Build a MECE issue tree
After the prompt, ask one or two clarifying questions, take a minute, then lay out a structure. Interviewers grade it on being MECE: mutually exclusive (branches don't overlap) and collectively exhaustive (together they cover the question).
Split the question into three or four branches, each a question you could answer with data: "Is the market attractive? Can we win? Can we make money? Can we execute?" Then say which branch you'd test first and why.
Tailor it to the case. A memorized framework recited word for word reads as a script, not thinking.
Worked example
"Should our client, a regional gym chain, open in a new city?" Sort each question into its branch.
Market
- How many adults in the city already pay for a gym?
- Is gym membership in the city growing?
Ability to win
- What do the top three local gyms charge and offer?
- Does anyone there know our brand?
Economics
- What does a new site cost to open?
- How many members until a site breaks even?
Each question sits in exactly one branch, and the three branches together answer "should we enter?". That's MECE.
Quick check
What makes an issue tree MECE?
Its branches don't overlap, and together they cover the whole question. No overlap, no gaps. The number of branches and the framework's name don't matter.