Monday, July 4, 2016

Problematic Software Development Interviews

Interviewing in the tech industry is hard. Both for the interviewer and the interviewee. As an interviewer you've got 45 minutes to 1 hour to determine if someone has the right technical prowess and is a good fit for the company and for your team. As an interviewee you've got 45 minutes to an hour to show off what you know, how well you adapt to change, and how you respond when you don't know something.

In my 16+ year career I've been on my share of bad interview loops. The reason they were bad wasn't because of anything inherent about the person conducting the interview. They were bad because the person conducting the interview didn't know how to flush out what they needed to know about the candidate because they were asking the wrong questions or questions with the wrong focus. Here are four ways I've seen interview questions go down the wrong track. 

The Problem Requires More Than 20 Lines of Code

Writing code takes time. But even more so, solving the problem takes time. You need 3-5 minutes to go over the problem. The candidate then will likely 5 or so minutes to ask clarifying questions and make sure they understand the problem. Then there's another 5 or so minutes of them thinking through the problem and coming up with a solution that they can code.

Ideally, the candidate should be able to write the code for the solution in around 10 minutes. That's 30 seconds per line at 20 lines of code. If your problem requires more than 20 lines of code to solve, you're going to miss out on talking through their code with them. You'll also miss out on being able to scale your question to see how adaptable their code is to change.

The Problem Is Niche And Took You More Than An Hour To Solve

It's very rare that your problem space is so niche that you only want a candidate that has an EXACT set of knowledge in your problem domain. More likely, it's the case that you want to determine if the person knows her/his fundamentals well enough to be able to quickly adapt and learn a new problem space. Because of that, you should avoid niche questions or questions that require you to have super in depth knowledge about the problem space.

A better way to determine if the person is able to adapt and learn is to give them a problem to solve that requires them to critically think through an issue. Once they've satisfactorily solved that problem, change the parameters. You can do this by asking things like "how could you do this with less hosts" or "how could you solve the problem with half the memory usage" or something else that changes an external factor, an input, or an output.

Responding to change is what we do day in and day out as Software Engineers. It's more important to see how well they know their fundamentals and respond to change than it is to determine how deeply they already know your problem space.

The Problem Has Nothing To Do With The Problem Space The Job Is For

Unfortunately, too many times I've seen people ask questions that were not anywhere near the problem space that the job was in. I've seen front end web or mobile developers struggle trying to solve binary tree or problems involving Dijkstra's algorithm when the real problem domain is working with object life-cycle, state management, hardware constraints, and etc.

Make sure that your interview questions are able to tease out whether they will be able to be successful in your problem space. Not whether they know some arbitrary algorithm that isn't applicable or won't help them be successful in your problem space.

The Problem Determines How Clever They Are, Not What Their Skill-set Is

Trick questions, riddles, and brain teasers are not good interview questions. They help determine how clever a person is but not how likely that person is to succeed in your position. There was a time in the early 2000's when you couldn't attend a software development interview loop without being asked a question like "why is a man hole round" or some other riddle.

As an interviewer you want to determine if the person knows the fundamentals of software engineering, can think critically, can adapt to change, and fits in with the culture of your company or team. Your goal is NOT to determine of the person is clever.

No comments:

Post a Comment