Being in tune with listeners: Enhancing recommendations through a reimagined Explore Page

Client

Amazon Music

Skills Applied

User Interviewing, Affinity Mapping, Wireframing, User Testing, Journey Mapping, Prototyping (Lofi-to-Hifi)

Timeline

6 weeks (Feb 2024 - Mar 2024)

Highlights

I researched what users wanted: How might we foster community on a music streaming app?
The prototype shows promises in usability and delight from users

The Design Brief: Given the current state of Amazon Music, how might AI enhance a customer’s experience with music and/or podcasts?

My findings

Leveraging AI to better tune recommendations to Amazon Music listeners

Through interviewing avid podcast listeners I found that 50% of users:

experienced trouble with tailored recommendations
preferred short-term content
rely a lot on others for their next podcast

“I'm always taking recommendations, but I always like vetting, especially for something that I wanna invest in time and effort and energy into listening”

— User Feedback

Through interviewing avid podcast listeners I found that 50% of users:

experienced trouble with tailored recommendations
preferred short-term content
rely a lot on others for their next podcast

“I'm always taking recommendations, but I always like vetting, especially for something that I wanna invest in time and effort and energy into listening”

— User Feedback

Through interviewing avid podcast listeners I found that 50% of users:

experienced trouble with tailored recommendations
preferred short-term content
rely a lot on others for their next podcast

“I'm always taking recommendations, but I always like vetting, especially for something that I wanna invest in time and effort and energy into listening”

— User Feedback

Leveraging AI to better tune recommendations to Amazon Music listeners

Issue 1

Users experienced trouble with tailored recommendations

Recommendation 1

A Gen-AI powered Alexa pulls data from wish lists, shopping habits, and read books to better inform your explore page algorithm.

Issue 2

Users prefer short-form and easily accessible content

Recommendation 2

Display bite-sized recommendations of video clips, quotes, posters, and merchandise using tiles

Issue 3

Users tend to rely on their social network for their next podcast

Recommendation 3

AI uses Natural Language Processing to recommend the best pieces to users' friends

Initial user testing shows promises

Initial user testing found small prototype errors and desire for more onboarding for the AI process. Overall, they were pleased with idea of short-form recommendations for their next podcast. More user testings need to be conducted, however, to further validate initial results and gain new findings.