August 25, 2016
Employers: Don't Use GitHub Stats To Judge Developers
I'm seeing a growing trend among employers to use GitHub statistics as a measure of a developer's capability (or, at least, one of them).
I advise employers very strongly against going down this road. There's absolutely no statistical correlation between those stats and the quality of a developer's work, or their ability to "get shit done" under commercial constraints of time and resources.
GitHub - an extremely useful tool for developers to share and collaborate on projects - is also a social network, and like all social networks it can be gamed. In many ways, it's little different to online services like Soundcloud. Would we measure a musician's worth by how many tracks they upload, or how long those tracks are? Indeed, given the tricks that can be deployed to boost a track's apparent popularity, can we even trust the number of track plays and likes as a metric?
Developers who are very active on GitHub are very active on GitHub. That's all we can reasonably conclude. I know many really spiffy devs who don't even have GitHub accounts. And I've stumbled across more than a few who have amazing-looking GitHub profiles who actually suck as developers.
The only thing employers can usefully get from looking at a developer's GitHub profile is the actual code itself. Does it have automated tests? Is it readable? Is it simple? Can you see lots of obvious duplication? Is it basically a C program, but written in Ruby? You can tell a lot about a developer by what their code's like.
GitHub commits are social interactions. They carry information about the committer, just as track Likes on Soundcloud do. And any interaction on a social network that carries information about the sender can, and eventually will, become just another channel for self-promotion, every bit as much as Facebook "likes", LinkedIn endorsements, and Twitter "follows".
So, yes, if you're a developer, stick some of your code up on a site like GitHub so we can take a peek and see what it's like. But if you're an employer, take those stats with a big pinch of salt. If you want to know how good a developer really is, there are much better ways.
Posted 4 years, 1 month ago on August 25, 2016