28 ms·

We can only blame ourselves for Spotify's bad UI! (jk) Spotify uses an internal experimentation platform that guides which features they determine are successfu

by davmar 5y ago

We can only blame ourselves for Spotify's bad UI! (jk) Spotify uses an internal experimentation platform that guides which features they determine are successful, based upon how our usage moves their metrics: https://engineering.atspotify.com/2020/10/29/spotifys-new-experimentation-platform-part-1/ https://engineering.atspotify.com/2020/10/29/spotifys-new-ex... So, basically, the bad UI is improving their metrics. Or something. Maybe this is a case of experimentation gone wrong.

ec109685 5y ago
Sometimes it's hard to spot long term degradation of quality via short term experiments. Each little thing shows uplift, but overall it gets crapperier.
xmprt 5y ago
I can only assume one of their metrics is the classic "time on app" or "user engagement" which is a bullshit metric for a music player. If I'm spending more time on the app then that clearly means I'm not finding what I'm looking for.
- ajmurmann 5y ago
  I generally agree with engagement being a BS metric, I'm not sure I agree in this case, unless we separate out time I'm actively interacting with the app and time I'm listening to music and only call the former BS. When I first switched to Spotify from Google Play, I listened to a lot more music because the recommendation algorithm pointed me at a ton of music I ended up looking a lot. Now its recommendations have gotten to samey and my engagement declined. I'd love for them to improve recommendations and drive my engagement up that way. Now, similar example is Netflix. I've at times reached a point where I think I had watched everything on the platform I'd enjoy. Their UI keeps showing old stuff as new tough, different pictures for the same show, etc. This led to wasted time and frustrating engagement. (Aside: the best Netflix UI ever was a script I wrote myself when they still had a public API that just listed all movies ordered by how much they predicted I'd like it. That paired with IMDB ratings and filtered by what I've already have seen would be the holy Grail but it would be obvious when it's time to unsubscribe for a while)
adriancooney 5y ago
I'd imagine their metrics are aligned with user behaviour, not experience.
- g_p 5y ago
  Most probably - metrics tend to be focused on something advertisers like to hear, like page views (paginate your artist listings for more page views!), scroll-downs (add loads of whitespace and padding and large artist photos), video watches (play the video in the background of the page), song plays (make it hard to find the song you want, so you try 10 until you find the right one...) Given so many metrics are either time or attention based, they probably are opposed to user experience (less time needed to achieve the task).
- TeMPOraL 5y ago
  > their metrics are aligned with user behaviour Or rather, they're trying to align user behaviour. Since customers have very little choice in the matter (most popular SaaS products tend to resist commodification, so there's no equivalent competitor to jump to), these metric-driven approaches often effectively become PID controllers, driving user behaviours to a desired trajectory.

by davmar 5y ago

ec109685 5y ago

Sometimes it's hard to spot long term degradation of quality via short term experiments. Each little thing shows uplift, but overall it gets crapperier.

xmprt 5y ago

I can only assume one of their metrics is the classic "time on app" or "user engagement" which is a bullshit metric for a music player. If I'm spending more time on the app then that clearly means I'm not finding what I'm looking for.

ajmurmann 5y ago
I generally agree with engagement being a BS metric, I'm not sure I agree in this case, unless we separate out time I'm actively interacting with the app and time I'm listening to music and only call the former BS. When I first switched to Spotify from Google Play, I listened to a lot more music because the recommendation algorithm pointed me at a ton of music I ended up looking a lot. Now its recommendations have gotten to samey and my engagement declined. I'd love for them to improve recommendations and drive my engagement up that way. Now, similar example is Netflix. I've at times reached a point where I think I had watched everything on the platform I'd enjoy. Their UI keeps showing old stuff as new tough, different pictures for the same show, etc. This led to wasted time and frustrating engagement. (Aside: the best Netflix UI ever was a script I wrote myself when they still had a public API that just listed all movies ordered by how much they predicted I'd like it. That paired with IMDB ratings and filtered by what I've already have seen would be the holy Grail but it would be obvious when it's time to unsubscribe for a while)

adriancooney 5y ago

I'd imagine their metrics are aligned with user behaviour, not experience.

g_p 5y ago
Most probably - metrics tend to be focused on something advertisers like to hear, like page views (paginate your artist listings for more page views!), scroll-downs (add loads of whitespace and padding and large artist photos), video watches (play the video in the background of the page), song plays (make it hard to find the song you want, so you try 10 until you find the right one...) Given so many metrics are either time or attention based, they probably are opposed to user experience (less time needed to achieve the task).
TeMPOraL 5y ago
> their metrics are aligned with user behaviour Or rather, they're trying to align user behaviour. Since customers have very little choice in the matter (most popular SaaS products tend to resist commodification, so there's no equivalent competitor to jump to), these metric-driven approaches often effectively become PID controllers, driving user behaviours to a desired trajectory.