Live coding interviews don't work

A silhouette of a curly haired figure, writing equations on a white board
Don't do this. Photo by ThisIsEngineering from Pexels

If someone sent you to this page, it's probably because they are urging you not to put them through this capricious interview practice. Or they're explaining why they've already refused. I hope it's the latter.

Just don't do them. It's that easy.

There's plentiful research explaining why and how live coding challenges fail to serve any useful purpose. Live coding challenges are high pressure social performances that strip a candidate of all the tools, context, and support they would rely on to do real work. People's performance is a product of their environment, and live coding challenges are just about as unlike a real software development environment as it's possible to get. Doing these challenges on a whiteboard is worse than other methods, but screen sharing and pair programming tools have the same problems.

Researchers at North Carolina State University in 2020 clearly demonstrated that live white board interviews don't work. In fact, the actual effect of these challenges is to eliminate a great deal of capable and qualified candidates.

[...] the software industry has seemingly reinvented a crude yet effective instrument for reliably introducing stress in subjects, which typically manifests as performance anxiety. A technical interview has an uncanny resemblance to the trier social stress test, a procedure used for decades by psychologists and is the best known “gold standard” procedure for the sole purpose of reliably inducing stress.

Behroozi, Shirolkar, Barik, Parnin

Notably, those researchers found that zero of the women in the test passed the challenge when they were required to perform live for an audience. Every woman tested passed when they were allowed to work privately.