In different phrases, Denisov-Blanch’s competition that much less code is a powerful indicator of poor efficiency would possibly sign the other. At least, it doesn’t verify his and the opposite researchers’ finger-pointing at low ranges of Git commits as dispositive proof of builders “ghosting” their employers. Nor does it verify his “don’t-quote-me-on-this” argument that the analysis additionally reveals that “the highest 25% of engineers contributed about 50% to 60% of the output,” although that discovering could also be extra intuitively appropriate, given the 80/20 rule.)
Much less code could imply extra productiveness
Counting code commits, whereas an comprehensible strategy, is flawed. Sure, the strategy is a little more refined than that, however not as a lot because the researchers appear to assume. For instance, Nvidia Senior Engineering Supervisor Aaron Erickson factors out that the researchers would possibly discover “one other 10% of engineers who do add code, but it surely’s ineffective abstractions or vainness rework that provides detrimental worth and confusion.” Stanford’s analysis would say that these are priceless engineers, however in actuality, they is likely to be doing extra hurt than good. Their employers could be higher off in the event that they determined to ghost as a substitute of committing worse-than-useless code. The analysis doesn’t account for unhealthy contributions, by Denisov-Blanch’s admission. They simply count on unhealthy commits are resolved throughout evaluation.
All of this can be a good distance of claiming the analysis could not say what the researchers imagine. This wouldn’t be an enormous deal besides that the headline is clearly meant to drive employers to revisit how they measure engineering productiveness. (Denisov-Blanch says he did the analysis as a result of he believes “software program engineering may benefit from transparency, accountability, and meritocracy and [he] is looking for an answer.”) That’s a fantastic aim, however what about all of the CEOs who might even see the headline and demand that their ghost engineers be fired? Utilizing code commits as the one metric might find yourself eradicating a few of an organization’s high engineers, not essentially their worst ones.
