Ambiguity

you moved without enough information

What is scored: can you make progress when the problem is under-defined, without either freezing or charging off in an arbitrary direction?

This competency is weighted heavily for senior roles, because seniority largely is the ability to operate without a spec.

What the interviewer is looking for

Genuine ambiguity. Not "the ticket was vague and I asked a question". Nobody knew the answer, including the people who would normally decide.
A method for reducing it. Talking to users, building a prototype, looking at data, shipping a small version to learn. Something that converts unknowns into knowns.
A decision made under residual uncertainty, with the assumption stated. You will never get to certainty; the question is whether you can act anyway.
Reversibility considered. Did you notice which decisions were cheap to undo and which were not? Moving fast on the reversible ones and slowly on the rest is the specific judgement being assessed.
You revisited it when more information arrived, rather than defending the original call.

Weak and strong, same prompt

Prompt: tell me about working on something poorly defined.

Weak.

"The requirements for the notifications feature were unclear, so I set up a meeting with the product manager and got clarification on what they wanted. Then I wrote a spec, got it reviewed, and built it."

This is not ambiguity, it is a communication step. Someone knew the answer and the candidate asked them. It is fine behaviour and it demonstrates nothing about operating without information.

Strong.

"We were asked to 'reduce support load from billing questions'. That was the whole brief. Nobody knew which questions or how many, because support tickets were free text and had never been categorised. I could have built a categoriser, which would have taken a month. Instead I read 200 tickets by hand over two days and tallied them. Roughly 60% were one question — customers could not tell what a specific line item on the invoice was for, because we labelled it with an internal product code. That reframed the whole thing. It was not a support tooling problem, it was a labelling bug in the invoice. I was uncertain about two things and said so: my 200 were a convenience sample from one month, and I did not know whether fixing the label would move the number or just shift the confusion elsewhere. So I shipped the label change alone, in a week, before building anything else, specifically so we could measure it. Billing tickets dropped by about half. What I would do differently: I should have sampled across three months rather than one — a colleague pointed out afterwards that we had a billing change in the month I sampled, so my sample may have been unrepresentative. It happened to hold up, but I got a bit lucky there."

Real ambiguity, a cheap method to reduce it, assumptions named, a reversible small step chosen deliberately to get information, and an honest reflection that includes luck. Acknowledging luck is rare and it reads as unusually calibrated.

Question bank

Tell me about a time you worked on something poorly defined.
Describe a decision you made without all the information you wanted.
How do you start on a problem nobody has solved before?
Tell me about a time you had to choose between several reasonable approaches.
Describe a project where the goal changed partway through.
Tell me about a time you had to make a call and defend it later.
When have you had to say "I do not know" to a stakeholder?

Where these answers go wrong

It was not ambiguous. Someone knew and you asked them. That is a good habit and the wrong story.

Analysis paralysis presented as rigour. Spending six weeks on a decision document for a reversible choice is a negative signal, and candidates sometimes present it proudly. The judgement being assessed is matching the depth of analysis to the cost of being wrong.

No stated assumptions. If you decided under uncertainty and never said what you were assuming, nobody — including you — could tell later whether the assumption failed or the plan did.

Never revisited. The information arrives and you do not check the decision against it. Say what you looked at afterwards.

Manufactured certainty. Presenting a guess as though it had been analysis. Interviewers probe for the basis, and a story where you were confident without grounds is worse than one where you were explicitly uncertain.

How to say it in the interview

"Nobody could tell me what the requirement was, because nobody knew yet. So I picked the smallest version we could ship in two weeks and used it to find out."

Checklist

Nobody knew the answer, including the people who normally decide
You used a concrete method to reduce the unknowns cheaply
You stated your assumptions explicitly at the time
You matched depth of analysis to the cost of being wrong
You identified which decisions were reversible and moved fast on those
You revisited the decision when information arrived
You acknowledged where you were uncertain or lucky

Practice this interactively