Analytics at the till: A loss-prevention use case

By Shawn Smith, Sr. Solution Specialist, SAS Canada Retail Practice

Big Data analytics and retail are a natural fit. With huge volumes of transactions and relationships between them and the client, it’s a marketers dream—coming closer to delivering the right offer to the right customer at the right time. While the value proposition for customer-facing, front-of-house applications is clear. What about back-of-house applications? Analytics can play a vital role in optimizing supply chains, reconciling inventory, and loss prevention.

Shrinkage is the dirty little secret that retailers don’t want to talk about because the discussion inevitably ends up pointing fingers at staff. Thus, loss-prevention as a discipline tends to be understaffed. One national retailer employed four loss-prevention analysts across the country—one in the west, one in Ontario, one in Quebec, one for the Maritimes—until it cut back its LP efforts and reduced its analysis staff to one person for the entire country. Considering that, if this were a $20-billion a year retailer that grapples with the average shrinkage rate of 6-7%--that’s about $140 million—a solution to the shrinkage problem could literally pay huge dividends.


There are numerous sources of shrinkage.

  • Paper shrinkage: Supply chain processes aren’t followed properly, bills of lading don’t match the inventory received, or receiving staff divert incoming inventory to be stolen later.
  • External shrinkage: Retailers’ eternal bane, shoplifting.
  • Spoilage: Perishable merchandise that isn’t sold by its best before date.

While analytics can play a loss-prevention role in any of these areas, let’s examine a very specific use-case—loss prevention at the checkout. Cashier fraud and “sweethearting”—collusion between cashier and customer—is responsible for significant shrinkage in retail. Identifying at-risk staff before it becomes a problem—for the retailer and the staffer—is a valuable proposition. One large international retailer’s analytics regimen can identify a high-risk cashier within two hours of his or her start time at the desk.


This type of prescriptive analytics draws on huge volumes of data to detect patterns that can predict at-risk employees. That data includes:

  • Employment history. How long has the staffer worked at the store? Is there a disciplinary history with the human resources department? Is the employee satisfied with his or her job?
  • Transaction types. Frequent voids and returns, especially in a regular pattern—for example, every fourth or fifth transaction—can be an indicator of a high-risk employee.
  • Social network analysis. To whom is the cashier linked? Do certain customers exclusively transact with one cashier? When pairs or groups of staff work together, is there a change in the transaction patterns?

Not long ago, conducting these kinds of analyses on a huge body of structured—as in, transactional—and unstructured data (which doesn’t fit into the defined fields of a database) would have been nearly impossible. At best, it would take days of IT department time writing structured queries based on the business goals of the LP department. Data that wasn’t transactional couldn’t be included. Line-of-business users were dependent on technicians to get the job done.

But a new generation of tools has opened up the magic of analytics to the line-of-business user. Visual analytics tools don’t demand highly technical query strings; with some training, line-of-business users can isolate patterns and determine relationships among huge volumes of data in a point-and-click environment. They can also bring unstructured data into the equation. The LP analysts’ expertise doesn’t need translation through an IT professional to wring relevant information out of the database.

This frees up time—an invaluable resource—for both the DL professional and the IT professional, allowing them to pursue more strategic concerns.


Analytics can red-flag particular transactions and cashiers as high-risk, but that’s all analytics is—a red flag. It still comes back to the DL professional to understand the patterns. A wrongful dismissal suit over a firing based entirely on analytics is just as damaging, if not more so, than the problem of shrinkage itself. There’s more to the process. That’s where the general, rather than technical, expertise of the DL professional comes into play.

Analytics is merely the beginning of the investigative process. After identifying high-risk employees, a retailer must perform due diligence before taking disciplinary action.

  • Identify suspect transactions and link them to specific staff. This is where analytics can lighten the workload by discerning patterns and red-flagging activities.
  • Follow up with physical security measures to deepen the investigation, such as observation by physical security staff or using “secret shoppers” to confirm suspicious behavior.
  • Analyze social links to external and internal parties to determine whether collusion is a factor.


Sometimes, what analytics detects isn’t malfeasance on the part of staff. Sometimes, it can contribute to a more strategic function.

Is a cashier high-risk, or is he or she simply undertrained? Some of the same cues that suggest abuse—lots of voids and returns in transactions, for example—may be a result of incomplete training. An LP professional must be able to discern whether to refer an issue to physical security for investigation, or to human resources to improve training. In fact, an LP professional is in a unique position to advise HR on gaps in training—if indicators come up frequently, and among several staff members, perhaps it’s time to revisit the training regimen.

Analytics can also contribute to an understanding of how to staff particular shifts. If there is an abnormal number of unusual transactions during a particular shift, perhaps more supervisory, or even simply more experienced, staff should be assigned. How is the performance of the operation when particular staff members are paired or work together in a group? Analytics can help optimize staffing as well as performing a loss-prevention function.


Applying analytics to cashier transactions isn’t just about catching the bad guys. By detecting high-risk employees early, there’s an opportunity to intervene before they become a problem, for the store and for themselves. Analytics can find training opportunities that have been overlooked. Analytics can refine and optimize staffing processes.

There’s a lot more to analytics in retail than marketing.

About the author:

Shawn Smith is a Sr. Solutions Specialist currently employed in SAS Canada’s Retail Practice. In his current role, he is responsible for helping clients gain value from current and future investments in analytic retail technologies. Shawn has extensive knowledge and deep domain expertise in retail analytics and in devising and executing strategies to better understand shopper behavior, loss prevention and Omni-channel evolution. 

A frequent presenter, Shawn has participated in numerous panels and webinar series involving discussions on various topics that are relevant to retailers such as Omni-channel Strategy, Shopper Marketing, Loss Prevention and Information Management. Chief to all of these sessions, was emphasizing the importance of “listening to the data” and leveraging analytics to solve and identify business issues. Through these channels, Shawn has been able to share both strategic and tactical insight into operationalizing enterprise wide analytics.

In his earlier career, Shawn led many initiatives in implementing key system and software solutions within large Canadian Retailers such as Bata Shoes, Athletes World, Holt Renfrew and most recently Loblaws.  In his past roles as a Business Analyst and Chief Data Steward, Shawn has witnessed and experienced first-hand the value of data, and its significant role in the analytics practice.

blue squares flying out of computer monitor