[jpadkins]

I do a lot of PM phone screens. I don't ask brain teasers but I do ask basic math problems (i.e. drake's equation type problems). They are very good for screening out candidates who won't be able to do the analytical part of the job.

[mdorazio]

1) That's not a "basic math problem", it's exactly the kind of brain teaser/gross estimation problem this thread is saying doesn't work. 2) You're going to need to explain to me how estimating the number of communicative extraterrestrial civilizations in the galaxy (or a similar question) is an indicator of how well a person can understand a product's features, tech stack, and customer needs. Unless your product is something related to statistics, you're just screening out people who aren't like you.

[jpadkins]

One variant I ask is 'estimate the number of delivery drivers needed to start a same day delivery service in a city'. Do you think this is a brain teaser or not? This is clearly not a statistics product...

I find PMs need the ability to break down problems into manageable pieces to calculate, and estimate things with lots of ambiguity. How else do figure out whether an idea is viable or not, and worth spending more time on going deeper on?

And I meant to say fermi problems, not drake's equation.

[mdorazio]

Yes, that's an interview brain teaser of the exact variety Google and others have found doesn't indicate much in the way of actual job capability/fit. As someone else in the thread pointed out, anyone can look up the methodology for answering these types of questions in about a minute, and be able to tackle all of the variants easily with little actual thinking.

I agree that product managers need the ability to break down big problems (or requirements/client requests/company direction) into manageable pieces for their team to work on. However, there are better ways to gauge this ability, such as...

- Ask directly for an example of how they have taken a big problem or assignment and broken it down in the past

- Give them a real-world big problem you've had recently and ask them how they would break it up into team tasks/user stories/whatever you use

- Ask them to explain their methodology for determining how long something will take to build or test (ex. some people love planning poker, other people love logarithmic buckets, other people base estimates on experience, and others do something completely different)

Basically you'll get a lot more value out of interviews if you ask directly for experience or process they'll be using rather than trying to back into a skill estimate using a brain game.

[kybernetikos]

> They are very good for screening out candidates who won't be able to do the analytical part of the job.

It's very unlikely that you have any way of knowing that, since you don't know how good people who fail would have been at doing the analytical part of the job.