Measuring Success: Spotify Playlist Sharing

Spotify launched a new playlist sharing feature that lets users send curated playlists to other users, either in-app or via external channels (iMessage, Instagram Stories, etc.). As a Product Analyst, I was asked: how do we know if this feature is actually working?

This case study walks through the measurement framework I'd build, a mock rollout dataset, and the resulting analysis and recommendation.

• → Increase user interaction and engagement
• → Drive music discovery
• → Strengthen retention and community

"Share-to-Listen Rate — the % of shared playlists where the recipient streams at least 30 seconds of the playlist within 7 days of receiving it."

With ten-plus candidate metrics on the table, the first job is picking one number the whole team optimizes for — everything else becomes supporting or diagnostic.

Why this one over 'playlists shared' or 'unique sharers': those measure sending behavior, which is easy to inflate (spam, one-off curiosity) and doesn't prove the feature delivers on its actual goal — getting music in front of new ears. Share-to-Listen Rate captures the full loop: someone shared something, and someone else valued it enough to listen.

Every North Star needs diagnostic metrics underneath it so a single number never hides what's actually happening.

Growth metrics alone can hide harm. I'd track these alongside the framework above:

• → Spam / report rate on shared playlists — protects against abuse of the share mechanism
• → Unfollow / mute rate after receiving a share — signals unwanted or low-quality shares
• → Opt-out rate from share notifications — signals annoyance, not just disinterest

Rather than launching to 100% of users and eyeballing metrics before/after, I'd structure this as a phased rollout with a held-out control group:

• → Weeks 1–2: 5% rollout — validate the feature works, watch for bugs/spam, sanity-check event logging
• → Weeks 3–4: 25% rollout — start reading early adoption trends by segment
• → Week 5 onward: 100% rollout to treatment, with a permanent 5% control group withheld for at least one full quarter, so retention comparisons stay causally valid rather than just comparing 'engaged users who chose to share' against everyone else

The data below is synthetic — generated to demonstrate what the analysis and readouts would look like, not real Spotify numbers.

This is the view that matters more than raw share counts.

Three cohorts, all measured from first exposure, inside a randomised A/B test.

Useful for prioritising: mood- and genre-based playlists are the highest-volume input if Discovery / Recommendation teams want to leverage shared-playlist signal.

Based on this (mock) analysis, I'd recommend:

• → Ship to 100%, keep a small permanent holdout for ongoing causal reads on retention.
• → Prioritise the open → listen funnel step (68% → 41%) — likely the single highest-leverage fix, since it's the largest drop-off after initial send.
• → Feed high-engagement shared playlists into Discover Weekly / Discovery Mix — the mood/genre segment data suggests a natural connection to existing recommendation surfaces.
• → Watch guardrails monthly, not just at launch — spam and opt-out rates tend to creep up after initial excitement fades.

• → Real data would let me check statistical significance, not just visual separation between cohorts — with mock data I can't compute a valid p-value or confidence interval.
• → I'd want novelty-effect checks: re-measure the same cohorts at day 60 and day 90 to see if the retention lift holds or fades.
• → I'd segment retention lift by user tenure (new vs. established) — sharing might matter far more for onboarding new users than for already-loyal power users.