The Hershey Company is using a new in-house AI tool with computer vision and augmented reality to decide where s'mores displays go in stores and how well they perform, according to Stephanie Berman, vice president of retail at the company. Hershey also deployed AI in its marketing-mix model this year to track real-time consumer data and adjust ad spending faster.
S'mores-related sales of Hershey products have grown in the double digits year over year to date. Hershey estimates s'mores generate $200 million to $250 million in overall sales during peak season, roughly from Easter through Labor Day.
Computer vision for store displays
The display placement platform has been used in several thousand stores, Berman said. It uses augmented reality to visualize digital 3D product displays in real-world retail environments, and image recognition to provide insights on merchandising execution, location effectiveness, and sales performance. Sales teams photograph displays in stores, and the AI tool combines those images with point-of-sale data to determine where displays work best.
Hershey uses pallet trains - standalone displays - to feature its chocolate alongside graham crackers and marshmallows from other brands. During s'mores season, pallet trains outperformed end caps by 101%, according to the company.
"The retailers seeing the strongest returns are those that place displays where they naturally align with shopper missions and seasonal occasions rather than simply allocating more square footage," said Kiara Barrett, global head of thought leadership for Circana.
The tool also helped identify specific placement issues. "Like being below the fold on search results, being in low-traffic parts of the store or with the wrong neighbors (like non-organic displays) in an organic aisle will reduce the effectiveness of a larger display," said Michael Hill, senior partner and head of commerce strategy for Kantar. The company found, for example, that s'mores displays placed near camping products drew high consumer interaction.
Circana merchandising data shows marshmallow and graham cracker sales rise most around Memorial Day, Independence Day, and Labor Day, with lifts as high as 300% in peak weeks.
AI in the marketing mix
The AI marketing-mix model pulls and standardizes data from social, search, and streaming platforms to inform ad spending decisions based on current market conditions rather than months-old data. Hershey said this turned what once took five months of manual spreadsheet work into an overnight process. For s'mores, the company found a roughly 20% year-over-year increase in retail media return on ad spend.
Hershey also used social media to gauge consumer preferences this season. The company ran a poll asking people whether they like their s'mores more toasty or more gooey - 70% said toasty.
"The AI marketing-mix model was how we were able to understand consumer emotions, like the poll, and understand where and when to engage them, so we could make the right investment in real time," Berman said.
"The big thing for us as we look at AI: It's been about real-time data, it's been about getting closer to the consumer in more real time, and it's about partnering with our retailers on how we see the consumer behaving, so that we can iterate faster and more meaningfully with the consumer," she said.
Hershey's sales also got a boost from a limited-time caramel Hershey's bar. Berman said the company will look at what that means for 2027 product innovation based on the caramel item's success. For marketers tracking AI applications in retail, the case shows how tools like computer vision and automated marketing-mix modeling can shift decisions from quarterly cycles to near-real-time adjustments. The approach pairs well with broader training on AI for Marketing Managers, which covers using data for faster campaign decisions.
Why this matters for marketing professionals
Hershey's results tie specific AI investments to measurable outcomes: a 101% lift for pallet trains over end caps, a 20% increase in retail media return on ad spend, and double-digit sales growth. The key takeaway for marketers is that AI's value here came from combining real-time data with specific retail contexts - not from replacing human judgment, but from speeding up the feedback loop between display placement, consumer behavior, and ad spend. For those building similar capabilities, the practical starting point is identifying one seasonal occasion, one retail metric, and one data source to test against, rather than trying to overhaul the entire marketing stack at once. Courses on AI for Marketing offer a structured way to learn these methods before applying them to your own campaigns.
Your membership also unlocks:
