-- Loss prevention has historically defined its professional identity around catching theft. According to Mitchell Goldman, Chief Operating Officer at Edgify, the math underlying that identity has quietly shifted, and the LP function that recognises this first will be positioned for a significant change in how it operates and where it sits within the retail organisation.
Goldman points to the NRF's 2024 National Retail Security Survey, which put US retail shrink at $112 billion. The composition breakdown of that figure, which Goldman says is rarely quoted in full, attributes approximately 30 percent to external theft, 30 percent to internal theft, 25 percent to operator error and procedural failures, and 15 percent to administrative loss. Goldman says the operator error and procedural buckets combined represent more than $40 billion in shrink that traditional LP infrastructure is not built to address.
Why the Historical Stack Does Not Reach Operator Error
Goldman says the traditional LP technology stack was designed for the theft buckets. POS exception reporting after the fact, ceiling cameras streaming to remote monitoring desks, and computer vision alerting on gestures and behaviours can identify some theft. According to Goldman, operator error does not look like theft. A cashier who scans a six-pack as a single can, a self-checkout shopper who keys an heirloom tomato as a regular tomato, or a bagger who misses a wine bottle at the bottom of the cart does not trigger a gesture-recognition model but does appear on the shrink line.
The In-Lane Model and Its Results
Edgify's in-lane AI model sits on the POS terminal or barcode scanner, scores the basket as it builds, and prompts a quiet intervention to the cashier, self-checkout shopper, or floor supervisor before the transaction closes. Goldman says the intervention is the product, in contrast to the detection alert that characterised the previous generation of LP technology.
At a US natural-grocery banner, Goldman says LP incidents per week dropped from 60 to two after the in-lane intervention stack went live. He describes these as avoided-incident figures rather than detection figures. At a regional US grocer, the federated platform delivered an 8.9 times return on investment against the incumbent computer-vision LP system, according to Edgify.
A Shift in KPIs and Organisational Reporting
Mitchell says the KPI shift that follows from in-lane AI is one the LP profession needs to internalise. He describes the historical KPI as incidents detected, and the emerging KPI as interventions per million transactions and the ratio of interventions to confirmed incidents. In P&L terms this moves LP from a cost-centre activity to a margin-protection programme with a measurable yield.
In Goldman's assessment, this changes where LP reports within the retail organisation. Rather than reporting through corporate security, he says the function begins to report through store operations or directly to the CFO, with the coaching loop on intervention data flowing to store managers rather than a fraud investigations desk.
Mitchell also notes a workforce dimension. Real-time intervention turns LP from a punitive function into a coaching function, where a cashier who scans incorrectly receives a quiet prompt at the lane rather than a written report two weeks later. Goldman says Edgify's deployment partners have reported higher cashier retention alongside shrink reduction where this model is in place.
Edgify's federated edge AI platform powers self-checkout, staffed lanes, and self-service scales, across loss prevention, and item recognition across top-tier global retailers. For more information visit https://www.edgify.ai/.
Contact Info:
Name: Gianmarco
Email: Send Email
Organization: Xraised
Website: http://xraised.com
Release ID: 89197652
Google
RSS