The future of linking retail criminal activity

As an industry, efforts to connect criminal activity remain largely manual, time-consuming, reactive, and ineffective. As one ORC investigator called it “it’s based on individual investigators to link, it’s not systemised.”

We recently hosted an invite-only webinar with industry leaders and experts on “Suggested Merges: the future of linking retail criminal activity.” In this article we’ll share some of the key concepts discussed.

There are four building blocks to the industry effectively linking retail criminal activity:

  1. Make it easy to capture quality data.
  2. Structure the data.
  3. Build smart, simple workflows and controls.
  4. Supercharge with Artificial Intelligence (AI).

Before we dive into each of these building blocks, it’s important to acknowledge why connecting retail criminal activity is more important than ever. The risk/reward tradeoff of retail crime is attractive to retail criminals. This is because offenders enjoy anonymity in a complex environment while the risk of getting caught and prosecuted continues to decrease.

Why offenders enjoy anonymity:

  1. Choice of victims - nationwide retailers have hundreds to thousands of stores to target, and when bad actors look at verticals and the industry as a whole there are tens of thousands.
  1. Inadequate systems - retailers and law enforcement partners all have disparate systems not designed to aggregate criminal activity or share data in a safe and secure way.
  1. Frontline staff not empowered - systems tend to be one way information flow from stores to corporate which misses the network effect of your stores working together to raise awareness of bad actors in the area. This allows bad actors to go store to store often undetected.
  1. Under-reporting - on top of all of this, store teams are busy, pulled in different directions, reporting incidents is time-consuming. You can’t link what you don’t know!

Our estimate is that only 5% - 10% of external theft is actually captured/reported by retailers. If you want to see where your organization sits, look at the delta between the total value of the reported incidents of your organisation last year and the size of your unknown loss you believe is external. Then divide the delta by the average value of your external incidents. This number is an estimate of the number of external incidents your organization is not seeing or reporting each year.

For example:

Without even getting into the challenges the industry face linking this criminal activity, this equates to a lot of anonymity.

Challenges around jurisdictional boundaries, increasing decriminalization of shoplifting, and increasingly resource-constrained law enforcement act as a force multiplier on top of this anonymity bad actors enjoy further decreasing the perceived risk of the criminal activity. 

Retailers that are serious about reducing crime and associated losses in their stores must address offender anonymity and overcome the current challenges in our criminal justice system. 

In a complex environment with so many interdependencies, clearly there is no silver bullet to achieve this. However, we believe it is imperative that the retail industry get better at linking criminal activity and raise the perceived risk of retail crime.

The building blocks to linking retail criminal activity:

  1. Make it easy to capture quality data - solve for under-reporting. You can’t link what you don’t know. But be careful, easy does not just mean quick. You also have to make the experience intuitive and accessible on any device to ensure a high quality of data is captured that can actually help drive outcomes. This includes capturing images and video. User experience matters. 

  1. Structure the data. This enables aggregation and for connections to be made, fast. Open text fields often lead to poor accuracy and compliance. Structured data requires less computing power to find possible connections and it leads to more reliable search capability for your team.

  1. Build smart, simple workflows and controls. Take advantage of the power of a network effect and empower everyone to take advantage of structured data. A key concept here is “suggested merges”. Enabling your frontline team to suggest possible links both at the time of reporting the incident and after the fact. Think about this as crowd-sourcing intel in a systemised way rather than “do you know this person” emails. Then build an appropriate approval process - but remember to not make this cumbersome or it won’t work effectively. 

  1. Supercharge with Artificial Intelligence (AI) - once you maximize the capture of quality data, in a structured way, put the right workflows in place, you have set the foundation to supercharge with AI. Our view is that it is critical to have human oversight over linking criminal activity. We can see a pathway to leveraging computers to do the grunt work (some examples below), constantly analyzing the data for connections, then making suggestions to investigators and/or field leaders to approve. This allows your people to stop doing manual, labour intensive work, that is well suited to computers, and instead focus on high-value work like driving crime reduction outcomes with law enforcement.
Natural language processing to identify connections from text
Computer vision to leverage images (and video)

The four building blocks outlined here will help retailers that are serious about reducing crime in their stores better connect the dots and focus on the 10% of bad actors driving 50% of the external loss. 

For an industry approach, there is a fifth building block. Which is creating an effective way to data share with other retailers and law enforcement partners in a safe and secure way, where each retailer retains data ownership (and control) and does not breach privacy law. That means you go from connecting criminal activity across your organisation to connecting it across the industry. We refer to this as a connected community.

If you’re interested in seeing how these progress, get an invite to our upcoming webinars below