Coca-Cola’s headquarters building rises 29 stories above Atlanta. But the team working on a new kind of soda dispenser in 2007 was stashed away in the basement. Two levels of passcodes and one very stringent non-disclosure agreement were required to enter the lab. The endeavor, code-named “Project JET,” was kept separate from other R&D initiatives.
“Leadership set up a skunkworks team that was incredibly secretive and given resources, but kept very close to the vest,” says Jim Sanders, Vice President of Freestyle Business Development and Fleet Performance. Because the workspace was located underground, “the initial engineers called it the dungeon,” Sanders says.
The Project JET team was trying to tackle what Coca-Cola viewed as a big problem: Consumers wanted more choice. While Coca-Cola was addressing that by offering an array of flavors packaged in bottles and cans — like cranberry Sprite or Barq’s French vanilla cream soda—the self-serve fountain you’d encounter in a fast food restaurant was stuck in the 1960s, Sanders says. “At fountains, there were just six or eight choices, and it was really limited on diet or low calorie options,” he says. “The idea of more choice was the heart and soul of the project.”
Designing a new kind of fountain might also give Coca-Cola a competitive edge. Even though the company already commanded greater than 70 percent market share at fountain-based locations in the US, “every time we developed something new, the other guys in blue were quick to follow it on,” Sanders says, referring to Pepsi. Part of the problem was that both companies relied heavily on outside suppliers for technology.
This time, Coca-Cola opted to try to build something on its own, rather than just buy from third-parties, so that it could have “a unique solution that [restaurant owners] couldn’t get from the other guys,” Sanders says.
There are now about 52,000 of the Project JET machines out in the world, and odds are good you’ve used one, maybe at a Firehouse Subs or Five Guys. (It’s now known as Coca-Cola Freestyle.) But getting there wasn’t simple.
Reimagining the Soda Fountain
The common soda dispenser — push the lever and fill a cup — blends carbon dioxide, water, and concentrated syrup. But the flavor balance can be imprecise, as valve settings drift over time. And if the company wants to launch a new flavor, it faces a tough decision. Which existing flavor has to go?
The Project JET team, about a dozen engineers and managers, borrowed an approach called micro-dosing from the medical industry — a way to precisely control how fluids were mixed. What if you could do that same thing with the flavors of a soft drink?
“We initially thought we were going to use the traditional beverage suppliers to this industry,” Sanders says. “But in an effort to keep the intellectual property and the ideas to ourselves, we did a lot of the engineering, and we used a lot of new approaches to designing and manufacturing the equipment.”
Funding, and the approval to create a separate innovation team that was fully resourced with dedicated people, came from then-CEO Neville Isdell, Sanders says, as well as the company’s head of R&D.
The typical model for Coca-Cola was to give suppliers a design brief, and let them develop designs for the company to evaluate, Sanders says. “Then, we’d forecast how many we might buy,” he explains. “It was an arm’s-length…relationship.”
But Coca-Cola wanted to own the product — from the fluidic system that would allow precise flavor control to the refrigeration technology to the software. And, says Sanders, after a few false starts, the Project JET team realized that the company’s existing supplier base didn’t have sufficient “capability to work in some of those areas,” he says.
The Project JET team didn’t do everything itself, however: It found an outside partner that had worked on micro-dosing before, New Hampshire-based DEKA Research and Development, run by inventor Dean Kamen. (Kamen’s first successful product, invented in the 1970s, was a medical device that delivered precise doses of insulin for diabetics.) A Seattle company, Bsquare, brought expertise in internet-linked products.
The system Coca-Cola developed invited customers to use a touchscreen to select a beverage. Then, as carbonated water is shooting from a nozzle into the customer’s cup, “the drink is being mixed in mid-air,” Sanders explains. “The ingredients are being fired from about 36 different pumps, which gave you an absolutely fresh and perfect drink every time.”
Even so, questions always surrounded the secretive project. Executives wanted to understand, Sanders recalls, “Is there a business case here? Do we want to invest at the next level?”
Before the project had been launched, one of its original champions, Isdell, retired, and a new CEO, Muhtar Kent, took over. “One of Muhtar Kent’s first inputs to the project when he came on as CEO was that he brought in Pinanfarina [the Italian car design studio] and they helped us with the external look of the device,” Sanders says.
By the end of 2009, the company had given Project JET a real name — Coca-Cola Freestyle — and green-lit the plans to commercialize it. Freestyle would be able to serve up about 100 different beverage flavors.
The Alpha Test
But first, the company needed to test the device in a real restaurant. It chose a suburban Atlanta burrito shop, Willy’s Mexicana Grill, for the alpha test of the Freestyle machine, which had been hand-built by the engineering team.
Why Willy’s? “We knew the owner [Willy Bitter] very well,” Sanders says. “It was close to headquarters. And they allowed us to have three or four engineers in there constantly” to babysit the prototype machine. The team also shot videos of how customers interacted with Freestyle.
“We’re watching not just how consumers react,” explains Scott Harrison, Vice President of Platform Leadership at Coca-Cola, “but we’re trying to find the bugs we didn’t catch in the lab. … We can go back and look at the film, see the use case, and fix it.”
Another reason for choosing Willy’s as a test site was its size — at the time, Willy’s operated just a handful of locations around Atlanta. “Consumers could’ve rejected Freestyle,” Sanders says. “We might’ve pulled it out. And you wouldn’t want to disappoint one of your large national customers” if that happened.
And there were definitely the sort of snafus one would expect from what Sanders dubs an “alpha test.” Pininfarina had designed a swoopy, elegant lever for dispensing ice into the cup. But if you pulled your cup away too fast, and the machine emitted a few extra ice cubes, the lever would essentially punt them onto the floor of the restaurant. The team realized that “we’ve got to go back and get much more of an industrial, usable device, instead of this thing that was more about looks,” Sanders says. It was also too complicated to reboot the Freestyle, something that restaurant operators would need to be able to do on their own, especially if the machine was having issues connecting to the wireless network.
But consumers who tried Freestyle at Willy’s “fell in love with it,” Sanders says.
Not long after, the company created a Facebook page — one of Coca-Cola’s first — to see what consumers were saying about Freestyle. “One thing we’d see on that early Facebook page was, ‘Bring Freestyle to Peoria,’ or wherever the person was from,” Sanders says. “And you also had people telling each other how to mix different brands to get something that tastes like an ice cream treat or a certain candy. There was that notion of the consumer owning the experience.”
As Coca-Cola continued to shepherd Freestyle toward the market, it needed to make several big changes. First, the skunkworks group had to move out of R&D — and the basement — and into a more commercially-oriented part of the company, Coca-Cola’s FoodService and On-Premise division.
The company would rely on contract manufacturing firms to build the machines — the same way Apple builds iPhones. A new supply chain had to be designed, and Freestyle required different training for service technicians. And the machine “needed quite a lot of maintenance in the early days,” says Thomas Stubbs, Vice President of Engineering and Innovation for Freestyle.
Testing a Rental-Based Business Model
Coca-Cola also planned to price the Freestyle at a premium to the traditional fountain dispenser. And it wasn’t exactly selling Freestyle machines to customers — it was renting them. “The idea of a program fee is one of the other big, courageous steps we took,” Sanders says. “The company made the decision that we would own all of these devices,” and the customer would pay a monthly program fee ranging from $160 to $300 a month, depending on the size of the machine. That fee includes use of the equipment, the software, installation, and any service; it doesn’t include the cost of flavor cartridges that go into the machine to enable it to produce beverages.
The pitch was, according to Sanders: “[I]f you’re looking for low price, use our traditional fountain. But if you want to innovate and grow your beverage business with young consumers, [Freestyle is] a premium option.” The company hasn’t disclosed how much it spent on the Freestyle project, but it spent $100 million upgrading an Atlanta plant so that it could produce concentrated ingredients for Freestyle cartridges.
Taming a complicated new technology while deploying a new business model could be headache-inducing. “We would tell each other, ‘If you ever have a day when you feel this is too difficult, go out and just watch the consumer in restaurants that have [Freestyle],'” Sanders says. “They’d be giggling, and sharing with each other how to use it. The engineers would have to rub that into their wounds like a salve. There were plenty of technical hurdles we had to go through…but we knew we were onto something that would change the game.”
Once Freestyle machines started showing up in restaurants around the country, in mid-2010, the wounds began to heal. “We have seen it, the promised land,” wrote Elina Shatkin in LA Weekly. “The soda dispenser of the future.” The Disney Food Blog dubbed it “the coolest Coke machine you’ve ever seen.” Esquire doled out negative reviews to nine of the more exotic flavors (Powerade ION4 Lemon “tastes like the kind of lemonade you can make at home if all you have is a lot of water and a lemon rind”), but reluctantly admitted that Sprite with grape was pretty good. The company’s Facebook page quickly amassed 8,000 fans. Pepsi — those “guys in blue” — raced to catch up, eventually launching a rival machine called Pepsi Spire in 2014.
Over the next nine years, Coca-Cola deployed more than 52,000 Freestyles to customers like Qdoba, Wendy’s, Burger King, Subway, and AMC Theatres. (While beverage consumption at Freestyle restaurants tends to increase about 8 percent, according to Coca-Cola, calories-per-serving goes down, as more consumers choose low-calorie or no calorie options.) It was granted more than 80 patents for the technology inside.
‘How Do We Scale This Thing?’
After the Freestyle had been in market for several years, the team overseeing it realized the need to create Freestyle 2.0. The first device “was about discovery, and proving the model,” says Scott Harrison, Vice President of Platform Leadership at Coca-Cola. “Now, it’s about how do we really scale this thing?”
Stubbs says that there were several components to the business case. The first was security. “We needed to update our software and the way we connect to the machines to keep the bad guys at bay,” he says. “These are wirelessly-connected devices with screens, and our first goal is to not repeat the mistakes of some companies that have been hacked.” The second was about creating a single Freestyle operating system for the entire “fleet” of machines, rather than running different software on the different models of machines.
The third component was maintenance. “We were aiming to be about 90 to 95 percent more efficient than we were,” Stubbs says. “We needed to make [maintaining the machines, including cleaning them and swapping out flavor cartridges] simpler and more efficient.” The team also wanted to enable Freestyle to handle a wider range of beverages, including juices and coffees.
And the final component of the business case was the ability to innovate with customers, creating specific beverages exclusive to their chains.
Rather than trying to add features or enhance the first version of Freestyle, they redesigned it from the ground up. “It was just more efficient, better quality, and less expensive to go with a fundamentally new design,” says Stubbs.
While the original machine ran on Windows, the new one relies on a Linux operating system that supports tech standards like Java and HTML 5. That “allows us to more rapidly innovate,” Stubbs explains, “and it means that other teams [both within Coca-Cola and at its restaurant customers] can build apps for the machine.” The touchscreen display has much higher resolution and can play videos. The “push to pour” button relies on haptic feedback to vibrate a bit, so the consumer knows the beverage is coming.
And, Stubbs adds, “We filled it with electronics that we know we’re going to need in the future, like bluetooth and RFID radios, and we made it more modular for hardware and software expansion in the future.” The Freestyle’s wireless connection allows for its software to be regularly updated over the air — not unlike a Tesla sedan.
Halfway through 2018, the Freestyle team decided to run an alpha test of the redesigned machine — known internally as the 9100. This time, they chose a Firehouse Subs location in Marietta, Georgia. Firehouse had been one of the first Freestyle customers to roll out the machine to all of its locations in 2011. Firehouse was “eager to be first-to-market with the latest and greatest,” Sanders says.
This time around, the alpha test was different. “We just needed a close partner, and someone who wouldn’t mind having three engineers in the restaurant constantly,” Sanders says, and executives dropping by to have a look.
But while the engineers were busy keeping the Freestyle up-and-running during the original Willy’s test, “the second time around, the engineers were bored,” says Stubbs. Sanders adds, “We had to make sure the location had wifi. They weren’t having to deal with the equipment constantly, so they needed to be able to work on their laptops.”
After the alpha test, Coca-Cola typically goes to a larger rollout — anywhere from 20 to 50 locations “where we’re trying to simulate ‘real life,'” Harrison says. “We let the typical [restaurant and Coca-Cola] operations take over. We’re testing the supply chain, the install processes, the training for the crew. Then, fast-forward a few more months, and we start to build inventory and manufacture” the machines, and install them in restaurants around the country.
One important test for the new machines: How should Coca-Cola use the screen’s new video capacity? “Consumers have told us, ‘I’m there to get a drink—don’t show me a commercial,'” Harrison says. “When they’re up in front of it, they want to discover something new, or find something fast. When it’s not pouring a drink, that’s when the video might show up. We’re figuring out as we go.”
The Debut of Freestyle 2.0
The Version 2.0 of Freestyle was unveiled in May 2018 at the National Restaurant Association Show in Chicago.
It has an array of new features, like tracking usage patterns on the machine so that a restaurant manager can better predict when a flavor cartridge will need to be replaced. The original model would report on what percentage of a cartridge was remaining. But “now we can predict that you’ve got ten hours [of use left], or two hours,” Harrison says. “It makes them more efficient and effective in how they maintain the machine.” And the dashboard of information about the Freestyle 2.0 can be accessed using a smartphone, tablet, or PC.
Data about how consumers in various geographies use the Freestyle machines also plays a role in how Coca-Cola makes marketing and distribution decisions. “Last summer, the Sprite team used Freestyle consumption data to understand that Sprite Cherry was a fan favorite,” Harrison says. “So they developed that into a ready-to-drink version you could buy at retail.”
While the first model of Freestyle offered 106 different flavor choices, both that model and Version 2.0 can now serve up about 200.
Consumers can also craft their own custom beverage and store it in the Coca-Cola mobile app; then, on a future visit to that location (or any other Freestyle machine), they can “call up” the same recipe using their phone.
And Coca-Cola has been working with restaurant chains to add their signature drinks to the Freestyle. At Firehouse Subs, that means Cherry Lime-Ade. “They were hand-making it before they had [it on the Freestyle],” says Sanders.
The Freestyle team was at the massive South by Southwest festival in Austin earlier this year, showing off the new machine and offering samples at a park just outside the convention center. As of May 2019, there are about 525 of the Version 2.0 Freestyles out in the wild.
The team has also been collaborating with Coca-Cola’s Powerade brand to use the machine to create customized drinks for individual athletes “with the electrolytes you need,” says Harrison. “We’re testing that with Louisiana State University’s football programs.”
“One of our big lessons learned is that you can’t wait eight years to keep developing and investing in a technology,” says Harrison. “So we’re working on getting to shorter cycles, which will feel more like incremental innovation. We’re thinking about wearables, and facial recognition, and how those [might relate to the Freestyle]. Is it relevant to us and our consumers?”
Harrison says that 52,000 Freestyle machines served up about six billion beverages last year: “That’s a material chunk of our food service business, and it’s growing each and every night.” Freestyle General Manager Chris Hellmann said in mid-2018 that Freestyle had grown into a billion-dollar business for the company.
That makes it one of those rare skunkworks innovation projects that began in the basement, survived three different CEO switches, and eventually had an impact on a mass scale.