SPOTIFY - Product Teardown
7 years as a user. Paid since 2023. 550+ liked songs. I think that qualifies me to tear this down.
I’ve used Spotify since 2019. Paid user since 2023. 550+ liked songs, 15+ playlists, 10+ blends. I think that qualifies me to tear this down as a user and as a PM.
WHAT IS SPOTIFY
A platform to listen to music, discover new songs and podcasts. Simple as that.
The north star metric: Monthly active listening hours.
WHAT SPOTIFY GETS RIGHT
The first thing I do after opening Spotify is head straight to a top album or a Blend playlist with a friend.
Blend is my favourite feature on the platform. And I think it’s underrated. It doesn’t just play music. It uses music as a social connector. It calculates a match score based on both users’ listening history and liked songs, then builds a mixtape mixing both parties’ favourites. It updates every single day based on real-time listening trends.
I’ve used it as an icebreaker with strangers, cousins, and new friends. Works every time.
My only question for Spotify: why stop at two people? This could easily be a group activity.
The second thing I recently noticed is the mood-based filter under Liked Songs. My friend claims it’s old, but I noticed it this year. Either way, it’s genuinely useful. For someone with 550+ songs, creating separate playlists for every mood was time-consuming. This filter cuts that entirely.
3 INSIGHTS
1. SMART SHUFFLE - right intent, wrong execution
Smart Shuffle was built to solve music discovery. The intent is good. The execution is not.
Here’s what actually happens: I’m listening to soft pop. I turn on Smart Shuffle. Spotify throws a high-energy Punjabi song in between, completely shattering the mood and tone of what I was listening to.
Discovery should feel like a natural progression, not an interruption.
2. MOOD FILTER- solves sorting, misses language
The mood filter is genuinely useful, I said that above. But it’s incomplete for multilingual users.
I listen to Hindi, English, Kannada, Bengali, Telugu, Nepali, and Spanish. My liked playlist has songs across all of these. The mood filter doesn’t account for language at all.
This isn’t just an Indian problem, though India makes it extreme given the number of regional languages. Countries like Spain, with co-official and immigrant languages, face the same gap.
3. RECOMMENDATION ALGORITHM- the data is there, the dots aren’t connected
Spotify’s algorithm fails to connect language preference, mood patterns, and social signals, even though it has access to all of this data.
People often say YouTube’s recommendation engine is better because it has Google’s data behind it. Here’s my counter: I don’t use Google to search for songs. So Spotify actually has almost everything it needs: search history, liked songs, language patterns, and Blend social data. It’s just not using it well.
WHAT I’D BUILD
Fix the recommendation algorithm, and fix it properly.
The algorithm should graduate through genres gradually. If I’m listening to soft pop, recommend pop next. Then pop-rock. Then rock. Then maybe something more energetic. Don’t jump straight to Punjabi.
This one fix cascades into everything else. Smart Shuffle becomes usable. Discovery feels natural. Listening hours go up. The north star moves.
One fix. Multiple problems solved.
WHAT I’D NOT BUILD
Short-form video content.
Instagram Reels owns that space. Spotify’s strength is audio — the immersive, uninterrupted listening experience that nothing else does as well. The moment it becomes a visual platform, it loses that identity.
I noticed Spotify has already started experimenting with short clips. I hope it pulls back. Stick to music and podcasts. That’s the moat.
This was my attempt at a product teardown. Follow along if you want to join me on my job hunting journey!
Thanks for reading, more soon :)
Cheers,
Apoorva 🪩
Quick correction from a reader - turns out Blend can actually support up to 10 people, not just two. I stand corrected!
Really enjoyed this teardown, Apoorva. I used to love the Blend feature, too! I think it's one of Spotify's most underrated features because it does something most music products fail to. It turns music into a shared experience. The fact that you've used it as an icebreaker with strangers and friends says a lot about the value it's creating beyond just listening. I also thought the language gap in recommendations was a sharp observation. For multilingual users, language isn't just metadata; it's often tied to mood, context and identity and most recommendation systems still seem to treat it as secondary.