site banner

Small-Scale Question Sunday for June 2, 2024

Do you have a dumb question that you're kind of embarrassed to ask in the main thread? Is there something you're just not sure about?

This is your opportunity to ask questions. No question too simple or too silly.

Culture war topics are accepted, and proposals for a better intro post are appreciated.

1
Jump in the discussion.

No email address required.

I'm trying to fill a position at work right now. A sys admin role. I want to ask that gas mileage question during technical interviews but I'm afraid the people who will get it right will be so insulted that they can't believe I'm asking this, while the people who get it wrong will feel very unfairly brutally discriminated against because I could pop such an irrelevant-to-their-job question on them.

We use a question like this in our interviewing process. There's much more to it (this is just one component; they then have to design a software system around it), but it requires some basic multiplication.

For the first 2 years, we'd pop it on them live. The results were extremely poor. 90% of people had to pull out a calculator, and many of them would add or remove a 0 if they didn't. I hate to say it, but we had to provide the whole question beforehand to even get through the thing in an hour. I was originally horrified.

Then again, I don't think that solving this sort of complex problem with a timer over your head is empathetic or reasonable. In reality, when I come across a system constraint I have to engineer around, I at least have hours (if not days). My gut feeling was that many people failed because of the pressure and nerve aspect as opposed to the ability to solve the problem in a reasonable timeframe.

In any case, that's a long rant to say: "Feel Empowered to include basic shit like this and use it as a qualifier, but I suggest making it part of a larger question provided to the candidate the day before the interview."

Then again, I don't think that solving this sort of complex problem with a timer over your head is empathetic or reasonable. In reality, when I come across a system constraint I have to engineer around, I at least have hours (if not days). My gut feeling was that many people failed because of the pressure and nerve aspect as opposed to the ability to solve the problem in a reasonable timeframe.

I've been doing hiring off and on for about 20 years so I feel confident offering two quick thoughts.

  1. Have you done mock interviews with the rest of the team to try these questions out? How do your own people do?

  2. Is failing to complete the problem in time that big a deal? In general I find the journey of problem solving more important than reaching the end in the allotted time. You usually know after an hour if someone is too much of a dick to work with, or if they could finish if you gave them another 15 minutes. OTOH, if you and the candidate spend an entire hour struggling with what you thought would take ten minutes, that's probably bad.

  1. The scenario is interesting because it's a real problem we had to solve at the company early on. We haven't had to change it at all for around 5 years because it's such a high-quality signal (and interviewees rate it highly). We did have to modify the approach though - we finished iterating on it after around 1.5 years and probably 30 interviews.
  2. Once the question has been provided a day beforehand, being unable to come up with an adequate solution during the interview is a deal breaker. The good news is twofold: The answer has a correctness gradient, and the whole question is designed to be iterated on as a discusssion. Did you put together a maximally elegant and simple solution? Ok, now make it enterprise-grade. Did you bring your resume-driven development tendencies to the table? Now simplify.

Our entire interview suite uses those approaches, including our small take-home project, so we can easily cross-reference someone's performance with their salary requirements and stated YOE. The tech interview team explicitly defines the tiers of answer performance, and we all collaborate on them.

There have been some calls to introduce a new question to more closely represent the state of the art in modern development. After ~7 total years anything gets long in the tooth, and I'd be lying if I said I'm not a bit concerned about the performance of LLMs in the space. With excellent prompting, they can succeed in our interview if not blow us away. There's no substitute for seeing if someone can verbally describe how they'd solve or problem or change a solution.

Hmm, if it were me, I would be worried that this scenario question would get leaked online. But perhaps your company isn't famous enough that people collect interview questions on forums.

Lol ding ding ding! ~100 people so we are under the radar. Glassdoor doesn't have it on there yet.

But we've interviewed probably 2,000? I think it's still worth developing something, and even if someone knows the answer you can still qualify folks.

I’d personally love to give someone a blank Excel and that mileage problem, if I was interviewing them for a job with Excel use as part of the job description/job ad.

Seriously? My Excel job interview question looks more like this:

  • on worksheet A you have schema and table names and their sizes in the main DB
  • on worksheet B you have schema and table names and their sizes in the backup DB
  • show me all main DB tables that are either missing or have the wrong size
  • bonus question: show me which schemas are completely fine

Seriously. I’m not thinking of a job where Excel will be a primary tool for database work, but admin jobs and other jobs where Excel is an ancillary tool.

  1. Full outer join on table name
  2. Then just null and equality checks until you are there.

Admittedly I can't do this in Excel (because I don't use it), but it's a trival squeal (SQL) or any other tabular tool question

Left join, but ok. The trick is doing this in Excel.