Product design question

How would you design a recommendation engine page for Myntra (fashion e-commerce) for its Product Detail Page for a newly onboarded customer?

Practice this question out loud. An AI interviewer asks it, follows up like a real interviewer would, and scores your answer. Type or speak.

What this question tests

Cold-start recommendation design: solving personalization for a user with no purchase history, a common hard product design problem.

Identify the challenge explicitly: a newly onboarded customer has no purchase history, so standard collaborative filtering recommendations will be poor or empty.
Identify available signals despite the cold start: the specific product page they are viewing, any onboarding style quiz answers, device type, and broad demographic or location signals.
Design the recommendation logic: lean heavily on item-to-item similarity (visually and category similar products to the one being viewed) rather than user history based collaborative filtering.
Add a lightweight signal capture: a quick style or size preference prompt at signup that seeds early personalization without requiring purchase history.
Prioritize item similarity and the style quiz over pure popularity-based fallback, since popularity alone ignores the user's actual expressed interest on this page.
Define success as click-through rate on recommended items for new customers, benchmarked against the recommendation performance for returning customers.

Names the cold-start problem explicitly and correctly reasons that user-history based collaborative filtering will not work for this segment.
Proposes item-to-item similarity as the practical fallback, a standard, well grounded technique for cold-start recommendation problems.
Benchmarks new customer recommendation performance against returning customers, giving a clear, fair way to judge success despite the harder starting conditions.

Proposing a recommendation approach that implicitly assumes purchase history exists, ignoring the stated cold-start constraint.
Falling back only to generic bestsellers, which ignores the specific product context the new customer is actually viewing.

How would recommendations improve as this customer accumulates more browsing and purchase history?
How would you balance a style quiz's added signup friction against its personalization benefit?