Where Shrink Actually Happens: A Walk Through the Store

A retail store is a busy place. Shoppers are selecting products, cashiers serve queues, and self-checkout attendants help a few customers complete a transaction. At the entrance, customers come and go. If you were walking through such a chaotic scenario, how is it possible to identify whether any customer has left the store with unpaid merchandise?

And knowing where to look is just the beginning. The more useful question is where an item’s journey stops matching the payment record. To detect theft and non-payment, you need to follow the goods through key areas of the store such as aisles, assisted checkout, self-checkout, and security gates. 

Carrying out such a comprehensive walkthrough multiple times during a busy day cannot be done manually in a reliable fashion. This is precisely where Visual AI technology comes in handy. SAI One is a store intelligence platform built on SAI's patented Vision Language Model (VLM). SAI One turns the existing cameras in retail stores into continuous store intelligence and moves beyond alerting and points to the next best action that the store needs to initiate.

Aisle is where the customer journey begins

The first stop must be the aisle, as this is where merchandise starts moving from the shelf into a shopper’s hands. Aisle monitoring can flag suspected concealment and rapid product removal, while connected checkout and exit coverage can add context about whether selected goods were presented and paid for. A CCTV video clip from an aisle is just a starting point for investigating potential theft. Do not consider it as a complete account of loss.

• Concealment is an important event to examine. The shopper appears to move an item into a bag or clothing rather than the basket. The situation needs to be reviewed before deciding on the next course of action. The aim is to understand the evidence. SAI One detects and raises a concealment alert to notify store staff that they need to start monitoring the specific customer’s activity. 
• Shelf-Swipe involves rapid removal of multiple products from a shelf. SAI One detects this pattern and raises an alert along with a video clip. The store team can examine whether the associated video footage shows any unpaid product removal.
• Picked in aisle but not scanned: In this scenario, the attention shifts from a single gesture to the whole shopping journey. The store staff needs to know that all products selected from the aisles later appear in the checkout record. SAI One offers connected monitoring that automatically compares items picked up in the aisle with checkout activity to identify possible non-presentation or partial payment.  The store staff can review the associated SAI One footage to verify whether the shopper put the item back on the shelf or moved on to the checkout area.
• Consumption in store: Opening a drink or eating a food item inside the store is a scenario that can have physical evidence, but it is important to verify whether a corresponding payment follows at the checkout. Consumption of unpaid goods is a documented retail monitoring scenario, but it cannot be considered as confirmed theft unless the video footage clearly shows non-payment for the same. SAI One is integrated with store’s payment systems and automatically flags discrepancies that the store staff can verify before the shopper leaves the checkout. 

Assisted Checkout: A staffed lane can also be a weak link

At assisted checkout, the review shifts from shelf activity to the relationship between merchandise handling and the transaction. A staffed lane is still a place to examine missed scans and possible internal theft.

For instance, sweethearting involves a cashier who deliberately allows an acquaintance’s item to pass without scanning. Void is another scenario where the context needs to be analysed. 

SAI One’s transaction-linked monitoring flags patterns such as sweethearting or unauthorized voids and connects them with video footage. This enables the store staff to review the footage before deciding what response is appropriate.

Self-checkout: Scanning does not equate to paying

At self-checkout, several different routes can lead to unpaid goods. Non-scans, item stacking, payment skips, and walkouts that bypass checkout altogether are scenarios that need to be detected and addressed in realtime. 

SAI One detects all these behaviors and provides video clips that the staff can review. Here again, review is the operating word. The aim should be to establish what happened and allow possible mistakes to be corrected before classifying any incident as self-checkout theft.

• Items not scanned: A product remains in the trolley or moves through the checkout area without a matching entry in the transaction. The reviewer needs to review the evidence to confirm whether the mismatch is a human error or deliberate theft.
• Stacking: Here, the shopper presents multiple products together and fewer items get registered. For instance, in such a mis-scan scenario, two products can move together across the checkout while the transaction records only one. SAI One compares the merchandise quantity with the recorded quantity and provides reliable evidence to reviewers. 
• Part payment: While a receipt can be an important piece of evidence, it does not answer every question about the merchandise. Sometimes, the customer scans the items but leaves the store without completing payment. SAI One distinguishes payment skips from non-scans and surfaces mismatches between basket content and transaction data.
• Full trolley pushout: This scenario goes well beyond an individual scan discrepancy. The customer skips the checkout and leaves the store with a loaded trolley full of unpaid goods. SAI One connects self-checkout coverage with exit monitoring so that reviewers can quickly check where the trolley went, and the evidence of non-payment.

Across all four situations, build the response around verification. SAI One sends a nudge (a friendly prompt) to give a shopper an opportunity to correct a possible error.  If the situation remains unresolved, such events can be reviewed by trained staff. The staff needs to use SAI One’s real-time correction capabilities and evidence-backed notifications to ensure that a completed, accurate transaction takes place while avoiding unnecessary confrontation.

Security Gates: The final handoff

The security gate is the last stop in the current shopping journey as it ensures that customers scan their receipts (often using the QR code on their receipts) before exiting the store. If security gates remain open for too long, shoppers may leave without proper verification. SAI One monitors the status and operation of gates positioned at store exits, especially those outside self-checkout (SCO) counters and raises alerts when a gate remains open for prolonged duration and send alerts to trained staff. 

Watch out for repeat offenders

When a person linked to a previous theft incident arrives at the store, SAI One raises Revisit Alerts with historical evidence for the staff to review. While the historical data can provide context, it should not be considered as proof of wrongdoing during the new visit. The staff must review the current image against prior evidence before initiating further action. 

Store-Walk Checklist

For each of the following areas, identify competent reviewers, the evidence needed, and the recommended safe response.

• Aisle Monitoring: Review coverage for concealment, shelf swipe, picks not scanned, and consumption in store. 
• Assisted Checkout: Check how missed scans, suspicious voids, and possible internal theft are reviewed against transaction records and footage. 
• Self-checkout: Check for items not scanned, item stacking, part payment in transactions, and full trolley pushout instances. 
• Exits: Monitor the security gates and departure evidence against payment context.
• Revisits: Check that historical incidents and current matches are reviewed by staff before any action is initiated.

Frequently Asked Questions

Q. Where should a store-walk review begin?

Start in the aisles, then follow merchandise through assisted or self-checkout and the exit. The aim is to connect product handling with registration and completed payment, rather than relying on an isolated observation. SAI One’s connected aisle, checkout, and exit monitoring capabilities simplify these tasks. 

Q. Does an aisle alert prove that theft occurred?

No, it should be considered only as a prompt for review. For concealment, shelf swipe, picks not scanned, or consumption in store, examine what happened next (whether goods were returned or paid for) before reaching a conclusion. SAI One provides connected evidence comprising video clips and transaction data that the staff can use to verify these aspects.

Q. How should possible internal theft at assisted checkout be reviewed?

Match the relevant merchandise movement with the transaction record and video footage. SAI One provides transaction-linked monitoring and flags sweethearting and unauthorized voids. However, authorized trained staff must review and distinguish an exception from evidence of deliberate conduct. 

Q. What is the difference between non-scanning, stacking, part payment, and a full trolley pushout?

• Non-scanning leaves an item unregistered 
• Stacking involves multiple products handled together with fewer registered. 
• Part payment covers only some goods
• A full trolley pushout involves a loaded trolley leaving unpaid without passing through checkout. 

Q. What should exit monitoring cover?

SAI One provides departure evidence alongside information on the shopping and payment journey. Assign alert ownership, preserve relevant records, and prioritize safe responses while making interventions.

Q. How should staff use Revisit Alerts?

SAI One provides footage of verified historical incidents which should only be considered as context and not as proof of new wrongdoing. Review the current match before responding, and maintain clear access, retention, and correction procedures.