I saw Jeff Atwood’s post from a few weeks ago on iterating. One part really struck an chord with me:
Boyd's Law of Iteration.
Boyd decided that the primary determinant to winning dogfights was not observing, orienting, planning, or acting better. The primary determinant to winning dogfights was observing, orienting, planning, and acting faster. In other words, how quickly one could iterate. Speed of iteration, Boyd suggested, beats quality of iteration.
USAF Col John Boyd developed the concept of the OODA loop which is a widely used framework to drive operations (both in the military, but also in business).
When I think back to the many projects I’ve worked on, often we started out with little understanding, including of the scale of the task before us, but the most successful projects depended on the team’s commitment to rapid iteration. It is easy to say, but really hard to do well.
I was recently at a dinner with a large number of very bright government policy researchers and practitioners who were discussing cyber security policy. It struck me how hard it is currently for governments to rapidly iterate on policy. Even in cases where the cost of being wrong isn’t an existential threat to the nation, there is a very high cost, if only in political capital. I started to think what policy would look like where it was explicitly acknowledged that the means was unknown, but the ends were measureable. It is really hard for large institutions (government or otherwise) to invest in projects with similar characteristics mostly because they know they will have to live with the decision with little option to adjust course for the foreseeable future.
When I think about how to get large institutions to more rapidly iterate (also as a degenerate case), I see at least one important conclusion: you must empower the people at the edge to run as much of the loop as possible, understanding that many mistakes will be made along the way. Boyd was focused on dogfights – mano e mano. As you scale up the process to N people, you need consensus at each stage of O-O-D-A to get to the next, and the time involved in not a linear function (anecdotally quadratic, but can be optimized with some clever protocols). That has huge implications for out-iterating the other party since even modest improvements in pushing decision making power down yield huge cycle-time returns… and this very powerfully checks out with experience.
… now you just need to make sure you have the very best people at the edge and give them all the support in the world.