Cicada companies

Every now and then I think about the weirdness of big companies. The number of developers at Google is likely somewhere between 20,000 and 30,000. That is a mindbogglingly large number. To put it another way, 5 ex-Google developers would make a pretty decent software startup, and you could make 4,000 such startups. Whatever Google's doing, it's hard to believe that it has the same impact as 4,000 startups. On the other hand, maybe there's certain problems you can only approach with tens of thousands of engineers.

It strikes me that part of the problem with big companies is that they don't have the same volatility as small companies. Startups appear and disappear all the time, which gives them a certain flexibility, and a certain hunger for success that slows down as they gets larger. That may be a good thing when you have older products to support; the last thing you want is volatility when you're paying for some business-critical enterprise software. However, for new projects or fast-moving sectors, you need innovation and efficiency, and big companies rapidly lose that young and scrappy edge.

Many companies try building internal innovation departments, to a certain degree of success. But I think what's really needed is a way to build some degree of attrition into the company. Cruft slowly accumulates over years of operation, in the form of bad practices, bad projects, and bad people. Often the reason these things stick around isn't because anyone wants to keep them, it's because nobody wants to get rid of them badly enough. An attritional company would be one where the default is that all these things slowly disappear unless they repeatedly justify themselves.

A radical way to implement this would be what I'm calling a cicada company: a company that sheds its skin periodically, say, once a year. So 2012-Google works on all the projects launched in 2012. At the end of the year, you launch 2013-Google. 2013-Google operates new projects or projects specifically carried forward from a previous year. The only way that happens is by mutual agreement, and the different years don't necessarily share management. In fact, rapidly promoting (and testing) management would be another advantage of the cicada company.

2012-Google wouldn't disappear entirely. Instead, it would continue to run whatever projects it didn't want to give up, or that 2013-Google didn't want. Over time, it would get older and stodgier, and slowly make the transition from hot new 2012-Google to reliable, legacy 2012-Google. Any projects that stayed with that company would also slow down. Eventually, 2012-Google would wind down entirely, or maybe just bumble along indefinitely, as is often the way with old enterprise companies. Meanwhile, 2013-Google is where all the projects and people looking for fast-paced work go.

If this worked, it would mean a constantly-evolving CURRENTYEAR-Google, always working on the cutting edge and honed by endless attrition down to just the best and most innovative projects. Anything else gets left behind to stabilise and, eventually, stagnate.