Discovering where programs are on the productivity-to-team-size curve

Here is former China Lakes and Sidewinder missile inventor Bill McLean speaking with Congress in a 1971 hearing:

Only in listening to the problem of funding of competitive programs, it occurs to me that we really don’t know what a program should cost because there is an optimum size organization beyond which program costs go up without any greater output because it requires more coordination, more people, and more expenses, and we never know where we are on that curve which has an optimum number of people working on a project because at every point on the curve everybody wants more people.

 

I think when we are funding competitive programs it might be a good experiment to try funding them at a 10 to 1 difference in level so that you get some measurement of where you are on the curve of productivity versus number of people; and it is different, I think, for each kind of project that you want to fund so you almost have to do it for every case.

I’ve read that quote several times over the years and I still get different (or more) meaning from it. Funding of projects should be run in an experimental design to reduce the most important risk, which includes costs as well as technology. It is relatively easy to judge whether a technology works, but much harder to know whether it was at the right cost.

Collecting cost accounting data when there is just one monopoly program does not help determine whether it was developed at a minimum cost. Should cost studies are also not suited to the task. Moreover, the program’s no-fail monopoly situation means they can lobby for even greater resources, and policy makers have no external data to push back.

In some classes of systems, like drones, sensors, and software, there can me multiple competitive projects running in parallel. For the biggest platforms where there can be just one, then a hedging project at one-tenth the size provides valuable information.

It is acknowledged that SpaceX developed the Falcon 9 launch vehicle at one-tenth the cost estimated by NASA. If DoD can collect more of these game changers across system types — a one-tenth hedge to NGAD, to GBSD, etc. — then that could really make a huge difference.

Moreover, it creates data that otherwise wouldn’t exist. This vastly improves analytical methods for evaluating current programs and estimating future ones.