17 replies to this topic

[harit]

Hi everyone, Harit here. I joined yesterday, so this is my introduction.

My background is software. Labeling is what I've been learning, and I expect to be learning it for a long time yet. This looks like the place where people who do it every day talk about it honestly, which is why I joined.

Mostly I'm here to read and learn. But an introduction seems like a fair excuse to ask one thing: what makes labeling hard in practice?

I'm curious where the real difficulty sits. Whether the guidance itself is ambiguous, or whether it's clear enough and the work is just tedious and easy to get wrong. And whether, once you've settled on an answer, you feel confident it would hold up if someone asked you to defend it a year later.

Small annoyances count as much as the big ones. If there's a better board for this question, tell me and I'll ask it there instead.

Looking forward to learning from the discussions here. Thanks for having me.

 

[Tony-C]

Hi Harit,

 

 

Welcome to the IFSQN forums.

 

Labeling can be tricky because of legislative requirements (e.g. size of text and information required) and need to be approved by an authorised person at the artwork and printing stages.

 

Next is the practicalities in manufacturing, if you only run one product and have only one label then great. But, manufacturing may involve many products and hence multiple label changes.

 

Label/product changes need to be monitored closely during production to ensure that the correct label is applied and products don’t get dispatched with the wrong information such as an incorrect allergen warning. Monitoring typically includes checks at start, during and end of runs, and when there are batch/reel changes. Stock also needs to be reconciled at the start and end of a production run to ensure there isn’t an anomaly in the usage compared to the number of products packed.

 

I’m sure there are other areas where people have difficulties but I’ve changed time zones and my brain hasn’t caught up yet!

 

Kind regards,

 

Tony

[harit]

Thanks Tony. That's more concrete than what I came in with.

 

The split you drew is the part I hadn't seen. I'd been thinking of labeling as one problem: work out what the label has to say, get it approved, done. You've described a second one that runs entirely after that — the right label reaching the right product, checked at the start, during and end of a run, again at every reel and batch change, and reconciled against units packed at the end.

 

So the thing I want to ask next: when it does go wrong, which half is it usually? Artwork that got approved with something wrong on it, or correct artwork applied to the wrong product on the line?

 

My instinct says the second happens more often and the first costs more when it does. But that's a guess from someone who has never stood next to a line.

[GMO]

I'd put it at mostly errors on the line but I've experienced artwork errors at generation stage so they do happen and they do sometimes cause recalls.

 

As for costs? It depends what the error is but they can be pretty similar. The wrong approved artwork can even cost more because by the time the issue is noticed, it can span several production dates. With incorrectly applied packaging, normally it's going to be a specific shift (but not always and it can sometimes be hard to scope). So I'd say "yes" to the first but "not necessarily" to the second.

 

Tony's absolutely right. Stock reconciliation is part of it and this is often written in retailer codes of practice but have you ever tried to do it? If you have a big discrepancy, this will pick it up but a few missing labels or imprints from a reel? When you will use some for setting up? That's way harder. Most manufacturers are now using some automated on line checking of labels now in my experience. This may just be barcode, 2D barcode or might be a visual of the packaging. These systems are expensive but wrong pack is one of the leading reasons for recall so if you can find a way to do it I would. 

 

What makes it hard - artwork approval

 

You're right the work is tedious. Legislation is a little ambiguous in places, however, most companies have interpreted it sensibly in my view. Big companies in the UK (not sure where you are) can sign up to primary authority support if needed too so they can get official interpretations of legislation which are binding. 

 

But the work is dull, dull, dull. And sometimes marketeers and sales people design products with very similar looking labels. This makes it even harder if checking a lot.

 

Things which can help shift you out of the mind wandering mindset can include having different people check each version (highly recommended), printing out the physical copy to check and marking on it each aspect and having a checklist you work through. I suspect you might be alluding to this; "is this an area AI could help?" Yes. Absolutely. But be careful what it's checking. If it's checking inputs not outputs, that is still prone to error. I've known of "allergens in bold" being an input but the visual not looking bold for an unrelated reason (as an example). If you were to take out one of your human checkers and replace with AI as a starting point with human oversight for, say, version 2 and FTP, I'd be fine with that and it would give you the opportunity to assess how good it is.

 

Online application

 

Similar problems in a way. Visually similar packaging, lack of automation, high workloads and pressure on downtime (all lines should be cleared of packaging between production runs, recording use matches expected use but cutting corners there will save you time and therefore increase production but also increase risk). If you don't have an automatic checker, it's highly recommended to retain a sample of the packaging on any artwork or photograph it for electronic systems. But people do make mistakes. Some level of verification, e.g. a periodic supervisor check that operators are following the right processes and the pack is right is recommended.

[Tony-C]

 

Thanks Tony. That's more concrete than what I came in with.

 

So the thing I want to ask next: when it does go wrong, which half is it usually? Artwork that got approved with something wrong on it, or correct artwork applied to the wrong product on the line?

 

My instinct says the second happens more often and the first costs more when it does. But that's a guess from someone who has never stood next to a line.

 

 

Hi Harit,

 

IME more recalls are caused by the incorrect label being applied. The truth is I don’t think anyone really knows for sure, for example with food safety recalls the reason given may be undeclared allergen which could be either an artwork or labelling error.

 

This area is a massive problem in the Food Industry. From my analysis of recalls in 2025, close to 49% of Recalls from the US, UK, Aus & NZ combined were due Undeclared Allergens/Labelling.

 

See my Food Safety Essentials Webinar - Food Safety Review 2025 for more information.

 

Kind regards,

 

Tony

[TimG]

Backwards regulations like what we've discussed here California 'Best if used By' AB-660 Food and Beverage labeling - IFSQN can make labeling hard. There was also a pretty stupid FDA back and forth on labeling requirements for things like honey needing to list 'added sugars' on their label so that consumers could see the 'added sugars' in their diet. Guess what the average consumer jumps to when they see 'added sugars' on a label of pure honey? "OMG, IT SAYS YOU ADD SUGARS IT'S NOT PURE I'M GONNA REPORT YOU!" I get a few of those every month...

[MDaleDDF]

Hi Harit,

 

IME more recalls are caused by the incorrect label being applied. The truth is I don’t think anyone really knows for sure, for example with food safety recalls the reason given may be undeclared allergen which could be either an artwork or labelling error.

 

This area is a massive problem in the Food Industry. From my analysis of recalls in 2025, close to 49% of Recalls from the US, UK, Aus & NZ combined were due Undeclared Allergens/Labelling.

 

 

Yeah, this is kind of dicey they way they report it, eh?   The FDA cites percentages of recalls due to allergens/labeling errors, but it's kind of difficult to sus out.   Was the allergen in the product supposed to be on the ingredient listing, and is not?   Or is the allergen in the product added by accident, and was never supposed to be on the ingredient listing?    Because to me one is a label error, and one isn't, but they seem to count both scenarios a labeling issue?   Which I do get.   I mean it has something in it which isn't on the label.   But if it wasn't MEANT to be on the label or in the food, I don't personally see that as a label recall.

 

[GMO]

I have had two artwork error public recalls in my career. One mispack public recall but you only have to look at the FSA to see mispack recalls are not uncommon and probably more common than an artwork error. I agree it's sometimes hard to tell though but I read most of them as mispacks and when I've actively been involved in crises of this sort with other manufacturers it's always been completely avoidable mispacks.

[qa_maddy]

I keep having to straddle the worlds of FDA labeling guidance and USDA COOL regulations. 

yay seafood. yaaaayyyy

[liberator]

We had an incident for can artwork. It went through all the required checks, customer, regulatory QA etc, all passed. Can artwork print on tin plate, all checks good. Can plate arrives at can manufacturer, more checks, all good. Can arrives at our facility, our own checks, all good. Started to fill the cans operator noticed the ends were being put on the bottom, not the top as they should have been. Found the can artwork was upside down. Line stopped, CAR raised against the supplier

 

Supplier investigation found that when making the cans, their operator put the printed tin plate into the can former upside down. Their vision system, which is very robust, started rejecting the formed cans, as it was seeing these as defective. The operator looked at the cans, saw the artwork was correct, did not realise it was upside down, bypassed the vision system. The result a few thousand cans with upside down artwork. Their CA was to restrict the vision system bypass to QA authority only. You get so used to seeing something you start to fail to see. So even the best systems can still fail when we require on human intervention.

[Tony-C]

Yeah, this is kind of dicey they way they report it, eh?   The FDA cites percentages of recalls due to allergens/labeling errors, but it's kind of difficult to sus out.   

 

Indeed MDaleDDF,

 

The recall statistics I quoted were my own calculations based on analysis of over individual 600 recall reports from the US, UK, Aus & NZ in 2025. I excluded Canada from the final figures because they skewed the statistics due to an ongoing issue with importing Pistachios contaminated with Salmonella from Iran.

 

Close to 49% of recalls were allergen related but the cause of the recalls was either:

1. An error on the label including failure to identify allergen content in an ingredient or

2. The wrong label was applied

 

Recall data/reporting doesn’t normally provide sufficient detail to distinguish between the these possibilities but the statistics do show us that allergen management and label control are areas where attention is needed. Indeed in an SQF Review Webinar* a couple of years ago there were discussions on why this sort of thing is not picked up more often on certification audits.

 

* Tammie Van Buren, Compliance Manager, SQFI Food Marketing Institute - Food Safety Trends and Challenges - Insights from SGS and Not Your Top Non-Conformances ... But Should They Be? Links posted in this topic: SQF Certification Audit Non-Conformance Statistics 2021 2022

 

I also discussed causes and systems to prevent these errors in FSE Webinar What Food Safety Recall Data Tell Us - Part 2 (Key Food Safety Management Issues facing the food industry: Labelling and Undeclared Allergens).

 

Also note that the statistics I quoted only include allergen recalls and do not include other errors in labelling caused by anything else that is legally required on the label such as name, weight, handling instructions, cooking instructions etc.

 

And liberator, BRCGS has been quite keen on only authorised person making those sort of changes for a while.

 

Kind regards,

 

Tony

[harit]

This whole thread turned out more useful than I expected — thank you all.

 

Tony and GMO, the split you both landed on (mispacks more frequent, artwork errors sometimes fewer but longer-tailed and pricier once they happen) matches my instinct but I wouldn't have guessed the actual proportions. Tony, that 49% allergen/label figure across 600+ recall reports is a bigger number than I expected going in.

 

MDaleDDF's point about the ambiguity buried in that kind of statistic is the part I keep turning over — an allergen recall could be a labeling failure (allergen present, not disclosed) or a formulation/cross-contact failure that happens to surface as a labeling problem. Sounds like even people close to this data can't always tell the two apart after the fact, which says something about how these get investigated and closed out.

 

GMO — the "checking inputs vs. outputs" distinction (allergens marked bold as a field vs. actually rendering bold) is exactly the kind of failure mode that's invisible until it isn't. And liberator's can story is a great, concrete version of the same thing: a robust vision system doing its job, a human overriding it because the artwork looked right, and the actual defect being orientation, not content. Systems catch what they're built to catch; people stop noticing what they check every day.

 

TimG — the honey/added-sugars complaint made me laugh out loud. That's a rule that's technically correct and practically misleading, which might be its own special category of "hard."

 

If I'm hearing this right, the recurring theme isn't that anyone's guidance is wrong — it's that verification is manual, repetitive, and quietly assumes the checker's attention doesn't degrade over a shift or a hundred reels. That tracks with what GMO said about automated line checking being expensive but worth it because mispack is a leading recall cause.

 

Genuinely curious — for those of you running automated checks (barcode/2D barcode/vision), has it changed where the remaining errors happen? i.e. did it just move the failure point somewhere earlier (artwork approval, changeover setup) rather than eliminate it?

[jfrey123]

I've tried to weigh in on this thread a couple of times, but getting my thoughts out without creating a full-length college essay on the topic of labels for the OP becomes difficult. 

• Label design in of itself is a challenge, making sure you meet the sizing and margin requirements alone is a whole topic.  Then you can get into whether your claims are misleading, and then whether your sales people want to use a phrase that has specific requirements in of itself (like listing the number of grams of protein on the front vs just claiming "good source of protein").
• Then back labels become an issue, making sure your ingredients and nutrition panels are proper and up to spec.
• Then come the QA check that the label meets all the requirements you're aware of, in addition to a point of use QC check that labels were correctly selected from the approved and secure source in your facility prior to the production run.  Tracking their use in the plant, making sure copies are controlled and not thrown away where they can get stolen and used elsewhere to misbrand something...

There I go again.  The list of concerns just for a label can go on and on, and the amount of times a company I work for has been dinged over a label from either customers or GFSI auditors still gives me heebee jeebees as I type this out. 

 

Last time I had labels become a problem was in one of my USDA sites where we add RTE items into a tray to make a microwave appetizer type thing, and the sales people had us put "heat and eat" on the front label.  USDA was about 3 seconds from making us recall it after a couple of weeks because the backside instructions state to heat to 165F for food safety, which to them is a full cook step.  We lucked out by showing there was no true hazard to the public, but they still made us change the statement to "cook and serve."

 

Edit - Have to add-on one more bit:  I was part of a corporate team that visited that same USDA site back in 2023.  They were very proud to show off their label control program, including a chain link fenced area in the storage warehouse that had an electronic lock on the gate requiring a RFID card to access.  Only QA supervisors and above could get into the finished labels.  It was very impressive.  Until I grabbed a piece of wood from a broken pallet, stuck it through the fence, and triggered the manual release for the gate inside the cage.  I gained access, grabbed a roll of labels, and walked them into production and handed them to the guy running the label machine.  He said thanks and put them next to the other rolls they were actively using on that production run, which could've then been loaded into the machine when the current roll ran out and causes a mislabeling issue.

Edited by jfrey123, 05 August 2026 - 03:33 PM.

[GMO]

 

GMO — the "checking inputs vs. outputs" distinction (allergens marked bold as a field vs. actually rendering bold) is exactly the kind of failure mode that's invisible until it isn't. And liberator's can story is a great, concrete version of the same thing: a robust vision system doing its job, a human overriding it because the artwork looked right, and the actual defect being orientation, not content. Systems catch what they're built to catch; people stop noticing what they check every day.

 

 

Not quite. Inputs can actually be right but outputs wrong or inputs are pointing to the wrong data (both in artwork approval or online checks), or as the output is the important thing, the printing quality simply being poor. For example, bold print can appear to not be bold if there is a printing issue. I'm always curious why manufacturers seemed to have settled on that as a norm. Capitals would seem much less prone to this risk and equally valid.

 

By the way, curious why you chose AI to write that response? There are several tells. The em dashes of course but far beyond that.

[Bo16]

Don't forget:  processing aids not required on the label, spices not specific, enzymes listed as enzyme (what enzyme), ancillary ingredients in both of those and other formulations, GMO, Organic, non-GMO, Halal, Kosher, Mandatory vs Voluntary labeling laws, added vs naturally occurring ingredients, proprietary information?

 

Confusing.  What does or doesn't your ingredient supplier have to tell you and then does that need to go on the label.  If you put everything that was used to make the ingredient, then all the ingredients on your label, there would almost never be what the world is now calling clean label. 

[harit]

Yes. Boring reason though. English isn't my first language, and I write too long and lose my own point halfway. So I take help from AI to squeeze my notes down into something shorter and clearer. The thinking is mine, it just helps me to  tidies up the words.

[GMO]

Yes. Boring reason though. English isn't my first language, and I write too long and lose my own point halfway. So I take help from AI to squeeze my notes down into something shorter and clearer. The thinking is mine, it just helps me to  tidies up the words.

 

Ah, fair enough. Trust yourself though. AI sometimes gets meaning slightly wrong in a way that people don't even if they're writing in a different language. I think you probably understood my point then AI brought in misunderstanding. 

By the way, the tells:

The em dash

The use of the word "quietly" and other adverbs which don't need to be there

"Surfaces as" - not a phrase used in British English at least

"invisible till it isn't" - sounds very fancy actually says very little

 

Overall as well AI "speech" is incredibly bland. It has no personality. I'm now starting to look for work which has mistakes in grammar as a mark of authenticity.

 

When AI first became popular, I used to use it all of the time and it has it's uses but what I looked at and thought was great copy is now incredibly irritating to me. It's the unflavoured cottage cheese of writing. So please, feel free to put down what you want to say, mistakes and all. Nobody will judge you for it.

 

And just for fun, this is the above put through an AI tool so you can "hear" just how awful it is:

 

Love this perspective.

One thing I've learned over the years is that when it comes to communication, we sometimes underestimate the value of trusting ourselves.

AI is an incredible tool.

But here's the thing.

Sometimes what looks like clarity actually surfaces as complexity.

Sometimes what appears to be enhancement quietly creates distance.

And sometimes the misunderstanding isn't in the original message.

It's in the interpretation of the interpretation.

Let that sink in.

The signals are often invisible until they aren't.

You'll start noticing:

• The em dash.
• The adverb that quietly appears where no adverb was needed.
• Phrases like "surfaces as" that somehow sound professional and unnatural at exactly the same time.
• Statements that feel profound, impactful and deeply resonant while communicating almost nothing of practical value.

 

The truth?

AI-generated writing is increasingly polished.

It's increasingly sophisticated.

It's increasingly accessible.

But it's also increasingly homogeneous.

In the pursuit of perfection, we sometimes lose the imperfections that make communication human.

A grammatical inconsistency.

A sentence that runs on a bit too long.

An awkward phrase.

A personality.

That's where the magic lives.

When generative AI first emerged, I was genuinely excited by its potential.

Like many people, I embraced it wholeheartedly.

And for a while, I thought the output represented the future of effective communication.

Today, I've come to a different conclusion.

Great writing isn't about optimisation.

It isn't about refinement.

It isn't about leveraging language to drive engagement and unlock authentic connection at scale.

Great writing is about humanity.

The best writing doesn't always sound perfect.

It sounds real.

That's the lesson.

So if you're worried about making mistakes in your writing, here's your reminder:

Write the thing.

Use the awkward phrase.

Leave the imperfect sentence.

Be unapologetically authentic.

The people who matter won't judge you.

And the people who judge you probably weren't your audience anyway.

Thoughts?

[BM5]

Having just undergone 2 retailer artwork redesigns at the same time for produce, there have been a lot of errors at design stage and at legal (SGS), differences in interpretation between colleagues at the same legal approval firm (SGS), design agency not clear on the brief and carrying on anyway, multiple designers with what would appear to be some sort of free reign over artisitic impression!?! working on the same range with differeng position of nutritional tables, flash marks etc., a lot can go wrong, particularly if you're input is ignored and you're told to stick to the technical details.

 

Artwork not fitting in to cutters so varnish is applied to the wrong place and print comes of the film reel for example. Product names changing meaning you have to update all your internal docs to match the new name and it's possible to miss the odd one. 

 

The biggest issue (touch wood we've not had to carry out a recall for this reason) is line staff not changing labels or selecting the wrong labels, seeing as they "all look the same" due to the uniformity of the retailers packaging.