The Teleportation Device Test (Finding the Important Question)
Cap a candidate to two questions on a fictional problem to reveal whether they can find the decision-driving question.
- Difficulty
- Moderate
- Time to result
- ~days to results
- Steps
- 6
- Confidence
- 95%
A low-stakes, fully fictional case interview that tests one skill: can this person find the question whose answer actually changes the plan? The candidate is handed an absurd product (a teleportation device) and asked to bring it to market. After they fan out into unlimited clarifying questions, the interviewer suddenly caps them at two, forcing a ruthless prioritisation of information value. The quality signal is not the plan itself but whether each question maps to a branch in a real decision tree.
Origin
Shared by Shishir Mehrotra, CEO and co-founder of Coda (formerly YouTube product lead), on Lenny Rachitsky's podcast. It is his personal go-to interview question; Mehrotra notes he also uses the same 'what's the important question here?' habit with his own kids on everyday observations (e.g. why three gas stations sit on one corner).
Core principles
- 01The scarce skill is not answering questions, it is knowing which question to ask.
- 02A good question is one whose answer forks the plan; if any answer leads to the same action, the question is worthless.
- 03Fiction beats a real case study — with no domain knowledge to hide behind, only the method shows.
- 04Constraint forces prioritisation: unlimited questions reveal nothing, two questions reveal everything.
- 05Low stakes accelerate learning of the method, which is why children often outperform adults at it.
How to run it
- 1
Set up an absurd, unresearchable product
Tell the candidate a group of scientists have invented a teleportation device and hired them to bring it to market. Give no specs at all. The fictional nature is the point: no candidate can have prepped, googled, or memorised a framework for it.
Pro tip Any invented product works. The test is portable — the constraint, not the scenario, does the work.
- 2
Let them fan out unconstrained
Say nothing while they ask whatever they want: how big is it, how fast, is it safe, does it need a sender and receiver, does it disintegrate things. Let the full spray of questions come out and simply observe.
Pro tip Do not coach or narrow them here. The unconstrained list is the baseline you will later contrast against their constrained list.
- 3
Spring the two-question constraint
Interrupt: the scientists hate talking to people, are annoyed by the questions, and will answer only two of them — after which they expect a plan. Ask: which two questions do you ask?
Pro tip Deliver it as a plot twist, not a gotcha. The surprise is what prevents a rehearsed answer.
Watch out There is no single right pair of questions. Grade the reasoning, not a memorised 'correct' answer key.
- 4
Interrogate the decision each question unlocks
For every question they choose, ask what decision its answer would let them make. If they say 'I'd ask what size it is,' push: what would you do differently if it were large versus small? Weak candidates cannot answer; strong ones map each answer to a distinct strategy.
Pro tip This 'what decision does that let you make?' follow-up is the whole grading rubric in one sentence.
Watch out Candidates who cannot explain the decision behind their question are asking for comfort, not information.
- 5
Look for a spanning decision matrix
The best answers pick two orthogonal questions that generate a quadrant of genuinely different strategies. Mehrotra's favourite: (1) Is it safe enough for humans? (2) Is it more expensive to buy (capex) or to run (opex)? Those two axes produce four distinct go-to-market plans.
Pro tip Note the compression — 'is it safe enough for humans, yes or no' beats 'what is the reliability rate.' Strong candidates ask for the smallest fact that flips the biggest decision.
- 6
Practise the skill outside the interview room
Train the muscle on everyday observations: why are there three gas stations on the same corner? The important question is 'how do you place a gas station?' Any product, any oddity, has an important question hiding behind it.
Pro tip Doing this daily makes you a better grader — you can only score the skill reliably if you practise it yourself.
In the wild
One product manager, capped at two questions, asked: is it safe enough for humans, and is it more expensive to buy or to run? He then built four strategies from the quadrants. Safe plus cheap to buy but expensive to run means you blanket them everywhere like human fax machines and charge per use. Safe plus expensive to buy but cheap to run means you place them strategically and effectively replace airports. Not safe for humans opens an entirely different class of use cases around high-value goods currently transported at great cost.
→ Two questions produced four coherent, mutually exclusive go-to-market plans — the clearest possible demonstration that he was selecting for decision leverage rather than curiosity.
Candidates routinely burn a question on 'what size is it?' Mehrotra pushes back: why would you ask that, what decision does knowing the size let you make? Sometimes they can construct an answer; often they cannot.
→ The candidates who cannot connect the question to a downstream decision do not get hired — the follow-up cleanly separates pattern-matching from actual reasoning.
Mehrotra observes that if you give a kid the same teleportation prompt and the two-question cap, almost every child quickly lands on two decent important questions. Adults over-elaborate; kids simplify.
→ He treats this as evidence that important-question-finding is a skill we train out of ourselves, and that keeping the scenario low-stakes and fictional is what lets it resurface.
Common mistakes
Grading the plan instead of the questions
The plan is downstream noise. The signal lives entirely in which two questions were chosen and whether each maps to a distinct decision branch. Interviewers who score the go-to-market strategy end up rewarding fluent presenters over sharp thinkers.
Skipping the unconstrained phase
If you impose the two-question limit immediately, you lose the contrast between what they instinctively want to know and what they prioritise under scarcity. The gap between those two lists is the most revealing part of the exercise.
Using a real product from your own domain
A real case lets a well-prepped candidate substitute domain knowledge and rehearsed frameworks for reasoning. The fictional, unresearchable scenario is what makes the test unfakeable.
Expecting a correct answer
There is no right pair. Interviewers who hold a hidden answer key will penalise a valid but unfamiliar decision matrix and reward candidates who happen to guess their preferred axes.
Is it for you?
Best for
Hiring managers and founders interviewing product managers, engineers, or strategists where the core job is deciding what to find out next under ambiguity
Not ideal for
Roles where execution against a defined spec matters more than problem framing, or interviews where you need to assess deep, verifiable domain expertise
From the transcript
“they will answer only two of your questions and after that they expect a plan what two questions do you ask”
“why would you ask them what size what what what decision is that gonna allow you to make”
“I find that people's ability to learn the method is significantly higher if it's low stakes”
“you can almost take anything and say what is the question that really drives”
From the episode
My favorite interview questions from 100+ guests