How AI detects and resolves self-checkout missed scans

29^th April 2026

Let’s take a look at the most common type of shrink event that we see on the self-checkout in retailers all around the world. We call this miss scanning or skip scan. This is where any customer takes an item, and they place it in the bagging area of the checkout without scanning it. Now, this could be deliberate. This could be someone trying to steal from the store. Can also very frequently just be a customer making a genuine mistake. The good news is the system can handle it either way. And we intervene in a smooth, simple, low friction way for customers and for staff.

So here I am. I’m going to start transacting at the checkout, scan a couple of things through. And let’s say as I’m doing that, I fail to scan an item as I’m bagging it. Now looks to us like we try to scan that item. We didn’t scan it. We’re able to nudge the customer directly and present them with a video annotated to show with the red box the item that’s caused the problem, and a message saying was the last item scanned. Did you scan the last item? This allows the customer to self-correct. It’s a nudge. They can say Yes or No, and the retailer can decide what happens with each of those outcomes. I say, yes, I did scan it even though we know I didn’t. The retailer can then trigger a staff intervention if that’s what they decide to do. Equally, you could give the customer the benefit of the doubt. In this case, I’m going to say I’m an honest customer, and I made an honest mistake. Keep going.

But let’s do the same thing again within the same transaction. That’s my second missed scan within this transaction. And this time we say we’re going to block the checkout and trigger a staff intervention. You can see here the customer’s locked out of the SCO. A member of staff is going to come and assist them. Now critically the staff member can see the same evidence video on the checkout. And they can also see this on a dashboard or on a palm device wherever you need them to see it, to make sure when they arrive to help your customer, they’re able to deal with the situation as quickly and smoothly as possible.

So that’s mis scans, and that’s our rules based system for deciding how you act on them. That rules engine is totally flexible to you, the retailer. You can configure it however you like, for nudges for blocks and for other types of intervention.

One more thing I’ll just show you quickly is that the system is capable of handling all sorts of complicated customer behavior. For instance, if I get my phone out at the checkout and I put it there and accidentally drop it into the till, we’re not going to receive a nudge or a block for that, because the system knows my phone is not a product. It’s not worried about that.

Similarly, if I take two identical products and scan one of them twice, pop them into the bagging area. We’re not going to receive a nudge, because the system knows that those two products are similar, and we can let that go.

I’ll scan my last item. I have an empty basket here. When I hit Finish and Pay, we’re going to let the customer checkout. We’ve stopped 2 or 3 incidents of theft. We’ve done it in the most efficient, effective way possible. That is the most common, most typical example of theft that we see at self-checkout, but it’s by no means the whole story. The system can cope with all sorts of more complicated scenarios as well.

The single most common shrink event at self-checkout is an item placed in the bagging area without being scanned. It’s a missed scan, skip scan, or it can be deliberate. The advantage of detecting missed scans visually is that the Vision AI responds to what happened at the checkout, then lets the retailer decide how firmly to act.

A nudge that lets honest shoppers put it right

Vision AI sees an item bagged without a scan, it doesn’t lock down the checkout immediately. It nudges the shopper directly, showing them an annotated clip of the moment – the item in question marked with a red box and a simple prompt: did you scan the last item? The shopper answers yes or no, and the retailer sets what each answer triggers. This gives an honest customer the benefit of the doubt and the transaction carries on. The point is to let people self-correct before anyone is pulled into it.

Escalation when the pattern repeats

The response changes with the behavior. A second missed scan in the same transaction can move from a nudge to a block: the checkout locks and a colleague is called over. Crucially, that colleague arrives already informed – the same evidence clip is available to them on the checkout, on a dashboard, or on a palm device, wherever they need it. They can see exactly what happened before they reach the lane, so the intervention is quick, calm, and resolved without a guessing game in front of the customer.

A rules engine the retailer controls

SeeChange provides the rules engine. The retailer decides how it behaves. You set the thresholds, choose which events nudge, which block, and which trigger another response, then adjust as you learn what works across your stores. The result is a nudge-or-block decision that reflects your risk appetite, not a fixed setting imposed on you. Our AI rules engine covers how that logic is built and tuned.

Knowing what to ignore

A self-checkout that flags everything creates as much friction as the shrink it prevents so the harder problem is precision. Because SeeChange detects what’s happening visually rather than SKU-matching or weighing the basket, it can tell ordinary behavior apart from a real event. Drop a phone into the till by accident and nothing happens – the checkout knows a phone isn’t a product. Take two identical items and scan one of them twice, then bag both, and there’s no nudge: it recognizes the products as the same and lets it go. That’s the difference between catching theft and catching honest customers in a net of false alarms.

Why detection sees what the scanner can’t

The scanner records one thing: whether an item was scanned. It captures nothing about what actually happened at the lane. So a self-checkout built around a scanner alone cannot tell a deliberate skip from a genuine mistake, a slip of the hand, or a moment of distraction. It can only flag the absence and escalate, which is why first-generation self-checkout security tends to choose between letting events pass and frustrating honest shoppers.

By looking at what actually happened at the checkout, Vision AI can recognize a missed scan, judge whether it’s a one-off or a pattern, and respond in proportion without leaning on weight or SKU signals. How that works is explained in the computer vision self-checkout section of our self-checkout security guide, covering what vision AI can see that traditional systems can’t, and how it holds up across different checkout configurations.

For the complete picture of how SeeChange protects the self-checkout lane find out more on the AI Self-Checkout solution or speak to the team.