Spotify-This or That

Overview

Using a few simple interactive steps, assist users in discovering music that better aligns with their preferences and feelings.
‍

Background

Spotify uses machine learning to serve music and podcasts to its users in various personalized ways with Signals, like:
‍Discover Weekly, Daylist, Algotorial playlists, AI DJ‍

Signals are ways for platforms to understand what users enjoy, what they want more of, and what they don’t want to see.
‍

How might Spotify better serve music discovery behaviors using AI on new or existing surfaces based on both implicit and explicit signals?

Explicit Signals.These are signals that are given directly from the user to the platform. Much of the time, they manifest as UI elements indicating liking or disliking something.
‍
‍Implicit Signals.These signals that are gathered by a platform either through demographic data or are behaviors and motivations slowly uncovered over time of usage.

Design Process

User research

We gathered the problems and feelings people usually have with Spotify and grouped them.

Insight:
Spotify makes a recommendation they won't listen to -- they feel: doubt about it because: they don't know what it's based on
‍

HMW make recommendations be more reliable?
‍

Ideation

We had a brainstorm to generate ideas
- Combine the Explicit Signals and Implicit Signals
- User can be aware of the signal, knowing how it comes.

Ask users to pick up 3 choices they like. --> This or That!
The user will be asked to make a choice in a second, then the answers are used as explicit signals in recommend music.‍

Do you like Cat or Dog?
Do you want to relax in the City or Nature?
Are you feeling Happy or Sad?‍

Storyboard

Delivery

User flow

How user approach to recommend stream‍

UX design

Start at --‍

Search page and my Library where user discover new music and add create new playlist‍

Make your choice -- ‍

‍
User will be asked to choose 3 sets options about Situations, Actions/Things, Feelings

These questions and options could be different in each time. The questions are generated by AI.‍

Generate Playlist -- ‍

‍
Spotify’s user experience has encouraged playlist-first listening, so the recommend music is presented in the form of playlist. The length of the playlist is in a proper time, about 45 mins.
‍
Considered that people may not satisfied with the playlist, or they want to change their mind, I designed Regenerate and Reset in the end of the page.‍

Where to find -- ‍

‍Now the user like the recommend playlist very much! Where can they find it afterwards?
The user experience is consistence with playlists in Spotify, which is either find it in the Recent bar, or saved in my Library.‍

Iteration

1. The Access Point

In the user testing, participants barely find they can create the recommend playlist in Library. Firstly, the entry is very deep, and there is no text or icon clue outside. Secondly, the Library is a place to storage the added music, rather than discovery.
So we only keep the access point in the Search page.

2. The UI design of making choice

The user is not clear on how to choose a pic without a question. They need to answer based on a specific question otherwise they will choose a pic in a better look. So we add the text of questions.The UI is full with pictures, and mixed different visual styles. I developed a visual component that combines the SP and option images. The interaction should be light and fun. While pressing on the SP will rotate.

Reflections

This project posed unique chance due to the reveal of how Spotify recommendation system works with Signals, pushing me to innovate in conducting user research and gaining valuable insights. Addressing the interesting topic of music, I delved into the discovery problems and human desire. This deep understanding informed the design of our solution, This or That.