Improving Discovery and Creator Tools on Spotify

✳The problem they went after

Spotify's mission is to unlock the potential of human creativity for artists and listeners, so Samar checks the actual flows against it. Home page discovery, search, and the now-playing screen each get mapped for what works and where friction shows up, across six segments: active and casual music listeners, active and casual podcast listeners, and both music and podcast creators.

✳What they actually did

The friction list is specific. Songs cannot be played straight from the home page, since users have to navigate into a playlist first. Search is elastic, doubling as a discovery page, which confuses anyone arriving with clear search intent. Secondary controls on the now-playing screen are unclear. The teardown also credits what works: jam sessions for collaborative listening, offline downloads for premium, and fans-also-like recommendations. Top hypotheses land on missing per-episode podcast ratings, weak sorting for podcast creators, and new releases buried for followed artists.

✳What you can take from it

The transferable bitA model for structuring a teardown so it produces hypotheses instead of opinions. Strengths are named alongside friction, segments are defined before any problem is picked, and each hypothesis carries a consequence, such as the churn risk attached to missing episode ratings.

✳Everything else they shared

✳The problem they went after

Spotify wins on audio consumption and loses on podcast discovery. People find shows through short clips and social feeds elsewhere, then switch to Spotify only to press play. Research pointed at recommendations biased toward creators the listener already knows rather than topics or moods, and at the removal of the What's New feed, which left routine listeners without a daily anchor.

✳What they actually did

Samar ran five interviews plus a survey, and the findings sharpen the problem into intent mismatch and search friction. People choose podcasts on guest credibility, emotional resonance, and context such as a commute or a workout, not on popularity scores. When search fails to surface something relevant in the top three hits, they abandon the session outright. Length preference splits as well, with long-form for deep learning and bite-sized for news during a workout.

✳What you can take from it

The transferable bitThe research does something specific: it converts vague dissatisfaction into decision criteria. Knowing that listeners pick on guest credibility and context, and that they quit after three bad search results, gives you a design brief. The recommendations follow from that, rebuilding discovery around mood, intent and life stage instead of creator identity.

✳Everything else they shared

✳The problem they went after

New podcast listeners land on a Spotify home page that is cluttered, offers no directional cues, and asks for endless scrolling with nothing to narrow the field. All the work of finding and judging a show falls on the listener, and plenty of them drop off before committing to any podcast at all, which costs both acquisition and session depth.

✳What they actually did

Samar evaluated fourteen solutions across confidence tiers, and two tied at the top with a RICE score of 13.5. Intent and mood oriented categorisation gives a Netflix-style browse that maps content to context quickly. Short snippet discovery sections score the same by promoting behaviour users already have inside the app, letting them preview a podcast in a bite-sized clip before committing to a full episode.

✳What you can take from it

The transferable bitOne solution wins by adding structure, the other by borrowing an existing user behaviour rather than teaching a new one, and the reasoning behind that tie is where the value sits. Together they turn open-ended browsing into a directed journey, which is the actual goal here: reduce the effort of finding and evaluating a podcast, not add features to the home page.

✳Everything else they shared

✳The problem they went after

Spotify's podcast listeners keep discovering shows on YouTube and Instagram. The recommendation engine surfaces the same hosts repeatedly instead of adjacent topics or mood-matched content, and losing the What's New feed took the daily habit away from regular listeners. The app ends up read as a consumption tool while discovery happens on someone else's platform.

✳What they actually did

Two personas carry the research: a casual explorer who wants emotional and topic-based discovery without planning effort, and a knowledge seeker who expects intent-driven, credibility-vetted recommendations that fit his career stage. Both hit the same walls, a cluttered homepage that defaults to music, no chapter navigation inside episodes, and search that misses in the first three results. The proposal answers each one with mood and intent carousels, snippet previews in the style of short-form video, chapter-level navigation, context playlists for commutes and workouts, a revived personalised What's New, and host credentials as trust signals.

✳What you can take from it

The transferable bitWhat holds this together is that every feature traces back to a named friction point, and the two personas differ enough to force different answers. The explorer needs low-effort mood entry points, the knowledge seeker needs credibility signals. Designing for both is why the solution set is broad rather than one hero feature.

✳Everything else they shared