We Bring You Expert Insights and Industry Data, Monthly.

Click Here For Current Issue!


Retailers question the value of legacy forecasting systems as artificial intelligence accurately evaluates a broader variety of demand influences at unmatched speeds.

Troy Prothero

Using machine learning or artificial intelligence (AI) technologies to forecast grocery demand is not new. Some industry innovators began testing smart systems more than a decade ago. However, many others have gotten more serious about the technology in recent years in no small part due to the limitations of historical sales data-based legacy systems, which became painfully clear in the early days and months of the novel coronavirus pandemic.

“Historical reliance on time series data broke completely, as did the longstanding paradigm of JIT (just in time) inventory fulfillment,” explains Troy Prothero, vp of product management, supply chain solutions at SymphonyAI, Palo Alto, Calif. “The problems were both acute, as the pandemic created unanticipated patterns and changes, and chronic since we now have a three- to four-year rolling period of data that is out of step with historical and likely future patterns,” he continues. “With demand forecasting AI, we can project current, con- textual data into the future for more accurate forecasts even in the absence of relevant well-structured time-series data.”

Simon Joiner

“Everybody realized that their legacy systems and their legacy processes — whether a behemoth like an ERP system or spreadsheets — were too slow, their granularity was too high, and their ability to react was too poor,” adds Simon Joiner, director of product management at Dallas-based o9 Solutions.


To a man, subject matter experts we talked to say they have seen a marked increase over the past two to three years in the number of grocers onboarding AI-based systems to improve their forecasting performance.

Ben Wynkoop

Ben Wynkoop, global industry strategist for grocery and convenience at Scottsdale, Ariz.-based Blue Yonder, says that advances in AI technology have made it a dynamic predictive tool that continues to improve as it teaches itself. But generative AI is set to take those advancements to a whole new level.

“Imagine if I gave you 1,000 people with computers, and every single day they shared best practices and learned from each other. And the next day they got better at what they did. Now you could take the humans out of it and say that’s just computers. The computers are talking to each other and learning from each other,” explains Wynkoop. “Today, we’re using hundreds of data points in predictive analytics.”

He adds that AI’s ability to connect the dots cannot be matched by humans. He pointed to a retailer in California that experienced a sizable increase in the sale of Hall’s mentholated cough drops after the state banned menthol and flavored tobacco. Humans would have taken weeks or months to discover a correlation that took AI days. Retailers can use information like that to communicate with the brand manufacturer so it can build this new understanding into its production output.

Post-pandemic, more retailers are turning to artificial intelligence to more accurately forecast demand.

‘AI demand forecasting will indicate the product categories and items that have the highest demand and can help you identify which are ripe for private label alternatives with higher margins.’

Joe Smirlies

Experts see grocers being attracted to the ease of use and other benefits promised by generative AI. “The reality is I can adapt models with new data inputs without having to reengineer my algorithms. I just have to wait for the impact” says Joe Smirlies, senior vp of product management at Toronto-based Invafresh. “We still need industry expertise to understand what it means to value that impact for the algorithm. We have already made many efforts to score additional extensive data- sets. You know, cannibalization has factors that go beyond just the limited category. We’re starting to look at cross-department and cross-category influences. We’re clearly looking at the weather.”

SymphonyAI’s Prothero says contextual data takes into account both macro inputs such as seasonality and micro inputs such as sporting events in local markets and areas and has the ability to take “real-time supply chain disruption information to adjust forecasts appropriately.”


Prothero says AI was tailor-made for grocers to successfully grow their store brand businesses. “It’s particularly useful if you’re looking at adding new items in your private label portfolio,” he explains. “AI demand forecasting will indicate the product categories and items that have the highest demand and can help you identify which are ripe for private label alternatives with higher margins. You can also feed product attributes and characteristics to the forecast and do what-if scenarios to best understand what new products are going to drive the most demand for private label introductions.”

Santiago Garcia-Poveda Maria

“(Store brands) traditionally have been managed just like another supplier and that doesn’t give you a chance of making the most of it,” adds Santiago Garcia-Poveda Maria, global vp of retail, apparel and footwear at o9 Solutions. “[AI] can provide more visibility of the expected demand two months from now or three months from now, [so retailers can] plan the right raw materials, the right production levels, the right inventory levels and get better costs out of it. There are opportunities to leverage AI in the longer term whether to plan logistics capacity or private label sourcing on a more strategic level.”

Frozen and refrigerated are good places to start AI onboarding. “Fast-moving items lend themselves to better forecast accuracy, so the richness of data means that AI demand forecasting is particularly accurate for frozen and refrigerated products,” says Prothero. “It’s a great high-ROI area for grocers to [consider] adopting AI demand forecasting for their first foray into the new-generation technology, and then phase it into other areas of the overall store with some best practices already under their belts,” he says.


Experts say that retailers should not let unrealistic expectations color their views on AI’s utility as a forecasting tool. “After taking the ordering out of the hands of humans, they need to say, ‘OK, I’m gonna let automated ordering do its thing.’ Some mistakes will be made. But humans make mistakes. That’s what we always forget. We want technology to be 100% perfect, 100% of the time. Humans are not 100% perfect even 1% of the time, especially when it comes to planning out demand,” says Wynkoop.

“Computers won’t ever be perfect, but they will get better than a human will be if they can learn from themselves,” he continues. “Now’s a great time for regionals and smaller retailers that have not invested heavily in technology to be getting into it. Because all of that learning has already been done. All the technology development is there, and there will continue to be better iterations.”

George Anderson

George Anderson

Leave a Reply

Got News?

Let us know!

Email Warren at

On Key

Related Posts


Grocers are increasingly partnering with third-party logistics providers to bring together the physical, technological, and human resources needed to meet today’s supply chain demands. Grocers


Comments from senior industry executives in March. “When we launched Restock Kroger several years ago, we knew that a strong omnichannel experience was a key

Databank – April 2024

FASTEST FROZEN Sales of frozen products in supermarkets, drugstores, mass merchants, military commissaries and select club and dollar stores combined totaled $18.65 billion during the


Are grocery retailers ready if TikTok is banned? If TikTok is on its way to being banned in the U.S., what will it mean for



contact us

Or reach us directly:

Paul Chapa, Co-Founder & Managing Partner
913-481-5060 or

Warren Thayer, Co-Founder & Managing Partner
603-252-0507 or