Fixing Onboarding Friction and Interface Clutter on Spotify
A listener's-eye teardown of Spotify: what works, what grates, and which frictions threaten revenue
Phase 01 · Product teardown
Fixing Onboarding Friction and Interface Clutter on Spotify
✳The problem they went after
Spotify does a lot right, and the friction sits in odd corners. The homepage stacks overlapping sections until it is cluttered, podcasts surface inconsistently, and the subscription flow pushes users out of the app for payment and FAQs. Shereen also flags a strategic wrinkle: Sound Capsule may be quietly cannibalising Wrapped, the platform's flagship annual moment.
✳What they actually did
Strengths are named as precisely as the problems: playback continuity, recommendations built on long listening history, collaborative playlists and Jams for social listening, and Wrapped as a viral event. Three features are reverse-engineered for why they exist at all, Wrapped for nostalgia and shareability, pre-save countdown pages for release discovery, and collaborative playlists for social stickiness. Segments cover casual listeners, engaged fans, podcast-first users and social sharers, and the write-up lands on three problem hypotheses: playlist migration barriers, the broken in-app subscription flow, and interface redundancy.
✳What you can take from it
Worth stealing
- Every friction tied to conversion or retention risk
- Flags Sound Capsule as a threat to Wrapped
- Three features reverse-engineered for why they exist
- Off-app subscription flow named as revenue risk
Read this if
Doing a teardown and want every friction on the list to earn its place
The transferable bitEvery friction point is attached to a conversion or retention risk. That single rule turns what could be a list of personal annoyances into a prioritised set of hypotheses, and it makes the subscription flow, which sends paying users off-platform, impossible to dismiss as cosmetic.
✳The submission itself
Open the PDF →Reading on a phone? The embedded viewer is cramped — open the PDF in a new tab instead.
Phase 02 · Problem framing
Making Spotify Music Discovery Truly Mood-Aware
✳The problem they went after
A survey of 11 users aged 20 to 30 found the same split every time. Listening is mood-driven, searching is title-driven, and the mood and activity playlists meant to bridge the two feel repetitive and generic. When the vibe is wrong nobody complains; they open YouTube or another app instead, which is exactly why the problem stays invisible in support data.
✳What they actually did
Shereen traces it to infrastructure rather than interface. Discovery is built on metadata and behavioural signals, with no model of emotion or context, so there is a permanent gap between what a listener wants to express and what the search bar can process. The consequence is less listening time and quiet churn. A PIF-scored problem list ranks mood-music mismatch first, then repetitive playlists, then keyword-only search, and the top problem is tied straight to Time Spent Listening as the north star.
✳What you can take from it
Worth stealing
- PIF scoring ranks problems before any solutioning
- Silent churn to YouTube treated as the real signal
- Root cause traced to metadata-based discovery infrastructure
- Top problem tied directly to Time Spent Listening
Read this if
Building the case for a problem before anyone has asked for a solution
The transferable bitSilent migration is treated as evidence, since the users who leave without complaining are the ones your dashboards will never flag. The problems are also ranked before any solution is touched, so the feature work that follows has an argument behind it rather than a preference.
✳The submission itself
Open the PDF →Reading on a phone? The embedded viewer is cramped — open the PDF in a new tab instead.
Phase 03 · Solution exploration
Reimagining Spotify Discovery Through Emotional and Contextual Intelligence
✳The problem they went after
Spotify's search matches literal keywords and its playlists sit static, so a listener with an emotional or situational need has no way to say it. The product feels mechanical at the exact moment people want it to feel intuitive, and the result is disengagement and discovery that never happens.
✳What they actually did
Nine concepts get explored: NLP-based vibe search, an emotional gradient mood selector, a freshness slider controlling how novel recommendations feel, a conversational AI DJ, a discovery bridge for tracking down songs first heard on TikTok, and emotion tags with listening insights. Each is sorted into MVP, second phase or experimental rather than accepted or killed. The MVP comes out as three: Vibe Search, which takes a phrase like 'upbeat music to unwind after work', the Emotional Gradient Bar, and the Freshness Slider.
✳What you can take from it
Worth stealing
- Nine ideas tiered into MVP, phase two, experimental
- Vibe Search accepts plain mood phrases as queries
- Freshness Slider hands novelty control to the listener
- Emotional Gradient Bar replaces typing with a colour slider
Read this if
Deciding what makes the MVP cut and what waits for a later phase
The transferable bitTiering beats binary triage. Sorting ideas into phases keeps the ambitious ones alive without letting them bloat the first release, and the three that make the MVP cut work as a set rather than as three separate bets, which is what makes the shift from algorithmic to empathic personalisation legible.
✳The submission itself
Open the PDF →Reading on a phone? The embedded viewer is cramped — open the PDF in a new tab instead.
Phase 04 · The feature set
Emotion-Aware Music Discovery Features for Spotify Users
✳The problem they went after
The gap Shereen keeps returning to is that people experience music emotionally but have to search for it literally. Keywords and static playlists cannot take a mood or a situation as input, so listeners try again, try differently, get frustrated, and slip over to a competing app without ever registering a complaint.
✳What they actually did
Vibe Search reads queries like 'feeling good and energised' with natural language processing and returns songs with a Vibe Match percentage attached. The Emotional Gradient Bar puts a colour-coded mood spectrum on the home screen so a listener can slide to a feeling instead of finding words for it. The Freshness Slider gives direct control over how familiar or novel the next stretch of recommendations is, aimed squarely at habitual listeners hitting repetition. Measurement covers recommendation accuracy, first-month adoption and abandoned searches.
✳What you can take from it
Worth stealing
- Vibe Match percentage makes the model's confidence visible
- Abandoned searches measured as the direct pain metric
- Each feature covers a different discovery moment
- First-month adoption tracked alongside the north star
Read this if
Specifying AI features and want metrics that trace back to the problem
The transferable bitThe three features sit at different moments rather than competing for the same one, and abandoned searches is chosen as a metric that maps directly to the pain diagnosed earlier. Time spent listening stays the north star, so every feature has to argue its way back to the core loop.
✳The submission itself
Open the PDF →Reading on a phone? The embedded viewer is cramped — open the PDF in a new tab instead.
