Katie Bauer!

There are thousands of Twitter threads on what’s wrong with whiteboard interviews. Objectively they are bad measures because they are high variance—the best talent can easily fail on a lark if they don’t take 2-3 months of extracurricular time to prepare.

Less often but possible is someone with no real world experience studies to the test, hits these it out of the park, and then fails to have much impact in their job because they don’t learn to work with complex technical frameworks and organizational structures.

We don’t ask candidates to show the key skills to succeed in junior roles: navigate large, highly abstract codebases, makes sense of complex data systems, and complicated organizational structures—to ask the right questions and identify the right people to talk to.

In fact, we often don’t even ask candidates to work with data. Sometimes we’ll put a highly structured example where we ask candidates to stream over a data set using Python or write hypothetical code to create a reasonable metric.

I’ve had some success with take home interviews involving making sense of data. You get a good sense of a candidates strengths and weaknesses in exploratory analysis, a sense of their strategic thinking, and a sense of how well they form questions.

Of course that means a bigger chunk of time that not every candidate has, esp if they have kids, particularly so during this pandemic. But I reject the notion that take home interviews are somehow harder on candidates than your typical suite of whiteboard technical interviews.