Farewell Jibo and Kuri

Or why the pursuit of the killer app is fraught with peril

In the past two months, Mayfield Robotics, makers of Kuri the robot, has shut down sales and operations and Jibo, which has run through more than $70M of venture funding (according to CrunchBase), announced a significant downsizing of the company.

This marks a sad time for the personal/social robot market. There were amazingly talented and passionate people at both of those companies who drove themselves constantly in the pursuit of building awesome products that were well-liked. They solved some really difficult problems. Because these excellent teams with experience and knowledge will likely disperse, more difficult problems will either go unsolved or will be delayed without them. We know many of our compatriots at these two businesses and are dispirited by their loss.

What happened? I propose that the bottom line is that both Jibo and Kuri sought to sell a product to mass consumers with the belief that they had found the “killer apps” (or the features that make a product indispensable). In this case, those were social interaction (Jibo) and live streamed video/photos in the home (Kuri). Yes, they both did many other things pretty well — from independent navigation (Kuri), to face and voice recognition (Jibo), to basic home assistant abilities — but the real robotic differentiators for each were those killer apps.

Additionally, both learned that the required consumer sales volume didn’t materialize. For owners (venture and Bosch), the size of their investment and parameters of ownership was likely (experience-based conjecture on my part) predicated on an expectation of selling hundreds of thousands, if not millions, of units. To make Jibo worth >$500M (which is likely what an investment of >$70M would require for venture to get its returns), it had to sell at least 100K units per year. And undoubtedly, Bosch requires its business units to create “meaningful revenue” that is acceptable relative to its other business units (see: Clayton Christianson: The Innovator’s Dilemma).

I haven’t spoken to either CEO about their strategy, nor have I seen any public disclosures about them, but from afar, both companies’ strategies looked like: “build the killer app for direct-to-consumer robots, sell millions of them, and then create a platform upon which others could build more.” I.e. “the iPhone strategy.” In my view, this strategy was doomed to failure from the beginning.

Why? Because technology diffusion doesn’t work this way.

(To read our take on each of six market strategies in play today, see our 10 Year Plan post. For my take on “the killer app conundrum” please read this blog post.)

The study of how innovative technologies diffuse through a society has a long and storied history. More recently, it’s been popularized by Geoffrey Moore’s writings captured in Crossing the Chasm, Inside the Tornado, and other works.

The gist of all of that history and study is this:

There is no magic silver bullet by which one gets to pass Go, collect $200, and sell a platform to millions of consumers out the gate.

Sure, Apple did it with the iPhone. But they had $100B in the bank and the world’s most recognizable brand — and the presence of the 3 core innovations (cellular phone, email, and web browsing) diffused throughout the population for over a decade.

Robots are not cellular phones, email, or web browsing. We consumers haven’t been using them for decades. Some tiny amount of us are using them to perform single tasks, such as vacuuming — but we have no existing penetration or models of social, mobile, autonomous, multi-purpose robots cruising around. They have captured our imaginations for decades, which has had the unfortunate side effect of causing sky-high consumer expectations for what an autonomous robot could do (basically, everything). No company in the world today, next year, or even a few years from now can deliver to these expectations. The day of a robot doing everything a consumer wants is still a decade or more away, because each of the component technologies is still nascent.

Inevitably, a robot that claims to do more than a single task (such as vacuum) comes joltingly face-to-face with the reality of those consumer expectations and that innovation adoption curve. Which means that any consumer-targeted robot has to (i) get the single task it does perfectly executed, (ii) be highly valuable to the consumer, and (iii) still start with early adopters who are willing to take risks with unknown tech.

In other words, it has to have the killer app. Or does it? At least, the robot has to perform some profoundly valuable task for which consumers will gladly part with a significant amount of money. Autonomous robots aren’t cheap to make — practically anything with autonomy, a motor, and processing power is going to have a >$500 retail price.

So, was the “task” of a robot being a better version of Alexa or Google Home worth a cost of 10x either of those two products? Was the “task” of capturing candid photos and videos necessarily worth >$500? In both cases, the market seems to have said no.

Does this mean there was something wrong with Jibo and Kuri as robots? As stated, my answer is definitely not. Both products have been high-quality and important advances in consumer robotics. The problem lies more in the apparent hope that the consumer market for robots could be fully engaged at acceptable sales volumes based on a very limited “task” set. I don’t blame them for trying, but I do think that the probability of any small team of innovators — likely less than 20 who actually created the robot at both Jibo and Mayfield — finding the equivalent of a killer app is infinitesimally small.

Why do we even talk about killer apps? It started with the Apple II and VisiCalc. Once VisiCalc arrived, the Apple II’s growth took a sharp turn upward and became a true consumer success. But Apple didn’t create or find VisiCalc. Based on the Apple II providing a relatively affordable, sufficiently powerful, and easily enough programmed platform… VisiCalc found it.

This is our model at Misty Robotics. It’s why we’re different. We’re leaving the innovation side of skill development to the future creators — our first customers. Instead of first starting with the consumer market, our target customers are enterprise and individual developers, makers, and students. We believe the probability of them creating killer apps (or as we call them, skills) is much higher than if we tried to do so ourselves. We’re creating the platform that gives developers the foundation to build upon, which we believe will lead to greater success in the long run.

We’re also solving a core problem for our market of business experimenters and teacher/student learners. We know, for example, that professional developers lack an affordable way to explore how robots might fit into their businesses. We also know that high school and college students (and their teachers) have no powerful, general-use robotics platform on which to experiment, build, and learn. And we know that makers are seeking an adaptable, highly hardware-configurable robotics platform, as well. None of these prospective customers wants to spend 6–12 months building their own robot nor several months learning the sophisticated ins and outs of existing open-source robots.

Importantly, the expectations of our developer customers are very different from that of the sky-high, sci-fi visions of the consumer robot purchaser. Our customers understand the nature of the rapidly evolving capabilities of general-purpose robots, just like Dan Bricklin and Bob Frankston (the inventors of VisiCalc) understood how to get the best out of the Apple II.

We know our model is not a quick path to success, so we’re planning in enough time with our business staging for both our platform and our developer community to grow. Unfortunately, time is something neither Jibo nor Kuri seemed to have. But once those killer apps do emerge on the Misty platform, we believe consumers will have a high-value proposition to invest in a multi-purpose personal robot.

A version of this article originally appeared on IEEE Spectrum. Read the original article.