> These kinds of metrics inevitably generate perverse incentives.
Precisely. There's a post that goes around about finding good programmers, which ended up concluding that good developers almost inevitably make many commits with small per-commit changes. I had some issues with the methodology (their outlier-exclusion looked like it might force the result), but the outcome made intuitive sense.
I would be horrified if anyone used that metric to judge employee quality. Even assuming it's flawless as a retroactive assessment, it's still trivially gamed. As soon as something that blandly quantitative affects people's work outcomes, it's ruined by the fact that optimizing for the metric is more efficient than optimizing for good work which happens to meet the metric.
Remember though, that the MMPI personality test actually trolls for those gaming the test (diagnosing them with personality disorders, usually.) It does that by tempting people to exaggerate their virtues, for example. So if you're willing to dive into the details of those commits, trumpeting that metric might be a nice honey trap. You don't want the serious gamers and players in your workplace.
Yeah - basically all working real-world systems include "and don't get clever" as one of the rules. If you can't refine your rules to be unbeatable, just put in a ban on beating the rules!
There's an old story about a Soviet pin factory that was judged by number of pins, so it made huge numbers of tiny, useless pins. In response they were switched to a weight metric, so they switched to making giant, 100 pound pins. It's not a true story, because whoever tried it would have gone to prison immediately. You don't actually need to legislate exactly how to behave, because you can just demand good faith and punish people who don't offer it.
I might sound sarcastic there, but it's pretty true. Employee metrics can work on a code of "don't cheat, jerk" as long as the cheating is detectable.
Precisely. There's a post that goes around about finding good programmers, which ended up concluding that good developers almost inevitably make many commits with small per-commit changes. I had some issues with the methodology (their outlier-exclusion looked like it might force the result), but the outcome made intuitive sense.
I would be horrified if anyone used that metric to judge employee quality. Even assuming it's flawless as a retroactive assessment, it's still trivially gamed. As soon as something that blandly quantitative affects people's work outcomes, it's ruined by the fact that optimizing for the metric is more efficient than optimizing for good work which happens to meet the metric.