot codes, where the right answer for the current run is known before the first pack is printed.

Grading measures. A barcode verifier scores the print quality of a symbol against an ISO/IEC standard, so a plant knows whether a retailer's scanner is likely to read it weeks later. For 2D codes the standard is ISO/IEC 15415. A barcode reader on the line answers a narrower question: could this code be read here, today.

The three jobs often share one station, sometimes one camera: a reader for the barcode on the case label and OCV for the printed date beside it. They still answer different questions, and the proof that each one works is different.

OCROCVBarcode grading
Question it answersWhat does the text say?Is this the text we expect?How good is this barcode's print?
Needs the expected stringNoYes, for each runNo; it grades against a standard
OutputText, or a no-readPass or reject for each packA grade for each symbol
CatchesMissing or unreadable text, and it leaves a record of what was printedMissing, broken, smudged, misplaced or wrong charactersWeak contrast, ink spread and damage a scanner will struggle with
Cannot catchA wrong string that reads cleanly, unless the text is compared with somethingAn error that is also in the expected stringWrong data in a well-printed code
Where it sitsAfter the printer, or at goods-inAt the printer, on each packUsually on quality's bench, on samples; inline verifiers exist but may not test every parameter
Proof that it worksRead rate on your own packs and fontIt rejects a challenge pack, and its false reject rate on good packs is knownA grade from a verifier, to ISO/IEC 15415 for 2D codes
Where codes are printed and checkedA chilled packing line, three code points1. Lidding film printerDate and lot on every packOCV on eachpack, reject2. Case labellerGS1 label: AI 17 and AI 10Line reader+ QA verifier3. Pallet at dispatchPallet label to the orderScan atdispatch1.2 mmmin x-height,EU and UK dateYYMMDDGS1 AI 17,use by4 to 5 hcode change,US canned foodReg. 1169/2011; GS1 AIs; 21 CFR 113.60Feed the camera the date fromthe order, not from a keypad.
Example of a chilled packing line. Date rules from Regulation (EU) No 1169/2011 (UK text), the GS1 application identifiers and 21 CFR 113.60. See the source.
Track OEE and downtime in real time
Book a demo

Why the date and lot code carry so much weight

A date code is a legal label element, and the rules behind it decide what a camera has to check. The food information regulation, Regulation (EU) No 1169/2011, applies in the EU and, as amended, in Great Britain; the articles below are quoted from the UK text on legislation.gov.uk.

  • The date is a mandatory particular. Article 9(1)(f) lists the date of minimum durability or the use-by date among the mandatory particulars, subject to the Regulation's exceptions (Annex X, for example, exempts fresh whole fruit and vegetables, wine and salt from a best before date).
  • It must be legible at a set size. Article 13 requires those particulars to be printed with an x-height of at least 1.2 mm, or 0.9 mm where the largest surface of the pack is under 80 cm². It also asks for them to be clearly legible and, where appropriate, indelible.
  • A use-by date is a safety limit. Under Article 24, foods that are highly perishable from a microbiological point of view carry a use-by date instead of a best-before date, and after the use-by date a food is deemed to be unsafe.
  • The wording and order are fixed. Annex X says "Best before" is used when the date includes the day and "Best before end" otherwise. The date is the day, the month and possibly the year, in that order and in uncoded form. For a best before date, day and month are enough on food that keeps up to 3 months, month and year for 3 to 18 months, the year alone beyond that. A use-by date goes on each individual prepacked portion.
  • A lot must be shown, and the date may be the lot. Directive 2011/91/EU on lot marking says the producer, manufacturer or packager, or the first seller established in the EU, determines the lot, and that its mark is preceded by the letter L unless it is clearly distinguishable from the other indications (Article 3). Article 5 lets the lot mark be left off where the label carries a date of at least the uncoded day and month, in that order. On a pack that relies on Article 5, one printed field does two legal jobs.
  • US canned food codes are prescribed. Under 21 CFR 113.60(c), each hermetically sealed container of low-acid processed food carries an identifying code, permanently visible to the naked eye, that identifies the establishment, the product, and the year, day and period of packing. The period code changes at intervals of 4 to 5 hours, at shift changes, or by batch, as long as a batch does not run over more than one shift.
  • Canada writes the date the other way round. The Canadian Food Inspection Agency's date marking rules require a best before date on prepackaged food with a durable life of 90 days or less that is packed away from the store that sells it, under the words "best before" and "meilleur avant". The year comes first, if shown, as at least its last two digits; then the month, in words in both languages or as a bilingual symbol (JA, FE, MR, AL, MA, JN, JL, AU, SE, OC, NO, DE); then the day. Expiration dates are required only on foods for special dietary use, infant formula and human milk fortifiers.

The barcode carries the same facts in a third form. In the GS1 application identifiers, AI 17 holds the expiration date, titled "USE BY or EXPIRY", and AI 15 the best before date, both as six digits, YYMMDD. AI 10 holds the batch or lot number, up to 20 characters. GS1's 2D barcodes at retail point-of-sale guideline (Release 1.1.0, ratified December 2025) describes the industry goal, Ambition 2027, of tills reading the GTIN from linear and 2D barcodes by the end of 2027. Among the benefits of 2D codes that carry such data, it lists tills that block the sale of out-of-date product, date-based markdowns and more targeted recalls. It also notes that a 2D code carrying a batch number or best before date may have to be printed in real time on the production line, and that pre-printing the packaging may then not be possible.

So the code a plant prints is read by the shopper, by the retailer's warehouse, by the till and, on a bad day, by a regulator. For how a lot differs from a batch in the plant's own records, see batch vs lot production.

One date, three formats: what the camera must expect

A chilled product made for two markets can carry the same use-by date in three different strings. The examples use 23 March 2027 and are example data.

Where the date appearsRuleThe same date, writtenWhat the check must know
EU or UK pack, use-byAnnex X: "use by", then day, month and possibly year, in that order, uncodedUSE BY 23 MAR 2027The words, the field order, and whether the month is digits or letters
Canadian pack, best beforeCFIA: year first, shown when needed for clarity, then bilingual month symbol, then dayBEST BEFORE / MEILLEUR AVANT 27 MR 23Year first, and a month from a fixed list of twelve letter pairs
GS1 barcode on the caseAI 17, YYMMDD(17)270323Six digits, year first
US low-acid canned food21 CFR 113.60(c): coded establishment, product, year, day and periodThe plant's own codeThe plant's decoder, including the period code

Three things follow for any code check.

  1. Build each expected string from one date. The plant holds one fact, the use-by date of the run, and prints it in up to three formats. Generate every OCV template from that one date. Typing three strings by hand gives three chances to differ.
  2. Limit the characters by field. A day is two digits, a Canadian month one of twelve letter pairs, an AI 17 date six digits. Telling the software which characters can appear in which position removes confusions such as 0 and O, or 8 and B, from fields where they cannot occur. Watch the pairs that remain: AL is April and AU is August, so a smudged second letter moves a Canadian date by four months.
  3. Size the camera for the smallest legal print. If a pack prints its date at the 1.2 mm floor, or 0.9 mm on a small pack, the pixel on the pack must be fine enough for the strokes, and dot-matrix inkjet characters have gaps between dots. For example, a field of view 80 mm wide across 1,280 pixels gives 0.0625 mm per pixel: about 19 pixels across a 1.2 mm x-height and 14 across 0.9 mm (example data). Ask each supplier how many pixels their tool needs on your font, then test it on your smallest pack.

Case: two wrong use-by months, January 2024 and October 2026

Food Standards Scotland (FSS) publishes the recall notices that apply in Scotland. Two of them show the failure an OCV project exists to stop, and where a camera on its own stops short.

NoticeProductUse-by printedCorrect use-byError
FSS-PRIN-003-2024, 19 January 2024Aldi Specially Selected 2 Haggis En Croutes, 400g23/02/202423/01/2024Month one too late
FSS-PRIN-037-2026, 22 October 2026Waitrose Tuna Pate, 115g17 December 202617 October 2026Month two too late

Source: the two FSS notices. The haggis notice warns of possible bacterial growth after the correct date; the pâté notice says the product is unsafe to eat after the correct date. Neither notice says how the wrong date was set, and nothing here is a judgement on either company. A third FSS notice, of 16 October 2025, shows another kind of date error: YSSS Food recalled seven cheesecakes that, besides gaps in safety controls, were labelled with a best before date where a use-by date was needed.

What the case teaches about code checks

  • The day was right and the month was wrong. 23 stayed 23 and 17 stayed 17. Both errors sit in the month field. Neither notice says why; our reading is that a field which changes rarely on a dated run is easy to set once and stop reading. An OCV template filled from the same entry as the printer passes every pack. A template checked against a date computed from the production date and the product's shelf life stops the first one.
  • Both dates moved later. The haggis showed a month of extra life and the pâté two. Both notices say the food may be, or is, unsafe after the correct date, so both errors ran in the dangerous direction. A date printed too early costs shelf life; a date printed too late costs a recall. If a plant adds only one rule, make it this one: reject any date later than the computed date.
  • A legible wrong date passes print checks. Neither notice mentions faint or broken print. 23/02/2024 would have read perfectly, and a barcode carrying the same wrong date in AI 17 could still grade well. Print quality and correct data are separate checks.
  • The words are part of the code. The cheesecake notice is about the label saying best before instead of use by. Annex X ties each wording to a kind of date, so put the words in the template, not just the digits.
  • The shelf searches by date. Both notices name the product, the pack size and the date, and neither lists a batch code. Where a date stands in for the lot, as Article 5 of the lot directive allows, the date is what the recall works from. Record which run printed which date, so the plant can name every run that carried the wrong one.

The next case will be added under this one, with its date in the heading.

Where codes are applied and checked on a packing line

A chilled or ambient food line can put codes on three levels. Each level has its own printer, its own typical fault and its own check.

  1. The primary pack. A thermal transfer printer on the lidding film, an inkjet on a tray, a can end or a bottle neck, or a laser on a coated surface. This is where the use-by date and the lot sit for the shopper. The check is OCV on every pack, after the print and before the pack can mix with others, with a reject that removes the pack and a rule that stops the line after a set number of rejects in a row.
  2. The case. A print-and-apply labeller or a case printer adds the GTIN, the date and the lot, as text and as a GS1 barcode. The check is a reader on each case, comparing the data with the order, plus a verifier sample by quality at start-up and after each label roll or ribbon change. The labeller's own losses are covered in labeller OEE and labelling machine losses.
  3. The pallet. A pallet label, scanned at dispatch against the order, so the warehouse system knows which dates went to which customer.

Two moments deserve the closest watch. The changeover, when the date, the product and sometimes the market change at once. And the first minutes after a consumable change, when a new ribbon or ink cartridge starts to print. Put the code check into the first off approval after changeover, with a printed pack signed against the order, and log coder faults as stops with their own reasons, as described in date coder printer stops.

Read rate, false rejects and the challenge pack

A supplier will quote a read rate. On a packing line three numbers matter, and one of them cannot be measured from production at all.

  • Pass rate on good packs. The share of correctly printed packs the system accepts. The rest are false rejects: good product in the reject bin.
  • False accepts. Wrong or broken prints the system passes. They cannot be counted from production, because nobody knows which passed packs were wrong. They are measured with a challenge set: packs with a missing character, a smudged digit, a blank, yesterday's date and the other market's format.
  • No-reads. Packs the system cannot decide on, often because of glare, a pack out of position or a print outside the window. Decide in advance what a no-read means. On a date code it should mean a reject.

The authors name high variability across sites as a main challenge. In a 2020 paper by researchers at the National Technical University of Athens and the University of Lincoln, a model judging the quality of printed use-by dates was trained and tested on more than 30,000 images from six locations in the UK. The images varied in distortion, background, light and blur, date format, and the angle and orientation of the label. Adapted to an unlabelled new set of images from one labelled set, the model reached 84.14% accuracy; adapted from two labelled sets, 90.53%. So a trial on a supplier's samples is not enough on its own (our advice, not the paper's): a line needs its own acceptance test, on its own packs, fonts and films.

Then do the sums at your own speed. At 4,200 packs an hour, an OCV that rejects 0.3% of good packs sends 4,200 x 0.003 = 12.6 packs an hour to the bin, 189 in a 15-hour day. If rejects are scrapped, at the low value of $1.32 a pack used in the worked example below, that is 189 x $1.32 = $249 a day. A check so tight that operators stop trusting it is worse than its scrap: people start feeding rejects back without a look, and the check is gone.

Three habits keep the numbers honest.

  1. Run the challenge set at each start-up and changeover, and record that each challenge pack was rejected. A check that has not rejected a bad pack today has not shown that it works today.
  2. Treat a no-read as a reject, and count no-reads apart from rejects, because a rising no-read count can come from the light, the lens or the pack position as well as the printer.
  3. Verify case barcodes on a schedule. GS1's guideline calls verification highly recommended to confirm a barcode's quality and how likely it is to be read as intended.

Choosing the check for a code station

Each kind of check covers a different set of faults. The matrix sums it up; the table below gives the detail.

What each check catchesFive faults, four ways to check a date codeEyecheckOCVon jobOCVon orderBarcodeverifierMissing printPartYesYesNoFaint or brokencharactersPartYesYesNoWrong date inthe job entryPartNoYesNoWrong formatfor the marketPartNoYesNoBarcode printbelow gradeNoNoNoYesPart: only at check times, and only if thechecker compares with the specificationOCV is as right as its expected string
Our assessment of how each check works. "On job": the expected string comes from the printer's job entry. "On order": it is computed from the order date and the item's shelf life.

The cost column gives a sourced anchor, not a quote. No independent price survey for code vision systems was readable for this guide, and quotes depend on optics, integration and validation. What can be sourced is the value of the packs a gap in checking lets through. At the Example plant's rate of 4,200 packs an hour, each hour of output that a check leaves unchecked is 4,200 x $1.32 = $5,544 to 4,200 x $2.63 = $11,046 of sliced ham (the unit values are explained in the worked example).

OptionGood atMissesCost anchorPick it ifQuestions to ask
Eye check by the operator, logged at set timesAny format, judgement, slow linesEverything between checks; a wrong date copied from the job sheetEach hour between checks: $5,544 to $11,046 of packs at the Example plantSpeeds are low, few products run, and the checker compares with the specification, not with the printerAsk yourself: against what is the print compared, and how long ago was the last logged check?
OCV at the printer, expected string from the printer's jobMissing, faint, broken, smudged or misplaced print, on every packA wrong date, wording or format that is in the job itselfSupplier quote, set against the hour value above for print faultsDate entry is already controlled, for example by a second person against the order, and print quality is the risk leftWhat false reject rate on our packs at full speed? How is a challenge pack run? What happens after rejects in a row?
OCV with the expected string from the order and the item masterAll of the above, plus a wrong date, month, wording or market format, at the first packWrong master data, such as a wrong shelf life on the itemThe same quote plus integration; it also closes the window for keying errorsThe product carries a use-by date, the line changes over several times a shift, or one product ships to markets with different formatsCan the printer and the camera both take the date from the order system? Who owns the shelf life of each item?
OCR captureA record of what was printed, by time; strings not known in advance; supplier codes at goods-inNothing on its own: a read is not a decision until it is compared with somethingSupplier quote; the return is trace time savedYou need a log of printed codes for each run, or must read codes you did not printWhat read rate on our font and film? Where is each read kept, and for how long?
Barcode reader on the lineEach case label read, with its AI 17 date and AI 10 lot compared with the orderPrint quality drifting toward a fail at the customerSupplier quote; set it against a rejected delivery at the customer's dockCases carry GS1 barcodes with dates and lotsDoes it compare the data with the order, or only report a read?
Barcode verifier on samplesA grade for each symbol, to ISO/IEC 15415 for 2D codesEvery unit between samples, and all printed textSupplier quote for a bench unitA retailer asks for a grade, or 2D codes with dates go to the tillWhich grade does the customer require, and how often should quality sample?

Then read off the verdict for the station.

  • Choose OCV fed from the order and the item master if the product carries a use-by date, the line changes over several times a shift, or one product ships to markets that write dates differently.
  • Choose OCV fed from the printer's job if date entry is already controlled elsewhere and the risk left is print quality.
  • Choose OCR capture if the plant needs a record of what was printed on each run, or has to read codes it did not print.
  • Choose a line reader plus a verifier sample if cases or packs carry GS1 barcodes with dates, above all 2D codes headed for tills.
  • Keep a logged eye check if the line is slow and runs few products, and make the checker compare the print with the specification, not with the job sheet.

Worked example: a wrong month on a sliced ham line

Example plant. Every figure in this section is example data except the two unit values, which are sourced.

A chilled meats plant slices cooked ham and packs it into 8 oz (0.5 lb) packs on one thermoforming line, at 70 packs a minute, or 4,200 an hour. A thermal transfer printer puts "USE BY", the date and a lot code on the lidding film. There is no OCV yet. The operator checks a pack every hour against the job sheet, and quality checks one every 2 hours against the product specification.

At a 10:20 changeover the operator keys the use-by date for the next product. The day is right and the month is one too high, the same kind of error as in the case above. The operator's checks at 11:20 and 12:20 compare the print with the job sheet, which carries the same entry, and pass it. Quality's check, due at 12:20, runs late and finds the error at 12:50, against the specification.

What happensWindowPacks
OCV fed from the order: the line stops after 3 rejects in a rowThe first packs of the run3
OCV fed from the job sheetIt checks against the same wrong entry and passes every packSame as no OCV
Found in the plant by quality at 12:50 (the baseline)10:20 to 12:50: 2.5 hours x 4,20010,500 on hold
Missed, and the whole run ships10:20 to 15:50: 5.5 hours x 4,20023,100 to withdraw

The value of a pack, from two sources

  • Low: what prepared ham fetches in trade. UN Comtrade, the United Nations trade database, reports US exports of prepared or preserved hams and cuts (HS 1602.41) in 2025 at $121,456,212 (FOB) for 20,851,272.58 kg, a weight Comtrade marks as estimated. That is $5.82 a kg, or $2.64 a lb (our arithmetic). A 0.5 lb pack: $1.32.
  • High: what it sells for in shops. The US Bureau of Labor Statistics average retail price for boneless ham, excluding canned, was $5.260 a lb in October 2026, US city average. A 0.5 lb pack: $2.63.
OutcomePacksAt $1.32At $2.63
OCV fed from the order3$3.96$7.89
Found in the plant at 12:5010,500$13,860$27,615
Missed and shipped23,100$30,492$60,753

The low value fits product still in the plant, which is written off or repacked at cost. The high value fits product pulled back from shops, before freight, retailer charges and disposal, so it is a floor for a withdrawal. Both are revenue values, not margin, and neither is a figure for any company.

Against the baseline, a check fed from the order keeps 10,500 minus 3 = 10,497 packs off hold: 10,497 x $1.32 = $13,856 to 10,497 x $2.63 = $27,607 for one event on one line. At four such keying errors a year, that is 4 x $13,856 = $55,424 to 4 x $27,607 = $110,428. A camera fed from the job sheet saves none of it, because the error is in the sheet.

The same event shows up in OEE. In this guide's convention, the 10,500 packs are quality loss for the shift until quality releases or scraps them, and the 30 minutes the line ran after the late check was due are the price of a check interval kept on paper.

Book a demo and see good output per line and shift.

Five ways a wrong code still ships

  • The camera checks the job, and the job is wrong. If the expected string is typed into the same screen as the printer's date, the camera confirms the typing. Move the source of the date to the order and the item master.
  • The challenge pack is never run. A check that has not rejected a bad pack since its last software change may be passing everything. Run the challenge set at start-up and changeover, and keep the record.
  • Rejects find their way back. A full reject bin, emptied onto the conveyor in a hurry, undoes the check. Lock the bin, and count what goes in.
  • The pack and the case disagree. The pack is printed from one entry and the case label from another. Compare the AI 17 date on the case with the pack date of the run, at the reader.
  • The date rolls over mid-run. On a night shift, a printer that takes the production date from its clock changes the use-by date at midnight while the run goes on. Decide whether the date follows the clock or the run start, write it into the specification, and test it on the first night run after any change.

Code rejects and coder stops in OEE

A code check protects the customer, and it also produces losses that OEE should see.

  • Rejected packs are quality loss. Each pack in the reject bin, good or bad, was made and cannot ship as it is. Count rejects with a reason for each fault type: missing print, unreadable, wrong date.
  • Coder and checker stops are availability loss. A ribbon change, a printhead clean or a stop after rejects in a row stops the line or the labeller. Give each its own stop reason.
  • In this guide, held product counts as quality loss until it is released. In the example, the 10,500 packs on hold belong to that shift's quality loss, so the cost of a late check shows on the same chart as speed and downtime.

The vision system says which pack failed; OEE says what the failures cost across shifts and products. OEE software collects run time, stops and counts for each line, and a light MES ties them to orders and products, which is also where the expected date for each run should come from. How counts from cameras at manual stations become availability and performance is covered in OEE with computer vision.

A code check is an inspection job: a verdict on each pack, inside the cycle, with a reject. Counting output for OEE is a different job, with a wider view and no verdict on the pack. Fabrico's AI cameras (an add-on) count output at stations where a machine has no usable signal. They are not an inspection system: they do not read codes, check print, recognise stops or propose reasons.

Steps for quality, production and CI managers

This week

  1. Quality manager: take one pack from each line and each market format, write the expected string from the order and the item's shelf life, and compare. Note each line where the printed date is typed by hand.
  2. Production manager: at the next changeover, watch how the date is set: who keys it, from what document, and who checks it against what.
  3. CI manager: pull last month's rejects and holds for code faults, split into missing print, unreadable and wrong date, and book each as quality loss with its own reason.

This quarter

  1. Take the date from the order. Feed the printer and the check from the order and the item master, so nobody types a date at the line.
  2. Add a challenge set to each code station: a blank, a missing character, a wrong month and the other market's format, run at start-up and changeover.
  3. Generate one template per market from one date for each run: the pack text, the Canadian format if you ship there, and the AI 17 string on the case.
  4. Add a rule against late dates: reject any printed date later than the computed one, whatever else the template allows.
  5. Verify case barcodes on a schedule, at start-up and after each label roll or ribbon change, and keep the grades.
  6. Run a trace on one date. Pick a use-by date from last month and name every run, line and customer that carried it, the way a recall notice would ask.

Where Fabrico comes in

Fabrico is a cloud OEE and light MES platform: OEE and losses by machine, line and shift, from the PLC or a Fabrico IoT device, with reasons operators add on a tablet or phone.

On manual stations, AI cameras (an add-on) count output.

Book a demo and see OEE and downtime per line and shift.

What people ask next

What is OCR in manufacturing?

OCR, optical character recognition, is a camera and software reading printed or marked characters on a pack, label or part, such as a use-by date, a lot code or a serial number, and turning them into data. On a packing line it can sit after the printer or labeller, next to OCV, which checks the print against the string the line should have printed.

What is the difference between OCR and OCV?

OCR reads text it does not know in advance and asks what it says. OCV starts from the expected string and asks whether each pack carries it, complete and legible. For date and lot codes OCV is the check to use, because the right answer for each run is known before the first pack is printed.

Can a vision system catch a wrong use-by date?

Only if the expected date comes from somewhere other than the entry that set the printer. OCV fed from the printer's job confirms whatever was typed. OCV fed from a date computed from the order and the item's shelf life rejects the first pack with a wrong day, month or year, and a rule that rejects any date later than the computed one covers the dangerous direction.

What read rate does a date code check need?

Measure three numbers: the pass rate on good packs, false accepts on a challenge set of bad packs, and no-reads. Work out the false rejects at line speed: at 4,200 packs an hour, 0.3% is 12.6 good packs an hour in the bin. Treat a no-read on a date code as a reject, and run the challenge set at each start-up and changeover.

Is barcode verification the same as reading a barcode?

No. A reader on the line shows that a code could be read there and then, and can compare its data with the order. A verifier grades the print quality of the symbol to an ISO/IEC standard, ISO/IEC 15415 for 2D codes, which tells you whether a customer's scanner is likely to read it later. GS1 calls verification highly recommended.

Does a food pack need a lot code if it carries a date?

In the EU, Directive 2011/91/EU requires a lot mark, but its Article 5 allows the mark to be left off where the label shows a date of at least the uncoded day and month, in that order. The date then does both jobs, so a wrong date is also a wrong lot. In the US, low-acid foods in hermetically sealed containers, such as cans, carry a code for the establishment, product, year, day and period of packing under 21 CFR 113.60.

Sources

Sources accessed on 16 November 2026 unless marked as accessed on another date. Dates are publication dates where a page gives one; other dates are labelled as what they are.

Track OEE and downtime in real time
See where lines lose time
Book a live demo.
See where lines lose time
See where lines lose time

Real Fabrico tiles, sample data.

By clicking the Accept button, you are giving your consent to the use of cookies when accessing this website and utilizing our services. To learn more about how cookies are used and managed, please refer to our Privacy Policy and Cookies Declaration