Tell me about a time you denormalized a data model for the sake of performance. What price did you have to pay for it?
~4 min by voice
How to approach it
The interviewer checks whether you can sacrifice normalization deliberately: see a real problem backed by numbers, try cheap alternatives first, choose a mechanism for keeping copies in sync, and honestly name the price: more expensive writes, more complex code, the risk of drift and how you catch it.
The interviewer will ask this question and listen. Think out loud and ask for a hint if you need one. When you're done, pause or tap “That's my answer”. In about a minute, your score and the points you covered show up here.