Keecker

» see keecker.com as it lived, on the wayback machine

~ the last words ~

A home robot priced at one thousand seven hundred and ninety dollars in 2017 had no path into a flat in Ankara, and Keecker never looked for one. It sold in France and to a thin international audience of people who wanted a projector that drove itself, which is a small market inside a small market.

Pierre Lebeau had been a product manager at Google before he started the company in 2012. The thing he built was about fifteen inches on a side and weighed eighteen pounds. It ran Android and carried a projector with speakers, plus enough navigation to move between rooms on its own. It showed up at CES in January 2014 with a planned price somewhere between four and five thousand dollars. By the time it actually shipped to backers in 2017 the price had come down to $1,790. The distance between those two figures covers almost everything that happened in the three years between them.

It worked, as far as it went. The company counted over a thousand robots in homes and about three and a half hours of use a day, which for a piece of discretionary living room hardware is real engagement rather than shelf decoration. The thing it never managed was a second round of funding. Keecker had raised roughly eight million dollars across its life, and in early 2019 the existing investors declined to continue and no new ones appeared.

~ the record ~

~ the last update ~

On 8 March 2019 Keecker pushed a final over-the-air update to every robot in the field. Its purpose was to remove the robot's dependencies on company servers, so that when those servers went dark the hardware would still boot and still play local films and music. The cloud features went with the company, including the remote video access and the home analytics. The box on your floor kept working.

My own work is inference that has to run on the aircraft rather than in a datacenter, so that changelog held my attention longer than the shutdown notice next to it. Cutting a shipped fleet loose from its backend is not a trivial piece of work, and it is work with no commercial upside whatsoever, done by people who had just been told the company was finished. Four days later, on 12 March, Lebeau announced the end publicly.

~ the 2019 robot cull ~

Keecker died in the middle of a bad eighteen months for consumer robotics. Jibo's servers were going off around the same period, Mayfield Robotics had folded Kuri before most units shipped, and the pattern was consistent enough that Lebeau named it himself when he wrote about the closure. His summary of why was blunt: building a profitable business in consumer robotics is harder than it looks from the outside.

His other line is the one that stuck with me, because it is the thing every hardware team says to itself at the post-mortem and most of them are wrong about. Maybe they were too early, maybe they were too expensive. At $1,790 for a self-driving projector in 2017, set against the price of a projector that stays where you put it, the second of those was not really in doubt.

~ what survived ~

The robots did, which is the unusual part. Most of this category ends with a device becoming a paperweight on a specific announced afternoon, and the families who bought one get a refund policy instead of a product. Keecker's last act was the opposite of bricking: it spent its final engineering days making sure its customers would not need it anymore.

Lebeau went on to co-found an engineering consultancy. The company's own epitaph, posted the week it stopped, was the kind of thing founders rarely write down in public: everything is possible, but today it has to stop.

@}-->--

              )
             (_)
            .-'-.
            |   |
            |   |
            |   |
            |   |
          __|   |__   .-.
       .-'  |   |  `-:   :
      :     `---'     :-'
       `-._       _.-'
           '""""""

« previous  ·  random  ·  next »