Metrics question

How would you measure the success of Facebook Friends feature?

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 your ability to build a metric tree for a mature, low-visibility feature where the naive metric (friend count) is misleading.

State the goal: Friends should deepen real-world relationships and feed content relevance, not just inflate connection counts.
Reject raw friend requests sent as the top metric since it rewards spam-adding strangers.
Build a tree: friend requests sent, accept rate, and post-add interaction (messages, tags, comments between the pair in 30 days).
Add a guardrail metric, requests-to-block or report ratio, to catch abuse like mass-adding.
Pick a north star, share of new friend connections with at least one interaction within 30 days, since it signals a meaningful tie.

Distinguishes vanity metrics (total friends) from quality metrics (interaction after connecting), which is what interviewers look for on this question.
Uses an illustrative number, assume 60% accept rate but only 25% of accepted pairs interact within a month, to argue quality is the gap.
Segments by relationship type, close friends found via mutuals versus cold adds via People You May Know, since their health looks different.
Ties the metric back to News Feed relevance, since interacted-with friends produce content people actually want to see.

Stopping at 'number of friends' as the success metric without questioning what it actually signals.
Ignoring abuse cases like mass friend requests from fake accounts.

How would you detect and penalize spammy mass friend-adding?
What tradeoff exists between showing more People You May Know suggestions and interaction quality?