Microsoft has taken a lot of heat for its inability to innovate, but Kinect proved the old dog can still learn some new tricks.The controller-free system for the Xbox 360 received almost universally glowing reviews and shipped 8 million units in the first quarter, making it one of the fastest selling tech product debuts in history — and a new billion-dollar profitable business for Microsoft.
Kinect is also a lesson in how a big company like Microsoft does innovation. Kinect wasn’t created by a single product team working in an isolated lab, as it would have been at a startup. Rather, it was the product of collaboration between many different groups, which combined years of Microsoft research with some critical acquired technology and turned them into a real product.
Here’s how all the pieces came together.
Microsoft hired Don Mattrick away from EA to lead the Xbox business in 2006. Mattrick had been thinking about how to move beyond traditional game controllers ever since the early 1990s -- he recalls a conversation while on the board of directors at USC where a person from the film industry told him that the traditional game controller was the main barrier to making video gaming into a mainstream activity. He also needed to respond to the runaway success of Nintendo's Wii, whose motion sensitive controller had made gaming more interactive.
In 2007, Mattrick presented a challenge to the Xbox team: expand gaming to a whole new set of customers. Ged rid of the controller.
Xbox Director of Incubation Alex Kipman created a small team within Microsoft and began researching the possibilities. He used his home town, Natal, Brazil, as the code name for the project. The team quickly honed in on the toughest challenge: getting a device to track users' bodies as they move.
Computer vision works by sending infrared light into a space, then using the reflections of that light to figure out what the room looks like.
Israeli startup PrimeSense figured out how to encode patterns in the light beams it sent out, then measure the changes in those patterns to give a particularly accurate view of the room. PrimeSense also used a second camera to detect colours within the human range of vision, and figured out how to put its technology into a chip that could sit inside the camera. Microsoft quietly licensed the chip and reference design from PrimeSense in 2008.
Microsoft later bought two other companies with 3D sensing technology, Israeli startup 3DV systems Silicon Valley based Canesta. However, neither of these companies currently provides technology to Kinect, and Microsoft probably bought them mostly for their patents.
The PrimeSense technology was a good start, but Microsoft still found it very hard to get the camera to recognise the human body as it moved through different poses, and to recognise the same person regardless of changes in their appearance -- all in real-time, with no lag.
Andrew Blake and Jamie Shotton at Microsoft Research had both done research into computer vision. Together, they figured out how to write an algorithm that would let the depth sensor recognise and keep track of specific parts of a person's body -- an arm, a shoulder -- without getting confused by rapid movements or changes in appearance.
Getting Kinect to recognise voice commands would have been hard if Microsoft had started from scratch. But the company had been working on speech recognition technology for a decade, and had already built products like Ford Sync, which lets users control their cell phone or MP3 player with voice commands.
Just as Wii Sports made the Wii popular, Microsoft knew it had to have a similar game ready when Kinect launched. Rare Studios, which developed games primarily for Nintendo consoles before Microsoft bought it in 2002, began working on Kinect Sports in summer 2008. It tested more than 20 sports before finally deciding on a handful of 'evergreen' sports that would appeal to everybody, like bowling, boxing, and track and field.
Microsoft also revamped certain Xbox Live services to work with Kinect, so that users could start video chats and select movies to download simply by moving their bodies.
Microsoft unveiled Kinect -- then called Project Natal -- at the E3 show in May 2009. The company knew it wouldn't be out in time for that holiday season, but revealed it early to get gamers and developers excited early.
Over the next 18 months, Microsoft demonstrated Kinect in speeches at other big events. In summer 2010 the company sent a few units on the road in Europe, Canada, and the US, giving real people a chance to play with Kinect -- and tell their friends about it. For the big launch in November 2010, Microsoft reportedly budgeted $500 million for advertising and marketing deals.
Shortly after Kinect launched on November 4th, Microsoft revealed that it was selling about 100,000 units per day, and expected to sell 5 million by the end of the year. In fact, Microsoft shipped 8 million of them to retailers. At $149 a pop, that means Kinect contributed about $1.2 billion in revenue in less than a single quarter. And because each Kinect costs only about $59 to make (according to tear down estimates), that suggests it's a profitable business as well.