Sullivan says that such technology can help assortment planners leverage the forecasting found in what she calls “science-based pricing, which accounts for transference of demand, including halo and cannibalization effects, that can result if an item is understocked, creating stockouts, or overstocked and languishing at the shelf.”
Longer term, the use of AI for product assortment forecasting at food retailers can also play a role in helping to design more efficient, smaller or otherwise less costly stores. Artificial intelligence additionally promises to play a bigger role in helping food retailers grow more dominant in e-commerce — a job that includes taking on massive players such as Amazon and Walmart.
“Unlike in a physical store, where prices are more or less static for days and maybe even weeks, with online shopping, prices can change rapidly — Amazon uses its dynamic pricing algorithms to review prices of millions of its SKUs, including for groceries, every two minutes,” observes Intelligence Node’s Sularia. “To thrive in this fast-paced e-commerce environment, AI-driven dynamic and competitive pricing solutions will be the only way to really optimize pricing.”
Overcoming Inertia
The use of AI for product assortment planning and related forecasts does face some significant inertia, however.
“Right now, the opportunities for product assortment planning are limitless because AI isn’t really even being used,” says Stefan Kalb, co-founder of Seattle-based Shelf Engine, whose technology is designed to reduce retail food waste. “Most buyers still rely on syndicated data, whether from IRI or Nielsen, while large retailers are stuck with national planograms because managing a radically different assortment at the individual store level, or even regional level, is too complicated.”
But times are changing — thanks in large part to the challenges of the pandemic, but also because of the increasingly digital and mobile nature of grocery consumers.
“The drastic shift in demand across nearly every category, combined with continued consumer uncertainty, will make relying on traditional data for assortment planning increasingly difficult in 2021,” notes Kalb. “While some retailers may choose to stay the course and hope for a return to pre-pandemic norms, others may see this uncertainty as an opportunity to try something new with an alternative approach.”
That alternative approach of using AI depends on retailers taking a risk with new and unfamiliar technology, he adds. That’s not always easy, of course, especially for grocery operations with relatively modest technology budgets. But the long-term payoffs can be significant, especially when one considers the potential scope of AI forecasting systems.
“AI systems allow retailers to more quickly and accurately react to unexpected shifts in product demand,” explains Kalb. “These bits of data can account for things like local weather patterns, sporting events and other external factors that impact the inexplicably varying sales cycles of some stores. AI also incorporates other types of information affecting shopping behaviors, like patterns involving complementary and competitive relationships. So, say your store is out of crackers; there’s going to be a smaller chance that shoppers will buy cheese.”
Importance of Flexibility
Consider this thought experiment: What revenue might a food retailer have gained early in the pandemic if it used AI to forecast a run on hand sanitizer and bought supplies accordingly?
“The challenge is identifying those insights faster, and being flexible enough to execute,” says Matt Schwartz, CEO of San Francisco-based Afresh, whose AI technology focuses on accurately ordering fresh food to increase sales and reduce waste.
The challenge is also knowing what you’re looking for via AI forecasting technology — that is, knowing exactly how to use this new tool.
“AI is really, really good at optimizing decisions within areas where there is a lot of data,” observes Schwartz. “As a result, with granular sales history, a machine-learning model can learn the effects of assortment, and use that learning to predict an optimized assortment going forward.”
He adds: “‘Optimized’ here usually means profit maximizing, but can also consider maximization of customer lifetime value, revenue maximization and other factors.”
As Schwartz puts it, human decision-making also is important, depending on the retail function, as experienced, insightful workers will sometimes know more than the machine.
“This is especially true in the perimeter, where product attributes are so dynamic — produce availability, taste and cost change so rapidly,” he says. “One example: With weekly produce ads breaking, a retailer found success in leveraging the machine to determine the optimal number of cases of raspberries per store, but allowed for discretion in dummying up the display after a load of short-shelf-life berries were received — this enabled stores to maximize their sales while minimizing shrink.”
Flexibility also matters when it comes to choosing the right AI system to deploy.
“Given how dynamic the landscape is in 2021, it’s critical for any system to be able to adapt,” advises Schwartz. “Certain AI technologies are more flexible than others — both in the construction of their algorithms as well as in the way human-centered workflows are built around them.”
Those are all massive choices for any food retailer, but at least one thing is becoming simpler about AI technology: implementation, which in turn affects cost.
“Any system implementation is not trivial, but with that said, new cloud-based software-as-a-service technologies can be adopted much faster than legacy systems of the past,” says Schwartz. “Whereas on-premise older technologies might have taken years to get off the ground, new AI technologies can be up and running in a couple months.”
AI has a way to go in food retail, but interest is building, even as best practices are scarce. You can bet, however, that AI’s impact will be felt pretty deeply before too long.
“Using AI is like going from a screwdriver to a power drill, a hammer to a nail gun,” notes Schwartz. “AI is a supercharging resource that can take almost any manual process or decision and augment it into a superhuman one.”