My qualifications


As decision scientists, a decent part of our job can be best described as “throw math at parts of the business and see if value comes out.” In general, recruiting is an under-mathed part of the company. Today, I want to talk about 2 of my favorite concepts at the intersection of stats and recruiting: maximizing signal and asymmetric hiring risk.

What an interview isn’t.

A vibe check. In the absence of good principles, any non-technical interview quickly devolves into a vibe check. An unprincipled technical interview takes only marginally longer.

Studies show we end up optimizing for candidates that are similar in style to us. “These traits and thinking styles made me successful. Naturally this would make others successful in this role as well”. Absent conscious effort to avoid this bias, Dmitri would hire people with technical backgrounds and a good taste for fonts.

This style of interview is also much more easily targeted via bullshitting, personality, and pure practice.


What an interview should be (in 2 steps).

It’s actually super simple. For each interview, you just need to do 2 things:

  1. Maximize the signal you get about the candidate.
  2. Use that signal to make a calibrated choice about whether to move the candidate forward.

Let’s unpack those one at a time.

Maximizing Signal

It’s good to walk into every meeting with a purpose. Bias aside, we tend to make a judgment on candidates in the first 5 minutes before proceeding to mechanically run through the rest of the interview. This habit makes sense for how we normally approach the universe. Doing analysis takes work - there’s only so many complex decisions you can make. So you evaluate a few situations from the ground up, and then start looking for patterns. Eventually we’re so tied to our heuristics that you have a hunch about the interview before it even starts. But what’s the point of a 30 minute interview if you only use the first 5 minutes?

I’ll posit that the way to overcome this bias is to set an intermediate goal in the interview. Before you make a 1/2/3/4 decision - take advantage of your time to maximize the signal on the likelihood of a candidate’s success.

Probing to failure