Machine Learning Solves Retailers’ Biggest Challenges — Part 2

Retailers are looking for solutions to give them a competitive edge, optimize their operations and keep costs low as the COVID-19 pandemic continues to wreak havoc on supply chains and in-store shopping. One potential solution is leveraging mathematics in the form of machine learning.

A subset of artificial intelligence (AI), machine learning involves software and technology that identify patterns of behaviors and execute tasks without being specifically programmed to do so. For many retailers who hesitate to make the jump to this new technology, machine learning might not seem like a must-have addition to their technology suite. But the reality is that in the age of COVID-19, when every decision can make or break a brand, machine learning is quickly becoming the gold standard of retail innovation, by delivering quick, real monetary value, both for top line and margin.

Machine learning might be the solution you didn’t know you needed. In this follow-up to Part 1 of my machine learning series, we’ll explore four more burning questions from retailers that machine learning can answer.

4. How can machine learning identify areas for operational improvement?

Accurately locating opportunities for improvement, no matter how minor, can give retailers the competitive edge they need to ensure cost-effectiveness and operational efficiency. Most importantly, successfully identifying and acting on these opportunities positively impacts the customer experience. Machine learning accomplishes this quickly and precisely by identifying anomalies in retail data.

Anomalies occur when real-time activities do not properly align with “typical” behaviors. This can include anomalies in internal data, such as employee and delivery behavior, as well as external data, including vendor activities. Machine learning can continuously monitor this data in real time, verifying that it’s consistent with expected behaviors. This can range from a delay in delivery to a drop in employee sales.

The solution detects and flags any anomaly with a high true-positive rate. This notification comes paired with a step-by-step workflow to resolve the finding or guide the associates to investigate further for an effective resolution. As a result, retailers can identify new ways to maximize revenues.

5. Can machine learning forecast shrink and risk?

Accurately predicting and planning for shrink and risk is a notoriously difficult task due to the influence of unpredictable outside factors like mis-shipments, organized retail crime and operational losses. Rather than forecasting based on an accepted and planned timeline of events, the root causes of shrink and risk lie in outside factors that retailers cannot control.

Machine learning takes the guesswork out of forecasting these events through intelligent, data-driven algorithms. These algorithms analyze your shrink rates during your most recent cycle count and then combine them with proven shrink indicators such as employee attrition, theft, returns, associate tenure, receipts, new product introduction, traffic and more. This allows the solution to generate a reliable forecast. Over time, the algorithms “calculate” the variables that most strongly impact shrink to continue to evolve and improve forecasting results into the future.

By leveraging machine learning to forecast shrink and risk, retailers have advanced visibility between physical inventories and targeted cycle-counts (or prescribed counting), while simultaneously identifying high-risk stores or products that require corrective actions earlier rather than later, closing any gap between expected and observed inventory. Machine learning empowers retailers to plan for the unexpected.

6. How can I identify cashiers who are non-compliant, need training or even committing fraud?

With the current environment making margins tighter than ever, retailers are cracking down on internal fraud, especially fraud committed by their own cashiers, as well as non-compliance caused by training gaps. Retailers are wary of two types of cashier fraud, in particular: both of which can be difficult to identify as intentional fraud rather than a simple mistake:

• Sliding: Sliding fraud occurs when a cashier purposefully obscures a product’s barcode when passing it over the register’s scanner. The customer is then able to exit the store without properly paying for it.
• Sweethearting: Similarly, sweethearting occurs when a cashier scans the barcode of a cheaper item in place of a more expensive product. In this scenario, a television could be bought for the price of a pack of gum.

Typically, these instances of fraud would only be caught by a manager witnessing the act in person, or by watching hours of CCTV footage. However, with machine learning, cashier activity is automatically monitored to identify fraud. For instance, algorithms analyze cashiers’ behaviors such as per-minute or hourly scan rates to determine benchmark averages. Should a cashier’s scan rate deviate from this benchmark by three standard deviations, indicating there was a longer-than-normal lapse between scans, they are flagged as a potential case of sliding. Managers then conduct an efficient and impartial investigation using machine learning-backed evidence.

7. How can machine learning help my store mitigate risk associated with controlled-substance sales?

Retailers selling alcohol, tobacco or cannabis understand the risks around non-compliance better than anyone. A sale made to a minor, whether purposeful or on accident, could lead to fines, lawsuits and — worst of all — a PR nightmare. These severe consequences require a comprehensive preventative solution.

Machine learning algorithms monitor employee behaviors to identify evidence of non-compliance. For example, liquor stores can leverage machine learning to monitor the customer birthdays employees enter into their database upon checking an ID. Should an employee consistently enter the same birthdate or a suspicious date, indicating they are failing to properly examine customer IDs, the solution flags asset-protection personnel to follow up on it. Depending on the severity or frequency, the follow-up directives could range from a full-on fraud investigation to just asking the cashier to watch a compliance-training video to close a knowledge gap. Machine learning identifies other instances of suspicious behavior as well, such as unusually low sales alongside a boost in hand-keyed prices, indicating an employee is overriding prices to provide friends with unauthorized discounts. Machine learning allows controlled-substance retailers to ensure they are protecting their business and customers by remaining compliant with state and federal laws.

The answer to retailers’ biggest challenges is found in machine learning

An increasingly competitive industry requires competitive solutions. For retailers searching for ways to maximize sales and operational efficiency, the answer is likely found in their own data, and the key to unlocking that data is machine learning. Machine learning empowers retailers to identify opportunities for improvement every day, giving their employees the tools they need to succeed. By leveraging machine learning solutions, retailers are leading the way in innovation.