[Feedback needed] ListenBrainz statistics proposal

development
Tags: #<Tag:0x00007fde0264a950>

#1

This thread is a companion to Statistics in ListenBrainz and the GSoC 2017 proposal. However, this proposal focuses on what statistics to collect over the technical implementation.

You can leave comments in this Google Docs page too!

Summary

Implement new statistics to calculate insightful data on the listens submitted to ListenBrainz, to make the platform more competitive to other solutions and more interesting to the user

Background

Currently, ListenBrainz collects two types of statistics:

  • Sitewide (code)
  • Individual user (code)

Sitewide statistics

ListenBrainz calculates a single statistic for the entire site:

  • Total number of artists submitted by all users

User statistics

ListenBrainz calculates the following statistics for users:

  • Top recordings over a time interval
  • Top artists over a time interval
  • Top releases over a time interval
  • Number of artists listened to over a time interval

Why now?

ListenBrainz does not yet collect many statistics. In the future, this will become an important focus for the platform to compete with other proprietary services and provide insightful information to the user.

Details

Two types of statistics were gathered for this proposal. Some are already implemented on other third-party sites, such as Last.fm, and others are new ideas that are not implemented by other competing, proprietary services.

Note that “entities” refers to artists, releases, and recordings.

Third-party statistics

Ideas for new statistics from third-party sites include…

  • Time spent listening: Total time spent listening to music over a time period
  • User percentiles for entities: Compare your listens for different entities compared to other users on the site (e.g. 80th percentile for number of artists listened to in a time period)
  • Top tags from entities: See patterns in tags from listens (i.e. pulling tags from MusicBrainz entities)
  • Listening clock: See when a user listened to the most music at what time of day over a time period
  • New discoveries: Highlight new entities that a user has not listened to before
  • Musical matches: See other users who listened to similar entities as you
  • Mainstream meter: Percentage of how many other users shared your listens
    • Example: If every user on the site listened to one artist, and a user only listened to that artist over a time period, their “mainstream meter” score would be 100%

New statistics

New statistic ideas not yet implemented by a competing service include…

  • Revivals: Highlight when a user listens to an entity that they haven’t listened to since a period of time
  • Streaks: Entities that consistently appear in listens in a measure of time
    • Example: User listens to the same artist every week for six weeks
  • “Love at first sight”: New listens for entities that are over a certain threshold
    • Example: User listens to a recording for the first time, and over a week, they play that recording over 100 times
  • “Where did you listen?”: Highlight what client a user submitted listens from
    • Listens do not currently collect client identifier (I don’t think)

Proposed action

This proposal maps ideas for new statistics and provides a chance to consider a long-term approach to how new statistics are added to ListenBrainz. Feedback is welcome, and a concrete action for where these suggestions should go is needed (new tickets?).


#2

These are really good ideas!


#3

Thanks for making this list! We used this as a starting point for talking about LB graphs in our mini-summit here in Delhi. We’ll post summit notes with mock-ups of some of these graphs in the next day or so!


#4

Awesome! :smile: I’m happy these were helpful. I’ll keep an eye out for the summit notes.


#5

Just curious, did this end up anywhere yet?


#6

In case you are looking for the notes: