Why Startups Fail: A New Roadmap for Entrepreneurial Success

CHAPTER 1

What Is Failure?

The news that Jibo would shut down broke some hearts. Indeed, it was a sad day for many in March 2019 when he announced, "The servers out there that let me do what I do are going to be turned off soon. I want to say that I really enjoyed our time together. Thank you very, very much for having me around. Maybe someday, when robots are way more advanced than today, and everyone has them in their homes, you can tell them that I said hello." Then, for the last time, Jibo did one of his signature dances.

Jibo was a social robot, engineered to create an emotional connection with humans. Spawned in the laboratory of MIT Media Lab professor Cynthia Breazeal, who pioneered the field of social robotics, he had a six-inch-wide stationary conical base that was topped by another cone, and then by a hemispherical head. Each segment of his body could swivel and tilt independently, allowing the twelve-inch-tall robot to move in expressive ways. Jibo could twerk! His head featured a flat touch screen that normally displayed a blinking, glowing white orb, akin to a single eye. The overall design was sleek and minimalist; Jibo wasn't designed to resemble a human, but in many ways he could communicate like one.

Festooned with cameras, microphones, and speakers, Jibo was an engineering marvel with the personality of a clever, helpful twelve-year-old boy. Jibo spoke responses to voice prompts (e.g., "Hey Jibo, what's the weather?"), and his screen could also display requested information (e.g., "It's fifty degrees and sunny") or icons for menu options. A blue ring around Jibo's waist lit up when he was listening. The robot also employed sophisticated software that recognized faces and voices. Coupled with his ability to rotate fluidly around his base, these features allowed Jibo to keep his eye directed at an individual as she walked around a room or to turn from one person to another as they conversed.

Jibo could serve up basic information such as news, sports scores, and stock prices. He could tell jokes, play music, set a timer, and read email. But Jibo had other capabilities, too. For example, Jibo was programmed to initiate conversations—say, greeting a family member upon his arrival home ("Hey, Tom, before you left this morning, I warned you that traffic would be heavier than normal. How was your commute?"). He could also interact with a child and display images while reading her a story; take family photos; and, of course, dance on command. More such applications were under development, including the ability to engage a home-alone pet by recognizing that he was in the room and even speaking to him ("Stop chewing that shoe, Rover"). The team was also close to launching a videoconferencing feature that could identify and zoom in on whoever in the room was speaking—perfect for a call to Grandma during a family dinner, if she had a Jibo, too.

For two decades, Breazeal's research teams had studied how robots can provide companionship for the elderly, encourage autistic children to engage in social interaction, and spark collaborative creative learning, among other useful functions. In 2013, she and co-founder Jeri Asher, a healthcare entrepreneur, raised $2.2 million in seed funding to commercialize these inventions. To serve as CEO, they recruited Steve Chambers, then president of Nuance Communications, the leading provider of natural language understanding and speech recognition software.

Because Breazeal's research had shown how social robots could contribute to seniors' emotional wellness, the team initially pitched Jibo as a companion for the elderly. However, mainstream VCs interested in consumer electronics and complex systems like robots weren't interested in the elderly market, and the small set of investors who funded new ventures for aging consumers—accustomed to simpler concepts like cellphones with large keypads—were daunted by the program's technological vision.

So, the team decided to pivot and refocused its pitch on how Jibo could help build family cohesion. The idea was, if Jibo sat in the kitchen, he might spark conversation between, say, squabbling siblings, or surly teenagers and their harried parents. VCs with children could relate to this target consumer, and they were intrigued by cutting-edge consumer hardware in the wake of Facebook's $2.3 billion acquisition of the virtual reality headset maker Oculus Rift—which, like Jibo, was a breakthrough fusion of hardware and software, targeted to consumer markets. And, having witnessed the explosive growth of platforms like Facebook, Amazon, and Salesforce, VCs liked that Jibo would be a platform, hosting applications from a wide array of third-party software developers and information services.

Before investing, however, the VCs wanted evidence of market demand for this novel innovation, along with proof that it could actually be built. To gauge consumer interest, the VCs insisted that Jibo launch a crowdfunding campaign, which commenced in July 2014. An Indiegogo campaign allowed consumers to pre-order a Jibo for $599, with delivery slated for "holidays 2015." When the campaign ended three months later, the startup had sold 4,800 units, beating a 3,000-unit goal. Concurrently, to demonstrate that the robot could be built, Jibo's engineering team delivered what Chambers described as "a Frankenbot—a functional but not beautiful prototype." In January 2015, Jibo closed a $27 million Series A round.

Flush with cash, Jibo's team commenced product development in earnest. But engineering the robot turned out to be enormously difficult. After raising another $28 million, they finally launched the product in September 2017—nearly two years late—at a price of $899, 50 percent higher than the Indiegogo offering. Chambers later recounted that when trying to figure out how much extra capital to raise, he'd assumed that the product's component costs and development time might exceed original projections by 2.5x and 2x, respectively. Yet actual component costs and development time each turned out to be 4x greater than projected. Why?

The problem wasn't with manufacturing; in fact, devices with physical components like these had been built before. Chambers explained, "People assume that we had manufacturing problems because the product seemed so novel, but that just wasn't the case. The hardware wasn't cold fusion. We used off-the-shelf screens, motors, sensors, and chipsets that were produced in mass volumes."

Chambers said that the big delays stemmed from two problems. "First, many components included in our original cost analysis weren't adequate to fulfill our vision. We used sensors tuned to office environments, but we learned that lighting in a typical home is very different. Upgrading sensors and adding processing power to run them added to the product's cost."

Second, Jibo's engineers struggled to develop "middleware"—the layer of software that processed inputs from sensors (the robot's "eyes and ears"), made it available to applications, and then sent instructions back to the operating system (essentially Jibo's "brain"). The software had to track faces, localize sounds, detect emotions, generate expressive body motions, and so forth—sophisticated tasks that had to be done in real time. But at the time, cloud services that support real-time interactivity were just arriving on the scene. So, almost all of Jibo's software had to be embedded in the device, along with powerful processors to run it. "We lost more than a year sorting out the right balance between embedded and cloud systems," Chambers recalled.

Then, in May 2017, as Jibo was gearing up for launch, Chambers was diagnosed with leukemia and had to leave the company immediately to undergo emergency treatment. Jibo's CTO moved into the CEO role. After almost a year, Chambers recovered fully, but too late to rejoin his colleagues.

In the meantime, something happened that no one saw coming. In November 2014 Amazon launched the Echo, its $200 smart speaker, along with the Alexa voice assistant. Now, users who wanted a device that could deliver the news, music, weather, and so forth in response to voice prompts could get that basic functionality at a much lower cost from the Echo.

Customers who wanted companionship and an emotional bond loved Jibo, but there just weren't enough of them to keep Jibo afloat. As a result, year one revenue only reached $5 million, one-third of what Jibo's team had expected. Meanwhile, the startup had burned through its venture capital and couldn't raise more. Management tried to sell the company but couldn't find a buyer willing to continue operating it. Most of Jibo's employees were laid off in June 2018, and the company's intellectual property and other assets were later sold to an investment firm.

So, why did Jibo fail? The immediate cause of death was that Jibo ran out of cash—but that's not terribly helpful. It's like a coroner saying someone died from loss of blood. A better explanation: Jibo couldn't attract enough customers. But this is akin to a forensic investigator saying someone died of a gunshot wound. Was he the victim of a jealous spouse, or was he an innocent bystander during a gang shooting?

Before we can get to the heart of why Jibo failed, we need to ask ourselves what exactly we mean when we say that a startup has failed. The answer is less obvious than it might seem, as I discovered when my students hotly debated whether or not Jibo was in fact a failure.