Sweetgreen is building a 'Spotify for food'
Could it? Would it? What?
“We want to build what Spotify did for the playlist, for the menu,” Sweetgreen CEO Jonathan Neman said last week on This Week in Startups, a popular tech podcast.
What he means, I think, is that Sweetgreen wants to call out exactly what customers need and want to eat to an extremely specific degree. As Spotify offers listeners songs that can surprise and delight a listener during their Daily Mix 4, so can Sweetgreen surface a bowl packed with the right kind of nutrition and the right combination of ingredients to delight customers while keeping them healthy. It could even tap a robot to make it.
TWiS host Jason Calacanis is known inside tech circles as a successful investor and entrepreneur, and outside of them as one of the men that helped Elon Musk through his Twitter takeover. He’s an admitted fan of restaurants and food robotics, and will remind anyone listening that he’s an investor in the Bay Area’s Cafe X coffee robots. For all the men’s tech talk, the podcast interview does a decent job of reminding listeners that Sweetgreen is a restaurant, not a tech company — a distinction that’s frequently (and purposely, IMO) blurred as the salad-forward chain works to turn a profit.
Sweetgreen delivered its quarterly earnings two weeks ago, and Neman said he believes the business can turn a profit in 2024. He shared more plans for the year that include opening 11 restaurants staffed by its “infinite kitchen” robotic technology. Sweetgreen’s robot, currently working front and center at two of its 225 locations, can dispense nearly every one the chain’s 55 ingredients, with notable exceptions including avocado and salmon. (These delicate ingredients require a human touch, Neman said.)
The bots are expensive — it costs half a million dollars to build a robotic location and even more to retrofit an existing one, but the infinite kitchen’s early performance shows a 10 percent lift in average check size and 7 percent margin increase at these locations thanks to its error-reducing technology. It can also portion ingredients in extremely precise amounts, down to the gram, Neman said, in what could be a hallmark of Sweetgreen’s future personalization.
Listening to Sweetgreen’s leader on a technology podcast, one might get carried away by future possibilities. “Spotify for food,” with its promise of ultra-personalized menus, is one of them — though Neman said Sweetgreen is already testing it internally.
“Sweetgreen is perfectly positioned to be able to take what your [health] goals are, understand more about you — what’re you’re willing to share with us — and be like, ‘You don’t have to order off the menu,’” Neman said. Instead, the app could say, plainly: Here’s what you should be eating.
This healthy eating focus builds on Sweetgreen’s core DNA — to become the hip and healthy McDonald’s of our generation — and some of the chain’s targeted initiatives, like ditching seed oil in favor of avocado and olive oils, a popular switch championed by the paleo set. (It’s one of those things that a small subset of Sweetgreen’s audience cares about, Neman explained, but those that care really care.)
It’s one thing to talk about personalized health goals that might be supported by a restaurant known for serving healthy, delicious, and hip food.
It’s another to talk about those goals as they could be realized in a tech-infused future. Willing diners’ health info could feed into Sweetgreen’s app that could tell its robots to perfectly portion an Instagram-worthy bowl of nutrition, suiting you and only you. I see it! It’s provocative, it’s futuristic, it’s the kind of speculation that hooks a techie like Calacanis, who’s willing to extol Sweetgreen’s virtues on a tech pod even if he doesn’t get the name right.
But is it realistic?
I don’t know the answer, so I asked someone who does. Matt Newberg, fellow food technology enthusiast and founder of HNGRY, has spent many hours thinking about, covering, and testing technology that supports personalized nutrition, from blood glucose monitors to stool sample testing that uncovers gut microbiome data. On that last one: according to Newberg, two people selected at random might share just 10 percent of similar gut bacteria, even though they’re likely to share 99 percent of their DNA, which speaks to the extreme level of personalization possible in food. (If you’re really into this, watch this video Newberg produced about his personalized nutrition adventure.)
Newberg says that Sweetgreen’s ability to execute on detailed nutrition data — when they have it — won’t be unique. “There is a very wide canvas of potential players — the entire food system today — that is going to be able to act on this,” he said in a recent interview. “It’s simply technology that’s going to tell you, ‘Buy this, not that.’”
This sort of thing already exists. There are companies, Newberg told me, like January.ai, that track eating behavior, activity, and even blood sugar, to analyze which foods make a person feel best (and worst). Spend a few weeks recording data to unearth info and patterns that help build healthy habits. It’s reasonably personalized nutrition for fitness-tracker types.
“If you plug that [data] into a Sweetgreen, or if you plug it into any restaurant software interface, you could essentially score every meal,” Newberg, who presciently mocked up what this might look like inside Sweetgreen’s mobile app nearly four years ago, said.
But in terms of going further, offering even more detailed, personalized, hyper-specific ingredient combinations tied to one person and one person only, Newberg says we’re not there yet. There are promising studies on “food as medicine” and related topics, but overall, the burgeoning personalized nutrition business is fragmented, with many barriers to mass adoption.
“In order for this to work,” Newberg continued, “we need comprehensive data that is easy to collect, that will be a seamless opt-in for consumers — blood labs, sleep trackers, fitness tracking, gut microbiome tests, blood glucose levels — these are all different silos that need to be considered holistically. Then we’ll need the research to understand it and get to a point where you could [offer] a recommendation.”
In other words: “The output of this is the easiest part. The hardest part is the input.”
I’ve probably taken this idea far past Neman’s intentions when he spent only a few minutes talking about it.
But this level of detail was absolutely in the spirit, if not actual substance, of the podcast discussion. I think that we can agree that offering up a voluntary stool sample — which Neman does mention in his interview as an extreme example of how people are thinking about personalized nutrition — is only attractive to a certain subset of very interested consumers. But like the seed oil detractors, they’re passionate about the cause.
That doesn’t make personalization a bad idea, and it’s a goal worth working toward, whether in service of a restaurant that thinks itself a tech company or simply one that wants to broaden its total addressable market. Neman didn’t share details on exactly what type of personalization Sweetgreen is currently testing or when diners might see any version of it on the company’s way to creating the restaurant version of the perfect playlist. But it’s clear he wants Sweetgreen to get there fast.
“Today, the ‘Spotify of food’ looks a whole lot like ‘What was my last order? What are people who order like me ordering now? It’s called collaborative filtering, it’s something that Netflix uses and Amazon uses,” Newberg explained.
For Sweetgreen’s purposes, this could be enough. There’s plenty of value in the information we have and are willing to share about our own bodies. A person who knows they’re iron deficient doesn’t need a blood test to confirm the exact deficiency or a robot to portion out the perfect amount of protein. They just need to know it might benefit them to eat more red meat.
Coincidentally, they’re in luck. As of early February, steak’s on the menu at Sweetgreen.
