Measuring performance doesn't work, Joel says: http://www.joelonsoftware.com/oldnews/pages/July2002.html

I think it's both Joe Krause and Max Levin that both quote the HP founder that, "You make what you measure"

The problem is that sometimes it's not obvious what you measure, since what you might be interested is something vague like "effectiveness" or "progress". And, like all pattern classification problems, if you pick features that don't give a good signal to noise ratio, your assessment could be way off.

Even worse when that being measured knows what's being measured. A feedback loop, if you will. And thus, the fact that you observed something changes it, which sways the results from if you didn't measure it. Sounds a little bit like photons in physics, but I'm guessing photons work off of different principles.

Joel is merely saying that you should be careful of what you measure, because some metrics (lines of code, hours work) aren't indicative of what you're really trying to figure out (progress, effectiveness).

when I say photons, I mean, using photos to measure the position of a particle.

Joel is basing his argument off the book "Measuring and Managing Performance in Organizations." In points out that in situations where the employee gains natural satisfaction from serving the customer, measurement often makes performance worse. The additional effort spurred by the measurement is offset by the distortion to the effort based on what is being measured. However, in the case of doctors, measurement is already taking place, and because it is highly distorted. Thus switching the measurement from number of procedures to increase in health would be a substantial improvement. It may, however, be even more of an improvement to pay a flat fee, regardless of the number of procedures.

Joel talks about software development, where the real impact of individual contributions is hard to measure. Any metric you set will be inaccurate, and thus harmful, because people will be motivated to maximize the metric, rather than real value. You just have to find people you can trust and empower, then trust them and empower them.

Now, healthy days is quite a good heuristic to evaluate a doctor's performance, don't you think? How do you think it can be gamed; i.e., what exploits can you think of?