Was Joel aiming at me? Probably not, since his nemesis here is "stunningly-good-looking."

Oh, and if you've ever heard my rant about how metrics drive behavior you'll probably get the sense that I'm definitely not the person he was aiming at. I watch this all the time: driving a complex problem down to a simple set of metrics. While a useful exercise to attach metrics to goals you're trying to achieve (otherwise, how do you know when you're done?), you need to keep your eye on the "shades of gray." Rarely in business do we see a simple case where a single, easily obtained metric accurately reflects success or failure. Great: you have 75% share of that market segment. Of course, you lose $1,000 per customer... Wonderful: you have a $1B business, but your customer satisfaction is in the hole. Excellent that you have great customer satisfaction, wonderful segment share, and high revenue, but you have no management bench so your next generation is doomed.

You absolutely have to pick metrics, but you have to realize two things: there will always be a balance of several that you're weighing off against each other and that they will change over time.

And in the case of measuring developer productivity I have to say it's akin to measuring the productivity of a musician. It's more art than science, much to the chagrin of the people who aren't artists themselves.