On Fri, March 28, 2008 2:30 pm, Christopher Smith wrote: > Lan Barnes wrote: >> Managers are (usually) not stupid and many were once coders. >> Unfortunately, they need metrics to drive decision making. >> >> LOC is perhaps the least useful metric possible and can actually be >> destructive. It's alos toe metric with the largest body of accumulated >> data. >> >> This tells me that most accumulated data on code productivity is BS. >> >> How do you compare 20 lines of assembler with 20 lines of Python? >> > The closest to an effective metric for managers came up when I was doing > SCRUM. Part of the process with SCRUM is you do estimates in "units". > Units have no root valuation, but over time, as the team gets better at > estimating, you arrive at a decent sense of what a "unit" means in terms > of the particular team in question. You can then use "units/week" to > measure the relative productivity of folks on said team. Of course, it > is tricky to come up with a way to use this to compare programmers > between teams, and even trickier to avoid people gaming the system if > they know you are measuring their performance this way. :-( > > --Chris
My little brush with SCRUM left me vry impressed, and I'm usually highly skeptical of "go fast" techniques, most of which I discard as developer attempts to return to the irresponsible womb of undregraduate all-nighters at the computer lab. But what I liked about SCRUM was it combined visibility with ownership. It also fostered cooperation. I think measuring productivity becomes damaging as soon as managers invest it with winners and losers, favorites and denizens of the shit list. -- Lan Barnes SCM Analyst Linux Guy Tcl/Tk Enthusiast Biodiesel Brewer -- [email protected] http://www.kernel-panic.org/cgi-bin/mailman/listinfo/kplug-lpsg
