So, we hypothesized that just one great recommendation could have a large positive effect on satisfaction. The positive connotation of that hope seemed to outweigh the “risk” that they wouldn’t connect with any of their recommendations. That said, they generally hoped to find a new song or artist that they’d really love. Hypothesis 3: One great recommendationįinally, users tended to approach their recommendations with the understanding that not every track would be a hit with them. Therefore, we hypothesized that metrics should be normalized relative to each user’s typical behavior, rather than assuming the same standards for all users. For example, streaming five tracks could signal a “good” week for a casual user who typically streams one to two tracks per week, or a “bad” week for someone who typically streams 15. We found that users judged their playlists relative to previous weeks, and the level of engagement varied between users. Hypothesis 2: Context from individual habits For example, a skip might mean something different for someone listening to music in the background than it would for someone saving music to listen to later. This leads us to our first hypothesis: that on-platform behaviors provide clearer signals when they’re viewed within the context of the user’s goals. ![]()
0 Comments
Leave a Reply. |
Details
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |