Please, stop quoting the Grant/Sackman experiment, it's invalid
I'm not quite sure whether Paul Graham refers to the Grant/Sackman experiment in his latest essay, but I think so. Grant and Sackman reported in late 60s about a 28:1 ratio in interpersonal differences in programming productivity. But that number is just plain wrong. For details, see Lutz Prechelt's account on the subject [pdf].
I'm not saying that Graham's misunderstanding disqualifies the essay as such, but it's just so depressing to see people quoting the Grant/Sackman results without knowing the whole story.
(Update: and, for the record, I do realize that Graham's talking specifically about great hackers.)
[permalink] [2 comments] 29.07.2004, 09:42
- Comments:
Posted by Ian Bicking at 29.07.2004, 18:02
I've heard also that particularly junior (or just poor) programmers added to a project can have a negative effect on the project (and this isn't even isolated to programming projects). Which is to say, zero or negative productivity. Which leads to productivity ratios as high as you care to speculate, even when compared to merely average programmers.
At the same time, there are projects which are simply beyond the capabilities of many programmers, which again means there is no ratio of productivity. To the degree you factor such challenging projects into your metrics, your ratio can increase arbitrarily. Also, programmers create their own foundations, and the quality of those foundations can create geometric increases in productivity when layered and accumulated.
Because of this, it's just something of a truism that programmer productivity varies considerably. You can put a number on it, but you're always just going to be pulling the numbers out of your butt.