Authors:
James Hawk, Susan Brosnan
Abstract:
This solution proposes using a network of in-store shoppers to crowd source POS security. Specifically, create an infrastructure to report, using the shopper’s smartphone, suspected theft that does not interact with the alleged perpetrator. Machine Learning algorithms analyze the reports to calculate the reliability or credibility of each report. Based on this credibility score, store personnel could be alerted for appropriate action.
Background:
In self-service (self-checkout), shrink/theft is a major problem and an impediment for adoption. Recently, while in the queue for self-checkout the author saw a bag of candy being passed around the SCS system and put right into the shopping basket. The shopper assistant was alerted to this, but they were overwhelmed with other duties and merely shrugged.
If there was a way to augment the standard security to alert the central store security, this would reduce shrink. Further, this would reduce the risk of 'sweet hearting', which was another possibility for the shrug from the shopper assistant.
This method could also be used to validate the accuracy of existing technologies that detect retail theft, e.g., existing AI/CV theft detection solutions. This method could provide an autonomous method for validation.
Description:
Using an integrated system of smartphone apps, utilize the shoppers ("observer") in the self-service line (queue) to provide a method of security of the POS transactions. This could be in the form of a text-based report or a video recording. Further, proprietary algorithms utilizing ML would be used to score the "observer" for reliability and reduce the opportunity for abuse ("SWAT-ing"). This scoring would affect credibility of future reports from the same observer.
Incentive programs could be developed, or existing loyalty programs could be enhanced to incentivize observers.
Also, signage within the store, advertising this new feature, would itself be a deterrent.
TGCS Reference 3807
