Back home

Discovery

Music discovery with taste, not only tags

Music product · listening research

A calmer way to find the next song

Playlists often ignore where I am. Here, time of day and who is nearby count as inputs, not footnotes.

The exploration balanced suggestions with curation at scale. The UI had to show both without turning settings into a spreadsheet.

This write-up covers narrative, interaction, and visual direction. Implementation will change; trust and explainability should not.

Role
Product thinking, UX, visual design
Research
Listening diaries, competitive audit
Client
Self-initiated
Phase
Concept through interactive prototype

Try the prototype if you link it, then come back for the process notes.

01

Research and listening diaries

What people say versus what they replay

Short diaries captured where people listened, who held the queue, and when they skipped. Social context often overrode the taste model.

I skipped long surveys. Quick prompts after sessions gave clearer signal than abstract genre questions.

  1. 1. Map contexts

    Commute, focus, party, wind-down alone. Each context sets different priorities for energy, lyrics, and risk.

  2. 2. Find trust breakers

    When someone calls a pick “random,” trace it to missing context or a weak explanation.

  3. 3. Editorial gaps

    Spot genres algorithms flatten. Those gaps are openings for short curated capsules.

02

Information architecture

A hierarchy that protects focus

Primary navigation stayed small: Now, Collections, Live, Profile. Everything else stays secondary or contextual.

“Why this track” sat in the main flow. If the system cannot explain a pick in one sentence, the pick was not ready.

03

Visual identity

Dark UI with warm accents and readable density

Contrast was tuned for night listening. Accent color marks live or social features without shouting.

Display type carries editorial voice; UI type stays readable for long artist notes.

04

Motion and feedback

Micro-interactions that respect the audio

Control motion stays under 200ms so it does not fight the beat. Larger motion belongs between sections, not on every tap.

Waveforms and progress bars were simplified to stay light on older phones. Felt speed matters as much as benchmarks.

05

Outcomes and next steps

What would ship first on a real roadmap

A production pass would start with honest explanations and simple playlist handoff. Deeper personalization can wait until those feel solid.

Open questions: licensing, offline behavior, and how much social data people will trade for better matches.

What did I learn?

  • People forgive uneven picks when the system explains itself.
  • Context belongs beside taste, not in the footnotes.
  • Pair algorithms with editors for anything cultural or time-sensitive.

Notes from diary participants

I kept skipping until the app admitted it was guessing. After that I trusted the next songs more.
Participant A, Listening diary
I want fewer tabs. If live and on-demand live together, show me how they connect.
Participant B, Listening diary
1 / 2
Design
Laolu James