Daisy Intelligence’s Gary Saarenvirta dismisses the hype behind the ‘data-is-the-new-oil’ approach when it comes to AI (Photograph by Christie Vuong)
In the world of grocery retailing, the first months of the pandemic will go down as a kind of Battle of Waterloo moment, when all supermarkets had to face off against radically unpredictable consumer behaviour. The behaviour ranged from panic buying (toilet paper) to comfort cooking (flour, yeast) and stockpiling (beans, pasta).
For some chains, swaths of shelf space continued to present a troublingly gap-toothed look long after the reopening began, while others seemed to have worked the knots out of the supply chain. For the latter, their capacity to roll with the pandemic punches may be tied to the effectiveness of their recently installed artificial-intelligence systems’ ability to anticipate buying patterns in a crisis.
Daisy Intelligence (No. 118 on Growth 2020), one of the fastest-growing contenders in this space, says its clients adapted to the volatility thanks in part to its software. “Every week, new patterns emerged,” observes founder and CEO Gary Saarenvirta, a hard-driving engineer who boasts that his firm’s software is based on inviolable mathematical principles instead of the nebulous statistical models that drive most AI algorithms.
The pandemic outed the “pretenders”—those AI firms touting systems that couldn’t predict their way out of the frozen foods section, he says. “COVID was good for Daisy. Our system could see all the dynamism.”
The 48-person firm, founded in 2003 and now based in a brick-and-beam warehouse in downtown Toronto, has partners in nine countries and operates in two verticals: it helps supermarkets select and price products for weekly promotions, and it spots fraudulent claims for large insurers. In both cases, the company uses two years of its clients’ sales or claims data to build mathematical models that generate automated recommendations. Saarenvirta says its software “can bump revenues by 3% to 5%.”
He describes Daisy’s analytics—marketed (and trademarked) as its “Theory of Retail”—as an automation tool that’s meant to remove intuition and guesswork from decisions about those weekly, featured sale items, such as those that appear in the paper flyers that clog mailboxes. The system works through a granular analysis of how the purchase of one item produces a kind of domino effect.
For example, a flat of ground beef on special during a balmy stretch of summer may lead someone to buy buns, condiments, tomatoes, onions and pickles. “Our math finds the optimal combination of products [to promote],” he says, pointing out that in a supermarket with 100,000 SKUs (stock keeping units) and 2,000 products on special at any given time, the number of permutations is staggeringly high. “Every week in retail is unique.”
Yet unlike many AI entrepreneurs, he does not bow down to the computer-science gurus behind AI innovations that have yielded technologies like the Alexa voice-recognition software.
An aerospace engineer who spent years with LoyaltyOne (Aeroplan) and IBM’s data-mining group, Saarenvirta says Daisy’s algorithms aren’t designed to generate statistical guesses but, rather, to identify connections. In other words, under what conditions does the purchase of one item trigger the purchase of others? He rejects the idea that these relationships, and the predictions that flow from them, can simply be culled from reams of transactions records.
“We got all hopped up on ‘data-is-the-new-oil,’ ” he states provocatively. “Total bulls–t. We’re doing conventional AI. The computer-science version is totally off-base.”
For further evidence of Saarenvirta’s math-geek outlook, look no further than the name Daisy, selected because the flower’s yellow core has a pattern that can be described using a famous mathematical sequence of Fibonacci numbers.
The firm’s publicly acknowledged customers include a regional U.S. chain, Harps; an organic food chain based in North Carolina; Earth Fare, which succumbed to Whole Foods this spring; and Walmart. (Daisy declined to provide further details.)
Dalhousie University professor of food distribution policy Sylvain Charlebois observes that in a high-volume/low-margin business that’s been historically resistant to automation, dominant companies play their cards close to the vest when adopting new technologies. But, he says, most chains are now looking at AI-based systems, which perform tasks from automatically selecting promotional products (Daisy’s specialty) to managing supply chains and developing futuristic innovations such as “smart” shopping carts.
The pandemic, and the surge of interest in e-commerce grocery shopping, has drastically accelerated this transition, Charlebois adds. He points to Sobeys’ new Voilà e-commerce service, and Ocado, a U.K.-based online shopping platform that relies on smart systems. “COVID has built a case for predictive analytics,” he says.
Amazon is taking the revolution one step further with Amazon Go. The bricks-and-mortar supermarket has abandoned checkouts in favour of an extensive deployment of scanners and sensors that tally up what consumers have purchased, and also track how they move through the store and what they look at. The first Canadian location opened in Toronto’s Eaton Centre.
While behemoths like Walmart and Amazon invest heavily in tech, traditional supermarkets are more staid and see their business as unique—a culture that may take some time to overcome, says Charlebois. “Daisy’s very rational and Cartesian way of seeing a grocery store is upsetting to some people in grocery.”
But Daisy’s early-stage investors clearly agree that the company has plenty of runway. The firm has completed two financings, a $5-million debt round in 2018 with Espresso Capital, and then a $10-million equity infusion led by Framework Venture Partners (FVP) in September 2019. Saarenvirta says he plans to do more fundraising in the coming year, with net proceeds going toward international expansion, sales and marketing. He expects the firm to double in size within a year.
FVP partner Peter Misek says his analysts put Daisy’s algorithms through their paces to confirm the company’s claim that its system drives top-line growth. “We tore apart their code base and looked at all the algorithms and looked at the data pre and post,” he says. “We got comfortable that there was real effectiveness.”
Misek adds that FVP’s confidence in its investment in Daisy has only grown through the upheaval of the past several months. Grocery shopping, he contends, “will never go back. There’s been a fundamental acceptance of online [grocery shopping], and also the scope and style of physical retail. The future will not look like the present.”
