Skip to main content

Bringing Organized Retail Crime Into Focus With Computer Vision AI

Organized Retail Crime (ORC) incidents are on the rise and crime groups are becoming more violent, making tackling ORC a priority for the retail industry and the communities they serve. While recent research offers a bird’s eye view of incidents reported to law enforcement, researchers agree more information is needed to better understand how and where these incidents are happening.

As an industry, we need to re-evaluate the way we collect and share data across stores, regions, and industries so that we can learn and benefit from the insights, and put preventative measures in place. We need to measure better, faster and more affordably than ever before. Only then will we start to see real change.

We have a unique perspective on fraud and malicious activity at Everseen because we are able to analyze millions of retail transactions every day and spot potentially malicious behaviors by what we call “red actors” — those who have criminal intent.

We look at loss more thoroughly by putting “eyes on the scene” – computer intelligence can see everything happening in every part of the store at all times – so it can respond in a way that humans cannot. While it’s still difficult to distinguish ORC loss from other types of shrink (which includes waste, errors, and accidents as well as theft), visual data allows us to see trends that indicate how and where the crimes are happening and what we need to be looking for.

The beauty of AI is that it learns how to improve with every transaction – and we literally have hundreds of years of video from retail stores and checkouts. Seeing trends before and during a crime or loss taking place enables us to suggest corrective action to enable protection, accordingly. The more data collected, the more accurate the trends and recommendations are.

I attended a loss prevention seminar with ORC criminal law enforcement professionals earlier this year and they described the moment a crime is happening as a “bang” and ORC (by the nature of it being ‘organized’) is happening pre-bang.  Computer vision AI can see what is happening during pre-bang, bang and post bang; and by doing this, the next best preventative measures can be put in place.

How ORC is Happening
While adding convenience for regular shoppers, self-checkouts have created a pathway for ORC and this is where we see a healthy chunk of loss happening. In fact, $23 billion of the annual $100 billion in retail shrink is unaccounted losses at checkouts and 48% of malicious shrink happens at self-checkouts.

Examples of ORC include:

  • Deliberate non-scans bypassing the scanne;
  • Switching a cheaper item for a more expensive one;
  • Leaving items in the bottom of shopping carts or in baskets;
  • Walk-outs without paying; and
  • Sweeping products from shelves

Just solving the problem at the self-checkout isn’t enough because it will only shift theft to other parts of the store. This is why a holistic solution that can see what is happening at the shelves, in the aisles, at the backdoor and elsewhere, is needed to see everywhere theft could be happening and furthermore, contextualize how it is being done. This is another advantage of artificial intelligence that can learn over time and be adapted to other parts of the store and beyond.

Who is being targeted?
Large national chain stores are key targets for crime groups, who rely on advanced planning to study store layouts, camera and exit locations, types of anti-theft precautions, and policies for stopping and reporting theft.

If we stop them pre-bang by deterring a theft attempt at any store with computer vision AI, we have accomplished a key goal. This is where collaboration among retailers in the community is paramount and something we are working on as part of the broader asset protection and loss prevention ecosystem.

Drug stores are also a popular target for ORC groups as they sell high-ticket, often smaller packaged items like makeup, razors and medicine that are easy to hide or sweep into a personal bag. Being able to identify this kind of behavior is too hard for store staff but it is exactly this kind of task for AI.

Insights from Artificial Intelligence
We can identify suspicious activity as it is happening by configuring the technology to respond. For example, 10 bottles of Tide are removed from the shelf by the same customer… that’s suspicious. When that threshold is crossed, an alert is sent to store associates, security guards or integrated loss prevention and security solutions.

For many stores, the current approach is to put high-ticket items behind locked doors, which adds more friction to the shopping experience for well-intended customers and is driving people away from stores. A better solution is to identify malicious behavior and drive criminals away but not the honest shoppers.

Current loss prevention solutions are narrow in scope and ineffective in stopping organized crime. Weigh scales at self-checkouts, for example, are programmed to correlate what has been scanned with what is on the scales – but that is all they do and they are notoriously inaccurate – needing staff intervention 20% of the time; and frustrating shoppers with “unexpected item in the bagging area”... now even a meme!

The automation capabilities of AI combined with the huge amounts of data available means that retailers can put in place protective measures to stay ahead of organized crime… pre-bang. ORC is one area where retailers can and do work together, and having visibility into the different types of malicious behaviors is key to being able to address the problem.

While emerging technology is being used to benefit ORC operations, technology like computer vision AI needs to be used by retailers to hinder ORC operations. The potential for taking action on the insights derived from AI is huge and every forward-thinking retailer should be looking at computer vision AI as a strategic investment in their future to stop and recover loss.

More Blog Posts in This Series

X
This ad will auto-close in 10 seconds