Estimation question

Determine the right number of photos to show in Facebook's newsfeed.

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

Tests structured estimation reasoning for a ranking and content-mix decision, balancing engagement data with content diversity.

Frame this as an optimization question, not a fixed number: the right photo count depends on maximizing engagement without crowding out other valuable content types.
Identify the tradeoff: more photos per session may raise engagement, since photo content typically drives strong interaction, but too many risks feed monotony and reduced diversity.
Propose testing a range, for example varying photo density from 20% to 50% of feed items, via an A/B experiment rather than picking an arbitrary fixed count.
Measure both primary and guardrail metrics: session time and engagement rate as primary, content-type diversity and creator-type representation as guardrails.
State a directional estimate, informed by these tradeoffs, that photos likely make up somewhere around a third of feed items, pending real experimental data to confirm.

Reframes the question correctly as an experimental optimization rather than a number to be guessed outright.
Names the real tradeoff, engagement versus content diversity, that determines where the right number actually lies.
Proposes a concrete testing range instead of an arbitrary single guessed number with no path to validate it.

Giving a single confident number with no reasoning or proposed way to validate it.
Ignoring the diversity and monotony risk of over-indexing the feed on one content type.

How would you personalize this ratio per user instead of using one global number?
How would you detect if photo density is being gamed by low-quality content creators?