My Vibe for two
How we launched an endless music stream for two people from scratch in 2 months, combining the tastes and preferences of each participant.
My role
I was responsible for the full design cycle, shaping the concept, conducting interviews, participating in defining the MLP and target product version, defending the concept before the team and stakeholders, preparing the design for development, supporting development, and conducting design reviews.
About the project
My Vibe is an endless personalised music stream based on ML recommendations. The algorithms analyse the listener's musical tastes, respond to likes, dislikes, and skipped tracks, and build a personalised musical experience. However, this experience remained individual, users could share tracks and albums but not their musical tastes.
Mission
To turn an individual experience into a shared one and enable creating a joint My Vibe that combines the musical tastes of two people. The algorithms analyse each participant's preferences and form a unified recommendation stream interesting to both. A common use case is a car trip where people want to listen together rather than choose whose track plays next.
Goals
- ✦ attract new users
- ✦ increase NPS
- ✦ increase listening time within the service
- ✦ improve user retention
- ✦ enhance subscription value
- ✦ grow retention
Constraints
The project had to be launched from scratch in two months. The MLP version limited participants to two, allowing faster hypothesis testing, feedback collection, and providing ML team data for training the recommendation model. Technically, the model had to find intersections in musical recommendations of two users so the stream pleased all participants, not just a shuffle of favourite tracks.
Work process
✦ concept 3 days
The concept was created without regard to technical or visual constraints to align the team on the target project vision quickly. Daily meetings with stakeholders were held to challenge and validate the concept.
✦ research 2 days
Due to tight deadlines, I conducted hallway usability tests to validate the concept. I built a prototype in Cursor, selected respondents, and ran usability tests. The research helped identify points in the scenario where respondents faced difficulties.
For example, initially in the Collection section, I used only the second participant's photo to quickly show who you were listening with on My Vibe for two. But 4 out of 6 respondents needed time to understand it was a combined stream, not another person's recommendations.
✦ MLP preparation 10 days
Alongside development and product teams, we defined which scenarios were essential for launch. The first iteration included:
Creating My Vibe for two via invitation link and accepting the invitation by the second participant
Saving and managing vibes in the Collection
Renaming and deleting My Vibe for two
Listening to My Vibe for two
✦ development and support 1.5 months
The entire development from scratch to launch took 1.5 months. During this time, we encountered new constraints and found compromises. Before launch, I conducted design reviews across all platforms.
Working with constraints
The MLP version limited participants to two, enabling faster hypothesis testing, feedback collection, and providing ML team data for training the recommendation model. We understood users would want to create shared vibes for groups of friends or family in the future, so we designed for scenario scalability from the start.
Outcome
In two months, we went from idea to launching My Vibe for two. Users gained the ability to create shared vibes, save them in Collections, rename, and manage them. Project metrics are under analysis, but the key product hypothesis was successfully launched and forms the basis for further scenario development.
Launch results
soon
My vibe & AI-commentsYandex Music