How Does Spotify Wrapped Work?

It's the most wonderful time of the year: snow is gently falling, hot chocolate is flowing by the gallon, Mariah Carey has once again moved to the top of the music charts, and your Spotify Wrapped just dropped.

To put it simply, Spotify Wrapped is genius marketing. The company measures each user's top artists, genres, and songs and compacts the data into eye-catching, easily digestible "cards" that are then shared far and wide on social media. Wrapped is personalized to each user's experience; it surprises us (I'm in Taylor's top 1% of listeners?!?), and in marketing speak, it drives long-term engagement and loyalty.

So, just how does Spotify Wrapped work? Who are the people and algorithms behind it, and how do they collect this data with respect to users' privacy?

A brief history of Spotify Wrapped

Since Spotify hit the market in 2008, it has become the most popular audio streaming subscription service, with over 574 million users. It boasts over 100 million songs, 5 million podcasts, and 350,000 audiobooks, a feature added in 2022.

The annual event, now known as "Wrapped," was first introduced in 2015 as the fledgling "Year in Music," but it didn't go viral until its 2017 upgrade when the visually stunning Wrapped cards blew up.

Splashed onto a poppy background, 2017's Spotify Wrapped gave an overview of the user's total listening minutes, their top genre, artists, and songs.

An annual social phenomenon was born.

Over the next three years, Spotify continued to release year-end summaries of users' listening habits, and people ate it up. In 2019, the aesthetics improved with vivid colours and eye-catching contrast. 2020 saw the addition of podcast listening habits, and 2021 took a softer approach to their Wrapped design and text flow after a hard year. In the spirit of staying connected, 2021's Wrapped also let users compare their listening profiles with loved ones and introduced them to their "Listening Aura."

The biggest revolution came in 2022 when Spotify took Wrapped to a new level, introducing 16 different "Listening Personalities," or what your music listening habits say about you. They also introduced the "Audio Day," revealing to listeners how their music personalities fluctuated from morning until evening.

What awaits the green circle's eager listeners in 2023 is a mystery. No doubt we'll get an overview of our top songs and artists, but Spotify has shown they're willing to go beyond the simple top ten lists and connect our identities to our listening habits. Or rather, define us by our listening habits.

Why Spotify Wrapped Works

The music you listen to forms your identity

In Spotify's most recent upgrades to its Wrapped feature, their designers have mimicked personality tests, even going so far as to give users a four-letter acronym akin to the Myers-Briggs test. Giving each user a "unique" personality badge ties their identity to the music they listen to. This is somewhat ironic, as with a company with a large user base like Spotify, you most likely have the same "unique" personality as millions of others.

It's not only your identity but your musical story. When Wrapped drops, it takes you through your music journey, revealing your changing tastes throughout the day and building up anticipation to show your listening personality and what it means to you. Congrats, wanderer, your listening type is "The Nomad." And like that, you're hooked.

The nostalgia factor

Sure, we get hyped when November's end rolls around, and you wait to see your top five songs, but going a step further, the data experts also deliver a playlist with your other top 96 tracks. Assuming you stick around for the following years' Wrapped moments, you accumulate a library of songs tied to specific events in years gone by. You can then return to these lists to be reminded of good and bad times and how you've grown through song.

It’s a fricking beautiful social experience

Let's not mince words; the Spotify Wrapped cards are gorgeous. So gorgeous, in fact, they've won awards. It's a masterclass in design — little text that says a lot and a contrasting, satisfying colour palette. The marketing specialists at Spotify could do all the personalization in the world, and yet Wrapped would be nothing without the designers who create those oh-so-shareable cards.

The people at Spotify have grasped one simple concept: music is a social experience. When you share your Wrapped results across social media, you are generating free marketing for the company by driving web traffic, AdSense data, and taste profiles. Ultimately, what Spotify Wrapped is doing is taking your randomized music data, refining and organizing it, and giving it back to you all "wrapped up" in a pretty user experience, and they do it during the holiday season to boot. Genius.

How it works

I could write the word "algorithms," and this section would be complete. But, you're here to learn how mass amounts of data are amalgamated into a neatly summarized social post for millions of people. So, let's get started.

Anyone who listens to at least five artists and 30 songs for at least 30 seconds is gifted a Wrapped. The tracking period starts on January 1 and ends on October 31, giving Spotify the time to organize and visualize the mass amounts of data. This may explain why that song you had on repeat that reflected your moody November didn't appear on your top songs.

Categorizing your top hits is simple metrics: you listened to Noah Kahan's Call Your Mom 557 times; that's song #1 (I'm making some personal predictions here). Endless Nights by Cian Durcot comes in at #2 with 502 streams, and so on.

But you probably would have never listened to most of the songs in your end-of-year bests without Spotify's recommendation systems. The short explanation is that Spotify tracks the music you listen to, categorizes it, measures listening habits against other users, and uses that information to suggest songs you may like.

Here’s a more in-depth breakdown —

Tools and algorithms that drive Spotify’s recommendations

Collaborative filtering

Collaborative filtering compares listening patterns to find similarities between users' listening habits and make recommendations. Analyzing this information and interactions, Spotify determines which songs are in similar categories and which users share preferences.

Collaborative filtering generates user taste profiles for more personalized recommendations. Spotify achieves this by logging user activity and categorizing it into listening sessions. This provides contextual information that is integral to predicting user activity. Data scientists can "weigh" certain aspects, known as explicit feedback, of this information as more important than other characteristics, known as implicit feedback. Explicit feedback is track saves, skips, and clicks, while implicit feedback is listening length and repeats.

Content-based filtering

Content-based filtering analyzes the track's metadata and raw audio to learn about its unique characteristics. This includes analyzing track titles, artist names, genres, moods, etc. Using "audio models," machine learning experts can pump out a list of sonic characteristics like key, tempo, and danceability to recommend songs with similar features.

Natural Language Processing

Natural Language Processing (NLP) allows Spotify engineers to transform playlists into textual documents to identify relations in lyrical patterns. NLP is also used in instrumental music as it easily goes between genres and introduces listeners to new music. Remember the digital nomad personality? That's thanks to NLP.

"ANNOY"

According to Erik Bernhardsson, the man responsible for building Spotify's recommendation systems, "ANNOY (Approximate Nearest Neighbors Oh Yeah) is a C++ library with Python bindings to search for points in space that are close to a given query point. It also creates large read-only file-based data structures that are mmapped into memory so that many processes may share the same data."

In human-speak, ANNOY can search for points close together in a particular query point and make recommendations based on them. For example, songs with a certain beat per minute would share a query. ANNOY can also allow indexes with the same data to be shared among multiple recommendation processes, speeding up the system.

Oskar Stål, Spotify's VP of Personalization, explains it this way, "Imagine you and another person have four of the same top artists, but your fifth artists are different. We would take those two near-matches and think, 'Hmm, maybe each person would like the other's fifth artist' and suggest it."

All of these processes are used to recommend music you may like via your "Discover Weekly." Usually, they can get at least a few songs per playlist recs that get you hooked. Then, once the song repeats enough, it shows up on your top 101 list, and Spotify can officially dub you a "Genre-bender" - not because you sought out new music, but because Spotify nudged you in the right direction.

Does Spotify know us all too well?

Spotify does a great job at repackaging our data and giving it to us as satisfying top 5 lists. As tech columnist for the Wall Street Journal, Nicole Nguyen, points out, you allow access to your listening history once you sign up for Spotify and agree to its terms and conditions. According to her, it's an example of "joyful data collection."

However, through this data collection, Spotify knows a lot about you. They understand how your mood fluctuates throughout the day, if you've recently gone through a breakup (Lewis Capaldi is your soundtrack), or potentially if you have a child (Let it Go on repeat?). Your podcasts reveal your political leanings (devoted to NPR or Joe Rogan?), religious affiliation, and interests (history buff and probable nurse into astronomy, crafting, and true crime). On top of that, Spotify has your location data.

Spotify doesn't sell user data to third parties, but all that precious data is stored for as long as the account exists. In fact, plastering users' habits on Billboards in certain locations may give advertisers and marketers all the information they need.

Unwrapping music, marketing, and privacy

Spotify Wrapped is fun; there's no denying it. I love mine, and I don't really care if Spotify knows my listening moods. I agree with Nicole when she says that it reveals something about our behaviour we didn't know before. But this joyful holiday greeting comes with the knowledge that Spotify knows you because they have loads of your data. The same goes for many other online services (Netflix, Amazon, Facebook, X, etc), so it's up to you how much you want big tech to know you.

No matter your position, you have to admit that Spotify may have come up with one of the greatest marketing campaigns of all time. This author wonders if Noah Kahan will finally dethrone NEEDTOBREATHE after years in the top spot. Either way, I'll raise my mug of hot cocoa, turn up the Holiday playlists, and await a digital package from a distant land.