)
A New Product Line for an Old Retail Problem
Scandit, a company best known for barcode scanning and computer vision software embedded in retail apps and handheld devices, has entered the loss prevention market with a dedicated self-checkout product. The system uses real-time vision AI to detect both accidental and intentional loss patterns as items move through a self-checkout station, catching the mis-scans, unscanned items and bagging errors that account for a large share of self-checkout shrink. Scandit says the product can recover or deter more than 75% of self-checkout losses through real-time detection.
Self-checkout shrink has been a persistent, expensive problem for retailers since self-service lanes went mainstream, and most existing solutions fall into two camps: weight-based sensors that generate frustrating false positives, or camera systems that lean on facial recognition and biometric tracking to identify repeat offenders. Scandit's entry is notable for what it deliberately leaves out. There is no facial recognition and no biometric identification anywhere in the product, a design constraint the company built in rather than added after regulatory pressure.
Detection Without a Biometric Database
The technical approach centers on analyzing the checkout process itself, tracking whether scanned items match what actually goes into the bag, rather than tracking who is doing the scanning. Video processing happens locally on the checkout station instead of routing footage to the cloud, which both reduces latency for real-time alerts and keeps the company on solid ground with GDPR and similar data protection regimes that treat biometric identifiers and centrally stored video as higher-risk categories requiring stricter handling.
Scandit CTO and co-founder Christian Floerkemeier framed the trade-off directly: retailers should not have to choose between a fast checkout experience and strong loss prevention. That framing matters because the two most common prior approaches each sacrificed one for the other, weight sensors slowed down honest shoppers to catch dishonest ones, while biometric camera systems caught more theft but exposed retailers to privacy complaints and, in some jurisdictions, regulatory action over unconsented facial data collection.
Self-Correction Over Confrontation
A key operational detail is how the system resolves most flagged incidents. Rather than routing every discrepancy to a store associate for manual intervention, the product is designed so shoppers can self-correct in most cases once they are alerted to a mismatch. That matters for staffing models that have already stripped labor out of self-checkout zones, since a system that requires an associate for every flag simply recreates the labor cost self-checkout was built to remove.
It also matters for customer experience, an area where loss prevention technology has historically done real damage. Aggressive AI flagging that treats every shopper as a suspect erodes trust and slows down the exact self-service experience retailers invested in for speed. A self-correction-first design keeps intervention as an exception rather than the default response, which is a meaningfully different posture than the accusatory tone many earlier loss prevention systems adopted toward ordinary customers.
Privacy-by-Design as a Sales Argument
The decision to exclude biometrics is as much a commercial strategy as a technical one. Retailers have watched competitors and peers get dragged into privacy complaints, lawsuits and regulatory inquiries over facial recognition and video surveillance programs, and procurement teams are increasingly asking loss prevention vendors to prove their systems cannot be repurposed for biometric tracking later. Scandit is selling the absence of that capability as a feature, not a limitation, betting that risk-averse retail legal and privacy teams will value the smaller attack surface over marginal detection gains a biometric system might offer.
That bet looks reasonable given how loss prevention technology has become a governance flashpoint elsewhere in retail, from the electronic shelf label backlash over dynamic pricing to advocacy campaigns targeting AI vendors with surveillance associations. A vendor that can point to local processing, no biometric storage and GDPR alignment out of the box removes several items from the privacy review checklist before a retailer's legal team even opens the contract, shortening a procurement cycle that has grown longer as scrutiny of retail AI has intensified.
What CTOs Should Ask Before Buying
For retail technology leaders evaluating loss prevention AI ahead of the holiday season, the Scandit launch is a useful prompt to separate two questions that often get bundled together, how well does a system detect loss, and what data does it need to store to do it. Those questions have different risk profiles, and a vendor's answer to the second one increasingly determines how fast a deal clears legal and privacy review, independent of how good the underlying detection model actually is.
The 75% loss recovery figure deserves the same scrutiny any vendor-supplied performance number deserves: ask for the methodology, the store types tested, and what counts as a recovered or deterred loss in that calculation. But the more durable lesson from this launch is architectural. Building loss prevention on process analysis rather than identity tracking is a design pattern other vendors will likely copy as retailers signal, through procurement decisions like this one, that biometric-free detection is worth paying a premium for.
Timing Ahead of the Holiday Shrink Season
The launch lands roughly two months before the peak shopping season, when self-checkout volume climbs sharply and loss prevention teams are already stretched thin covering extended hours and higher transaction counts. Retailers that want a new loss prevention layer running before Black Friday have a narrow window to evaluate, pilot and deploy a system, which favors a product built to run on existing camera infrastructure over one that requires new hardware, cabling or point-of-sale integration work across hundreds of stores.
That deployment speed is itself a competitive argument Scandit is making against slower, hardware-heavy loss prevention rollouts. A retailer weighing this purchase now is comparing both detection accuracy and deployment speed between vendors, weighing which one can realistically be live in stores before the volume surge hits. Loss prevention leaders should treat that timeline constraint as seriously as the underlying technology when scoping a holiday-season pilot, since a superior system that cannot be deployed in time delivers no protection at all during the highest-shrink weeks of the year.
