How Accurate is the Spotify Wrapped?

Hi,

If you're a spotify listener you probably have looked at your spotify wrapped, which is a website that shows your top songs and highlights that you listened to this year, since spotify tracks what you're listening to.

The final ranking was a bit surprising, too surprising.

Was congratulations by PewDiePie really first? There were a lot of slight errors.

Spotify wrapped most likely rigs it and puts some #10 artist on your top 5 just to make you surprised, so you will share it.

But how rigged is it?

I have last.fm which tracks every song you listen to. You can just connect it to your spotify and I started tracking form October. One of the confusing things about last.fm is that instead of a "play" like "you've listened to a song 10 times" they say "you've scrobbled a song 10 times." They have their own unique word called a "scrobble" xD. So whenever you see the word "scrobble" you can just translate it to "play".

https://www.last.fm/user/wizwilzo/library/artists


Looking at when I joined in March of 2017 till October, it's mainly Owl City and NCS xD. Lots of NCS uplifting and infinity.

50% of the songs I don't recognize today, so it still should be fairly accurate.



Let's go.


Right onto spotify wrapped in the fifth slide that shows your top artist I can check that.

Retrovision. Not a surprise. 

Last.fm approves.

Now how accurate is the next one?

This is when it gets messy. Alex Skrindo is 28th on my list...

Ok let's look at the other stuff first. They got Jim Yosef and Virtual Riot at the correct rankings, but Owl City is fourth on last.fm.

How did Alex Skrindo get here?

Let's look at something else to explain what my hypothesis is.

Look at my all-time

Skyline is by both Kovan and Electro-light, but is registered as only written by Kovan on last.fm while on Spotify it most likely registers a listen to both Kovan and Electro-light.

If last.fm did register collabs and gave my 468 scrobbles to both Kovan and Electro-light, electro-light would have a minimum of 468 scrobbles, and would thus be sixth place on my all-time artist, which is another reason why last.fm doesn't track collaborations.

So back to Alex Skrindo. He was featured in Ruby by Jim Yosef, which I listened to 116 times in the past year. 

Since Alex Skrindo has 81 non-collabed songs that are tracked on last.fm in the past year, 81+198=279, which would give Alex Skrindo the sixth spot.


Theoretically there are probably other collabs Alex Skrindo is in, which would bump him to fourth place, but if that happened Owl City would be fifth, not Virtual Riot.


They are 8 scrobbles apart, but both Owl City and Virtual Riot don't have any collabs I listen to.

I certainly don't know of anything that Virtual Riot is listed second on, so Spotify Wrapped rigged something about the artists. 




Moving on, slide 7 says stuff about the diversity music brings, and how I listened to people from 48 different countries this year.

Fun fact:

Panda Eyes is from Switzerland
Alan Walker is from Norway




Slide 8 is about genres, and since pop is very loosely defined I'm going to skip it since anything can explain anything when it comes to categorizing songs in a certain genre.




And then it's slide #9

Finally! 

According to spotify, my top 10 songs this year was:

#1=Congratulations by PewDiePie, Roomie, BoyinaBand

#2=Cake by Retrovision

#3=Skyline by Kovan, Electro-light

#4=Thunderstruck by Owl City

#5=Make Me Move(James Roche remix) by Culture Code, James Roche

#6=Cloud 10 by Fareoh

#7=Good Time by Owl City, Carly Rae Jepson

#8=Imagine by Jim Yosef

#9=Harvalien by LarsM

#10=Flutter by Diamond Eyes

And what does last.fm say?

The first two songs are swapped, third and fourth are on point, but the rest is a mess. 

I'm guessing Spotify put congratulations #1 because it was a song where I listened to it multiple times in a row to analyze it, as I liked it a lot when it first came out.

So I'm guessing spotify put congratulations #1 because it was more believable.

Now let's focus on that mess in the last half of the rankings. 

We can notice the margin from fifth to eleventh is 8 plays. 

We can break down  5-7, as those three are swapped with each other. 

I can understand why 5-6 could be changed, as they are one scrobble apart and some random glitch/start stop tracking could've been unclarified, but the margin between 6 and 7 is 5 scrobbles. 

Now the margin between 8-11 is two scrobbles, so I can definitely see some trend tracker to look at everything and determine the final tracks. 

so for that ninth slide? 

It's pretty accurate

Let's move on. 

The 10th and 11th slides are about podcasts

And 11.5 is introducing the decade stats.

I started listening and tracking in 2017 so all-time statistics are decade statistics. 

Since last.fm doesn't track minutes imma comment on the twelfth slide.

I've made an improvement(if listening to more music counts as an improvement) in the minutes every year, as 28000 minutes translates to 466 hours of music. That's legit more than 1 hour and 15 minutes of listening every day. Lowkey high school carrying xD. Imagine what the count would be if I didn't go on Euro Trip. As a total I've listened to 70,862 minutes of music in the past three years, which turns out to be a CONSISTENT 64 EACH DAY. 

I'VE LISTENED TO MORE THAN AN HOUR OF MUSIC EVERY SINGLE DAY IN THE PAST THREE YEARS ON AVERAGE.

That's kinda crazy. 

Aight back to the article

Slide 13 is nothing really that I can check. I did screenshot my last.year(ps I will make a last.year post later this month) last year and it does match up with the stuff in that image.

My artist of the decade is RetroVision which is cool and not surprising. 

Wide margin. 

This is the second-to-last slide, and the last slide is an ad to go premium. 

The squares do show the top 5 artists and songs.

In the all-time artists category the first two are spot on because of collabs Alex Skrindo from Ruby and Electro-light from Skyline by Kovan, Electro-light are squeezed in between second and third place. RIP culture code. 

I guess it's accurate.

In the all-time songs category 

Every single song of the five songs is correct. OKAYYY

Since we already analyzed the past year stats, let's get to the conclusion. 

Conclusion

       So nothing is really rigged or anything, it's just that Spotify has put other factors into your statistics, such as consistency, length, discovery, and trends. Despite last.fm and Spotify being different, leading me to make many assumptions in the collabs, it still can be concluded that the Spotify Wrapped for every user is accurate. There may be some slight fixed tweaks, but a solid majority of the statistics are accurate.

        So now you don't have to always be questioning the website whenever you see something surprising. You may just have listened to it a lot in the beginning of the year but it gradually died down.

Lyric(s) of the Post: Jingle Bell, Jingle Bell, Jingle Bell Rock.