At 3:00 p.m. on a recent Tuesday in August, a group of a dozen Clover Food Lab employees stood around a stainless steel kitchen slab in the company’s East Cambridge, Mass., commissary. After a short discussion about sourcing ingredients from local farmers, the hour-and-a-half eating marathon began with a lentil, curry, and coconut salad. Spooning bites out of plastic sample cups, Ayr Muir, founder of the Boston-based vegetarian fast-food chain commented that he could taste cider vinegar. Others picked up on the presence of cinnamon.
On the “FOOD DEV NOTES” sheet, under the “CONSENSUS?” column, Muir wrote: “Try again,” then moved on to the next item, a quinoa and cauliflower salad, which had too much sauce and not enough color. It too, got the “try again” status. This process continued through seven more salads, two spreads, and a Gobi Manchurian sandwich with cauliflower and pita bread.
This is the first step of Clover Food Lab’s data-driven menu development process. Every Tuesday, the company holds a tasting open to anyone, including customers, to sample or present experimental dishes for consideration to sell in Clover’s 11 locations across the Boston area. Before a dish goes mainstream, it has to pass through various testing phases: first, the food development meeting, followed by testing in one or two restaurants. If the data looks good, it will get sold across all food trucks and stores. The tests, however, never end. Even menu staples are constantly being tested and tweaked.
“In my former life as an engineer, testing out ideas and challenging assumptions is just what you do,” said Muir, who worked at McKinsey before starting Clover. “It’s hard for me to imagine not having this kind of approach.”
Clover’s constant recipe tweaks are a highly unusual practice for a fast-food restaurant. Most chains operate on consistency. Go into any McDonald’s on any day of the year and you expect the same french fries and Big Mac. “Most operators are too lazy and egotistical to want to change things” one restaurant industry consultant said. Over the course of a month, Clover changes up to 80% of its menu, says Muir. That partly has to do with seasonal availability; Clover won’t serve vegetables out of season because they don’t taste very good. In addition to that natural flux, Clover’s experiments often result in menu changes.
Muir, a descendent of the naturalist John Muir, is a scientist by training. He studied material science at MIT, before getting a degree from Harvard Business School. Following his stint at McKinsey, he decided to go into food service because of its potential environmental impacts. “I thought I was going to get into wind energy or something; I cared about environmental issues,” he said. “I read about food and the environment and started researching more. I could have a bigger impact by changing what people are eating.” He found that meat consumption has a bigger impact on climate change than transportation and energy. Therefore, Clover sells only vegetarian and vegan dishes. He hopes the food will draw carnivores, getting them to eat fewer meat-based meals, thus saving the planet one lunch at a time.
With no culinary background, Muir wanted to test the food at his first food truck with more nuance than sales figures. He started by doing what he knew best: collecting data by interviewing people who visited the truck. As the brand has grown, so have his experimentation methods. “What we’ve been trying to do is learn as much as we can, and our approach to that is to embrace failures,” he said.
In the spirit of gathering as much data as possible, Muir takes a big-tent approach to the food-development meetings. There’s no voting process (although comment cards are collected), and Muir tells me that certain voices have more pull than others. If everyone tastes it and likes it, then it gets the go-ahead, Muir says. But sometimes dishes that need work still make it to the next phase. Ultimately, Muir makes the final decision. At a recent meeting, five dishes got the okay to sell in select test locations. From there, they will get scaled up and put on menus, where the large-scale testing begins.
The information mining doesn’t stop at the menu. At Clover, every employee is a data scientist. When customers walk into one of five brick-and-mortar locations, cashiers take orders with iPods that double as data collection hubs. Are you a first time Clover eater? What did you think of the sandwich you ordered last time? Customer responses, along with any other useful observations, get logged into the Notes app, which syncs up to a single hub. In addition, Clover puts out a monthly survey and trolls Twitter, Facebook, and Yelp. All told, this process yields more than 3,000 comments a month that Clover analyzes for signs of success and failure.
Given the company’s focus on data, the analysis aspect of the process is pretty low-tech. Clover’s director of communications reads through the notes, bringing any issues to the food development meeting. Customer surveys run by both SurveyMonkey and Wuffoo include analytics tools on the back end. All of that intel, plus sales information, helps Muir determine if an item will find a permanent (for now) place on the menu.
First-time dishes often need tweaking. But even longtime menu favorites are subject to–and benefit from–data scrutiny. “Almost everything, even if we really love it, will go through a few iterations,” Muir explained. Most recently, Muir noticed complaints of a “too oniony” flavor in the signature chickpea fritter sandwich, the company’s version of a falafel. He brought it back to food development and found that their vinegar supplier had subtly changed its recipe. The investigation resulted in an overhaul of all the restaurant’s sandwiches containing pickled vegetables. This kind of thing happens all the time: Clover’s chickpea-fritter sandwich has gone through at least 34 changes since Muir founded the restaurant in 2008.
The process keeps egos in check, and ensures that quality, not emotion, wins the day. After that food development meeting, Muir’s opinions don’t matter. He might love an item, but if it tests poorly–like every version of a shredded-carrot salad he has tried–he will pull it. He believes that living by the data, rather than his gut, has made the food “markedly better” over the years.
While constant variety would work against any other fast-food chain–where customers have been trained to order old standbys by number–it has become central to Clover’s brand. When people go to Clover, they don’t expect the same burger and fries, they expect good food. (Clover has no tie to any food genre, serving dishes inspired from all over the world.) The dining experience feels, as one Cambridge local explained it to me, “very MIT.” It looks like a lab crossed with a Chipotle, with crisp, clean, white spaces. Clover, like any beloved brand of this era, prides itself on transparency. All of the food sits in clear fridges or bins waiting to be spooned into pitas.
Muir has ambitions of growing the chain nationwide. “I think we should be as big as McDonald’s one day,” he says. This fall Clover will open its first D.C.-based outpost, and a partnership with Boston-area Whole Foods stores is in the works.
But is experimentation scalable? Muir can’t possibly sit in food development meetings across the country giving the final word on what dishes make it out of the lab and onto the plates. After the tasting, Muir discussed various problems at the Boston locations: Are employees using the toasting timers? There was some scolding about substituting the wrong kinds of onions. Muir keeps tabs on each location’s implementation of food items, something he won’t have control over outside of the Boston area.
Muir is training someone to head up the D.C. location, and is confident that the product will only get stronger as his empire grows. “There’s this assumption that quality goes down as a restaurant expands or gets bigger,” he said. He doesn’t believe that. “The more data you have, the better job you can do. As we grow larger, our food quality ought to get much better, not worse.”