Item not on file - Why it worries retailers more than it should

Item not on file: why it worries retailers more than it should

The item not on file loss risk is smaller than the industry treats it as, and that is the most useful finding in this episode. On ECR’s own numbers, discussed in the Self-Checkout Loss Report 2026, the item that will not scan works out at roughly 34 cents per thousand transactions. It sits near the bottom of the report’s list of loss generators, well below missed scans and walkaways. Professor Matt Hopkins is blunt about the implication: as a retailer you go where your risk is greatest, and on this measure item not on file is not it.

What makes it worth attention is not the direct loss. It is the frustration. A shopper who cannot get an item to scan is a shopper who may decide they are entitled to bag it anyway, and that is a different problem with a different price on it. So the case for fixing it is a customer experience case that happens to reduce loss, rather than a loss case in its own right.

The cause is almost always upstream. Locally sourced produce is packaged differently from store to store, and packaging changes over time, so the master data file is chasing a moving target across thousands of products and hundreds of sites. The direction retailers are converging on is bottom up rather than top down: systems that recognise an unfamiliar item locally and push it to the master file quickly, with AI and computer vision doing the detection instead of a busy self-checkout host filling in a form. One retailer described a threshold of three scans in a single store before an item goes on file across the whole estate.

More on the wider picture at the self-checkout hub.

The full conversation follows.

Item not on file at self-checkout: Colin Peacock and Matt Hopkins in conversation

COLIN:

Welcome to the podcast, Sell more, lose less. And this week I’m joined by Professor Matt Hopkins. And we’re looking at self-checkouts, something that Matt’s really been investigating with us and writing up research reports on, one just launched, for the last two to three years.

One particular aspect of this whole problem is when you get to the checkout and your item doesn’t scan, and it says you can’t sell it. Cadbury Creme Eggs, with their barcodes, which are very untidy and get crinkled, are one of the sort of classic products that have this problem. That’s more linked to their packaging. But we’ve had a chance to do the research and learn a bit more about it.

So I was going to ask Professor Matt just what it is about this, how big this problem is. What are some of the drivers of this problem? Where might some of the interventions be going forward? What evidence do we have that those interventions are then making a difference? Because ultimately, I think this is a frustration that can be and could lead to loss.

And we already know from Professor Matt’s report that up to 25% of a store’s unknown loss can be attributable to the losses at the self-checkouts. So with that Matt, what do we know about items not scanning or not beeping at the self checkouts?

How big is the item not on file problem, really?

MATT:

Well, we know this is an issue. And we have some data on this in terms of the prevalence of this as an issue. We don’t think that this is, when you compare it to walkaways or missed scans, the prevalence of the issue is not as great, but still it is an issue for retailers.

And I think what we do know is that this can be incredibly frustrating to customers. So if you think of a customer going through their customer journey, and obviously we want those customer journeys to be frictionless and to be as straightforward as possible. When they get to SCO and they’re trying to scan items and the item is not on file, that will lead to frustration in the customer, and they’ll either need a quick intervention from a SCO host here, or they’re going to be in a situation where those frustrations might drive loss, in that they become irritated and they get that sense maybe that they’re entitled just to bag their item without scanning the item.

So we know this is an issue. We know it generates loss. And we think we have some good ideas around why this is being generated and possibly what we can do about this as well.

COLIN:

Yep. As I say, we heard from some retailers that it can be down to the master data file. One retailer was generous enough to show us. They developed a system so that all the master file was accurate. So what were some of the upstream interventions we heard about, beyond just this sort of master data file accuracy cleanup?

Why locally sourced produce breaks the master data file

MATT:

Yes, I mean the master data file is one of the things where there’s been a lot of discussion about this. What retailers want to do is obviously make sure that their master data, in many ways their big data about the products they’re selling across the company, is as accurate as possible. The challenge you’ve got, as retailers will know, is that you’re operating across multiple sites. You’ve got thousands and thousands of products, which should theoretically be on the master data.

But what you want to be able to do as a retailer is also to be able to offer a variety of products in different locations. So what you’re going to have is lots of locally sourced products as well, for example. And it’s really produce here which is the issue, and it’s that produce which is locally sourced, might be packaged slightly differently from location to location, and also those products where the packaging can change over time as well. It becomes an issue.

So that’s where you tend to get those issues here. And one of the ways of obviously dealing with this is to be able to deal with the problem locally. So where you do get item not on file events at the local level, there’s some way in which updated very quickly and to deal with that.

At the moment, what a lot of retailers are doing, they’re waiting for stores to report this. They’re waiting for the people who are doing the interventions to report this. And that can become problematic.

And really, the long term solution to this is to get to a situation where the systems learn very quickly. They recognise those products on file. The systems learn. So there’s a machine learning process here, which can then from a local level update the master data file very quickly. So your learning here comes from the bottom up rather than the top down.

COLIN:

Yes, we heard about retailers who are scanning the barcodes as they arrived at the distribution centre. They were using other feedback mechanisms such as the sort of electronic shelf labels to check that all the items were on file. So you’re right about that feedback mechanism. At the moment, what did we learn about how retailers were actually handling the risk, if you like, of the frustrations boiling over into a loss, or just simply managing those interventions where there is an item that does not scan.

MATT:

Well, at the local level, unfortunately, a lot of retailers don’t seem to be dealing with this particularly well. They don’t seem to be updating those lists particularly quickly, particularly accurately. A lot of the SCO hosts as well, we know that a lot of those items are actually just getting through without being scanned properly. There tends to be scanning of the wrong products in these situations, which of course has a massive impact then on inventory. So particularly with fresh produce, what you’re seeing is the wrong types of apples, grapes going through, and so on, which any retailer will know that creates a huge problem in terms of items on file and scan accuracy.

COLIN:

Yeah, I can see that. And I could also see how you don’t want to stop every transaction if you’ve got a number of different interventions to make. So you might wait to the end for an age verification as well as a sort of item check, and often default to using the sort of dump code, the generic code, if you can’t see it through.

So did you come out of the discussions with the retailers feeling like they’ve got a grip on this? Do you feel that there’s a clear direction and sort of best practice emerging that retailers can sort of look to, in terms of benchmarking their own performance? Were there some sort of golden rules, if you like, to sort of follow?

Hard stop or delayed review: two controls, neither settled

MATT:

Well, I think there are a number of things which are coming out of this. And I think one of the things we’re starting to see is that in terms of interventions, some retailers were talking about using hard stops as a sort of preferred method of control here. The concern with this is obviously again that kind of creates more customer frustrations here.

And retailers, they tend to talk about two types of control. As we mentioned earlier, we’ve got that delayed review sort of control. You allow the transaction to continue and you resolve later, which obviously can be quite a risky strategy. Or we go for the immediate hard stop. The attendant will come immediately and try and work out what’s going on here.

I think one of the problems with either one of those is you’re getting those customer frustrations, and you’re getting that intent from customers again to sort of compensate themselves in certain ways. I think one of the things we need to be very clear on going ahead is probably to develop some rules around those sort of stops, those sort of interventions that are required here.

So I think at the moment, retailers are very much feeling their way through this and there’s not a great deal of clarity on what should be done around this in relation to stops or in relation to updating their master file data.

COLIN:

Yes. Well, I think to your point, probably this sense of urgency for many retailers might not be there relative to some of the other risks that they’ve got. You pointed out earlier on, the numbers we printed in the report was that for every thousand transactions, the risk of loss here might be about 34 cents, which is based off a very, very low probability… low number of transactions. Whereas the missed scans you talked about, the risk per thousand transactions is around eighty eight euros. So I think that brings some sort of scale and dimension to context to what we’re dealing with here. Yet it was very well attended, both sessions that we’ve run on this.

Putting the loss in proportion

MATT:

Well, it stands to reason, doesn’t it? As a retailer, you’re going to go to where you think your level of risk is greatest. And if you do look at those sort of prevalence rates per one hundred, per thousand transactions, the items not on file comes fairly low on the list.

The missed scans comes fairly high. So as a business, where you’re going to put your resource. Walkaways we know is the highest individual loss generator here, about eighty eight euros per event.

So again when you compare it to item not on file, the scale of the problem doesn’t appear to be that great. But it still is a challenge. It still is an issue here. It still is an issue for any business not to have items on file where customers are trying to process that data.

COLIN:

Yes. And I think as you say, it could lead into temptation into the wrong behaviours, especially if you don’t have other controls like weight control at your self-checkouts. So yeah, I think this will be an ongoing issue. Well, thank you very much for that. Oh, sorry, do you want to say something else?

The fix is bottom up, not top down

MATT:

Well, I’m just going to say, and I think ultimately the way to deal with this just has to be that learning from the bottom up. It has to be that AI learning and computer vision for this. They’re going to be the detection tools. They’re going to be the way that businesses can really streamline their operations on this and to make sure they can still source those products, maybe local products, and make sure that master data is absolutely bang up to date.

COLIN:

Yes, I think, to your point, one of the retailers on the call said that after three scans of a particular item in any one store, that item will be on file for all stores if they wanted to use it. So they very quickly bottom up were able to build up a very accurate master database. Not only just of the sort of the product description, but in future perhaps even what the product looks like.

MATT:

Yeah, I mean, absolutely. And I think one of the problems with relying on human beings to do this, particularly with SCO hosts, is that they’re busy. They’re busy dealing with customers constantly. And I think there becomes an issue here with products which are not on file in terms of relying on that human type of intervention to update those products.

COLIN:

Yes, it’s probably my bias as well. But I sense that being the sort of leader of the master data accuracy team is possibly not the most exciting job in retail and takes a bit to

MATT:

But there are probably more interesting things to be doing. Yeah.

COLIN:

Yeah, so it’s one for the sort of specialists, but actually really critical. But perhaps it’s in the hands of other people who do not prioritise it enough, the commercial teams and other people in the whole ecosystem, which is where we are in terms of accuracy. But there we go. Okay, very good.

Thank you very much, Matt, for your time and your expertise as ever, and we’ll be in touch soon. Thank you very much.