Assisted curation
Designed at Spotify in 2018
Playlists has always been the bread and butter of Spotify but the experience has largely stayed the same since the inception. With assisted curation users can take full advantage of machine learning and personalisation to make the playlist creation easy, delightful and fun.
My role
Design lead • Wire-framing • Prototyping • Visual design • UI specification • Continues quality assurance
The team
Me as design lead • Fellow designer • PM • Squad of front and back end engineers • User researcher • Data analyst
The problem
People perceive it to be a lot of work to manually curate all songs and this is assuming that people even knows the name of the songs or artist they want to collect.
We had often heard people say in user research and diary studies that they wold love to create some playlists but they knew it would be a lot of work and had never found the time to do so.
The vision for this project was to create a solution that makes playlisting easy and fun by doing all the heavy lifting on the user’s behalf.
Principles
The first thing I did in this project was to work with the product manager to form design principles. These principles was then used throughout workshops and design work to ensure the right type of solution.
Ideation
The first workshop for the project was an office wide ideation session that I hosted. The participant was anyone who wanted to join including engineers, product managers, testers and more. This session allowed me to quickly collect loads of ideas and involve a larger group than I traditional design sprint affords.
In addition to the office wide workshop I also ran two design sprints with participants from Stockholm, Gothenburg, New York and Boston.
Design process
I often use an internal design framework that we call Thoughtful execution. It’s a method that ensures a shared picture of succeess and making sure I as a designer explores a wide set of solutions to prove or disprove a wide set of well defined hypothesis, all backed up by data and insights.
Wireframes
Before committing to a final design solution I quickly created a number of diverse wireframes. This was generally done using Figma as this allowed me to easy collaborate and share my work with my fellow designers in other offices.
Prototypes
For the user research, I created several prototypes using inVision, Flinto and Principle depending on the requirements of the prototype.
Design specification
With all the design in place, I created a design specification for the engineers and testers. The structure was to first describe the key screens followed by any redlines that might be needed. I then described the key flows and lastly described any error cases and other miscellaneous information.
Final product
At the core of assisted curations are cards with song recommendations that grow as the tracks are added to the playlist. It also include a new integrated search interface, previews and more.
Results
After two sprints and three AB tests, the ML powered playlist solution was shipped in Q2 2018.
This design was a part of a much larger, multiyear, project to redesign the free version of Spotify from the ground up. My design was the single most effective lever to longer-term retention in that project. Other stats include:
- 1.7 percentage points increase of week 2 retention
- 12.32 % cumulative week 2 music consumption per user increase
- 9.3 pp increase of playlists created