Design a recommendation engine for Spotify.

3. Users want new content, and do not want to listen to the same old songs/playlists, they want to be introduced to new music they would have never considered and informed of new genres

4. Users want to know what their family/friends/influencers are listening to

Brainstorm solutions

1. Sync contacts, follow certain people and get insights into what they are listening to and liked, (users can become influencers for their taste in music and garner many followers)

2. If users skip a song in a playlist before 3 seconds pass, that song should be removed from that playlist on their specific account (use ML)

3. Search for music based on mood (sad, happy, lonely) and/or event (pre-exam, happy hour)

4. Suggest new genres users might enjoy based on their interest in other genres

5. Option to bleep out curse words

6. Option to remove/block certain artists

Evaluate solutions

1. Spotify might already have this option, if not, it could be easy to implement as the technology to sync and connect contacts is already there (LinkedIn)

2. This can delight users, as users want customized playlists

3. Spotify likely already has this

4. Can be insightful to users and introduce them to new content

5. Should be easy to implement as technology already exists (TV/radio/podcasts bleep out certain words)

6. Should be easy to implement as technology already exists on other platforms (we can block calls and certain followers on social media platforms)

Summarize and recommend

Spotify likely already does #s 1 and 3. And #s 2, 5, and 6 are interesting ideas, but they do not solve the prompt to "Design a recommendation engine for Spotify." I would focus on #4, because introducing users to a new genre based on their interests can expose them to a rich new selection of content. With ML, Spotify can build an algorithm that suggests relevant/similar genres based on what other users with similar tastes enjoy. Similar technology already exists (recommending artists based on similar tastes) so the technology should be viable to implement.

Tanu Mehrotra · Answer 2 · 2020-03-26T21:20:53+0000

Clarifying Questions?

1) Who are the users here? Are these all the Spotify listeners or these are the listeners of any specific age group? All the Spotify listeners

2) What does the recommendation engine mean? Are we trying to recommend the music, or the advertisements popping up, or the news? Music

3) Why are we trying to do this? Are we trying to increase the DAU, MAU or trying to increase the conversion rate?

Goal: The goal is to recommend music to all the listeners of Spotify and increase the DAU.

First I will list who the users are then will list out the use cases. After that I will list out the pain points of the customers and offer recommended solutions and evaluate those.

Users:

1) All Spotify listeners

Use cases:

1) Listen to the songs of their choice while driving or while playing or while running or gymming.

2) Listening to any specific songs of their favorite artist.

3) Maintaining the playlist of the user’s favorite songs.

4) Searching the songs of the specific genres like Romantic, party, wedding, or workout.

5) Listening to the radio of different kinds like Mirchi Masala, or Retro hits etc.

I will prioritize the use cases #1,#3,#4

Pain points:

1) Cannot pick a specific song of its choice while driving or running or gymming as it will be a distraction from the activity one is doing.

2) Some Random songs will start playing from a shuffle list which might not soothe the listener and he/she might try to change the song which might cause some sort of mishappening if the person is driving or running etc.

3) The choice of the song varies from people of different age groups so they might not get attracted to the songs which randomly start playing from the shuffle list.

4) Users want to know what is the new type of music which is there in the market and want to get attracted to those.

Solution :

1) When a person registers to Spotify, they can ask the user what kind of music they like and accordingly suggest to them the songs of that choice.

2) A person would have created multiple playlists and if the users have finished listening to all the songs it should play the songs of the same artist or the similar kind of songs what a person has in the playlist.

3) When a person is driving it can be integrated with Alexa or google play so over the voice command person can change the songs, so it won’t impact them.

4) When a person is adding songs in the playlist Spotify should suggest the songs similar to the ones person is adding in the playlist.

5) When a person is listening to the songs, the age is registered so songs of that specific age group can be recommended for e.g. older age people might prefer listening to religious songs.

Evaluate Solution:

I will prioritize the #1,#2,#5 solutions as the reach of the customers will be more for these offerings and it will have a more impact and require less effort and can be developed with confidence. I think #3 Spotify already has, if not it might require more cost and more effort. Solution #3 can be integrated with #2 but that #3 can be of less priority and can be done later.