Spotify now counts over 140 millions users. 140 million users, meaning 140 million unique music profiles with unique music tastes (unless we’re talking about reggaeton). Massive amounts of user data are churned out every day, making personalization overwhelming at first, but increasingly feasible once the algorithms are tweaked just right. And thanks to its popularity, Spotify has developed one of the most refined renditions of personalized services in its market: Discover Weekly.
For those who aren’t familiar with it, Discover Weekly is a unique 30-song playlist crafted individually for Spotify users every Monday. Matthew Ogle’s brainchild, the service has gotten people hooked. I myself am constantly impressed by how accurate it is at pinpointing my quirky, ever-changing music taste. But most importantly, Discover Weekly does more than just provide me with what I like; it steps out of my comfort zone, surprising me with fresh songs that I didn’t even know I’d like. In fact, the playlist feels more like a music compilation crafted by a close friend rather than an automatized, algorithm-driven service.
So how is it so accurate? Turns out, Spotify is great at mixing machine and human data together to make Discover Weekly feel so spot on. While it utilizes powerful algorithms that scan raw audio tracks and acoustically matches your own songs to previously undiscovered ones, the service also looks through curated playlists – both by Spotify experts and close friends – to find your songs in their lists and suggest the ones that you haven’t listened to yet. The result feels, well, human.
This only goes to reinforce just how much Spotify knows about its users. And while digging deeper, I stumbled upon a feature that seems to only be available internally for employees and the occasional outside news editor. It seems like Spotify is capable of creating stunning visual representations of our music taste, both in heat map form and music genre form. The first version portrays how each user’s preferred music genres blend into each other, with darker hues representing the ones that they listen to the most. Also, this feature can name the exact genres and subgenres users listen to, divided into clusters that contain other similar genres. This is what Spotify calls the “Spotify taste profile”. Take the following example from Quartz reporter Nikhil Sonnad.
The feature has been made possible through the streaming giant’s acquisition of Echo Nest, a sophisticated music intelligence platform. But the question here is: why would Spotify not make its taste profile available to the public? Why, after many years of data mining, has the company not given its users access to such information? Backed by more than 200 users, some people even contacted Spotify directly with the same queries. The company’s answer? “Not right now.” I have asked friends and experts about it, and after reacting with confused faces and slight indignation, they gave me some ideas.
- Taste profile as a future monetized feature
While discussing with peers, the first argument to pop up was money. If Spotify has been hiding its taste profile from its users for years, it might be because it wants to implement it in the future (if you ask me, we’ve been waiting long enough) as a separate or additional feature. It might be branded as a premium and therefore sold to the public to provide the struggling music streaming service with extra cash in light of a dwindling premium subscriber base and skyrocketing licensing costs.
- It might render Discover Weekly obsolete
Another topic of conversation was Discover Weekly itself. Being one of Spotify’s most successful sub-services, we don’t know whether letting users have a clear overview of their music taste might actually disrupt its performance. Imagine knowing all of your favorite genres to the letter. Now imagine always being aware of which ones you listen to the most and what artists fall under each music category. While you might think you know that already, having such a sharp visual representation of your music taste might reduce the chances of you wandering outside your comfort zone. This would render Discover Weekly’s human-driven algorithms useless when faced with increasingly rigid audiences.
- Maybe Spotify isn’t as accurate as it wants us to think
Perhaps Spotify hasn’t cracked the code yet. After all, just like with Amazon’s product recommendations, some exceptions in user behavior might still stump the system, integrating unusual behavior – like friends using your Spotify account for a day – into habitual listening patterns. This might end up in displaying inaccurate data, and consequently inaccurate visualizations of people’s taste profiles. So maybe Spotify is still trying to learn from its mistakes and chooses to wait it out until it gets its cues right.
Regardless of the reason behind its shady behavior, Spotify still dominates the music streaming industry, providing people with awesome personalized services like Discover Weekly and acquiring more users by the day. So while we wait to see our taste profiles surface into the mainstream, we might as well enjoy what the company has offered us so far.