Cream City Calculation

Decoding the Melody

• Salim Fadel • Season 1 • Episode 4

Share episode

Podcast Description:

In this episode of Cream City Calculations, Colleen, Frankie, and Sal delve into the fascinating world of data analytics in the music industry. They explore how big data is being utilized to enhance everything from playlist recommendations to tour planning and music production. The team also discusses how platforms like Spotify use algorithms to shape your music experience and the broader implications for artists and the industry. Sponsored by Continuus Technologies, this episode sheds light on the powerful intersection of music and data, offering insights into how your favorite tunes are curated. Tune in to discover how data is transforming the way we listen to and create music.

Colleen:

Welcome to the cream city calculations podcast. We're three colleagues and friends that love data and to talk about how data is impacting our lives. I'm Colleen,

Frankie:

I'm Frankie

Sal:

and I'm Sal.

Colleen:

Hi, and welcome to episode four, Decoding the Melody. Today we'll be talking about data analytics in the music world. we're not going to be talking about AI and music, but rather how big data is being utilized to enhance the music industry.

Frankie:

But before we do that, let's dive into our data pulse. our quick hit source for this month's most impactful data news. So the first one I think I want to bring up is the Boston Celtics One, the NBA championship. And so one of the things they're talking about is that the Celtics think that they are utilizing analytics in the most effective manner and they're using it way better than any other team.

Sal:

This is why the buck should hire us.

Hire

Sal:

us guys are available.

Frankie:

So they're saying that like basketball is not just basketball anymore. And so there it's who's utilizing statistics in the best. to be able to win the game. And so the Celtics coach is saying that they are obsessed with data and that they hunt for the highest percentage shots, which happened to be their three pointers and they force low percentage opportunities from opponents. have you guys ever seen the movie money ball? No. it's about like baseball and how they use analytics. It's like the best optimal lineup. This is like exactly like that. But now you can see that analytics has now moved into the basketball realm, which is really they take a ton of stats there. They know that not hitting a mid range jumper, It's not efficient enough, I guess, because it doesn't add enough value, but there's enough risk there because it's far enough away. and so it's hit a layup or three, because those are the two main things. That's why the NBA has shifted that way. but it's funny cause actually like Jason Tatum is actually like one of the best mid range shooters out there and which is actually like a big thing. Of why probably they're, they are successful.

Colleen:

Yeah. I thought the article had some really interesting points about how they knew that they're going to be up against Antic, but they weren't going to try and completely stop him or double team them because he was going to get points, but they thought that their efforts or their energy was better spent on actually scoring more points than trying to stop him from scoring points. So it was almost like they were willing to give some up in order to get more.

Frankie:

I mean, that's why they let. Yannis shoot his free throws, right? Sorry, Yannis. but another interesting statistic is that the Celtics don't have a player who finished in the top five of MVP voting this season. So that's really interesting that this team who is not recognized as like those top players or anything like that, are doing so well and they were able to defeat everybody else.

Colleen:

Yeah, they didn't have one strong person. They had five strong people. Yeah, they

Sal:

I wonder how the balance between using this analytics to build a team where you're like, this is a more well rounded team. You maybe don't have that massive star. You could debate if Jason Tatum is a massive star, but whatever. he, but like using a team to be more well rounded versus having those elite, like the big three that they always have, or having that major player that just. Brings in all the fans and all that and can score 40 points by himself. which one is more effective and based on this article and based on what the results did this week, it, it definitely seems like this is a better strategy. I just don't know if the mentality of the NBA will like this.

Frankie:

Yeah. And then another strategy that they utilize was that, the Boston took 207, three point shots Series and Dallas took 152. I mean, it goes to show that like the more shots that you take, the more that you're going to make. And that's one of the strategies that Boston, was utilizing.

Sal:

That is crazy to think that's a lot of shots. I mean, that's a big differential, especially cause I, like the Mavs are not bad players. Shooter like like Luca is fantastic at shooting you've Kyrie. That's fantastic and dribbling like in getting in the lane Like yeah, that's crazy

Frankie:

That would be off by that much. Yeah. Yeah. Yeah And then one more quote that I just wanted to bring up was that they said Are we playing the right way taking the right shots and giving up the right shots? And I felt like that was just a really powerful like Quote around how to utilize analytics in basketball.

Colleen:

Yeah

Frankie:

the next article that we can talk about Is really incredibly interesting to me and it's around how pet data could help speed up your health tests. And so it touches on a couple of different companies. So there's one called tech site, and then that is like a human AI health company. And then one called Zotus, which is an animal health AI company, and they are partnering. And so they're performing AI based testing on animals, like using fecal and urine samples. So not hurting animals or anything like that. and they're collecting, around 50, 000 samples from vet clinics and then digitizing that into data and using an algorithm to understand what's a parasite, what's not a parasite, or what's a tapeworm, and what's not a tapeworm. So oddly enough, this information is like the same across humans as it is for animals, so they can utilize that animal health information. so

Colleen:

maybe train their models better so that they can make just as good predictions on humans,

Frankie:

right? Yeah, exactly so now they're like partnering with Mayo Clinic and they're gonna use some of those algorithms to try to help detect parasites and tapeworms and they're also going to be doing a Cancer detection in Mayo Clinic is digitizing 25 million tissue samples that they've collected over the years to train algorithms

Sal:

That's a, that's probably a model that you don't want to miss. Yeah. Cause the dog went for the,

Frankie:

Yeah, I think so. One important note here is that this is meant to be a second opinion for the doctor that does not replace the doctor. The doctor will still make the decision, but they are able to use this technology to help them like, either. help them with recognizing maybe there's a different pattern that they're not looking at or maybe it's just solidifying what their thought is. Yeah.

Sal:

It's really interesting. Like my, actually my, my sister in law's mom, owns a CT scan company. and I wonder if they're going to start selling, again, they probably can't sell that data, but I wonder if that data or those things can start to go into a larger,

Frankie:

can they sell de identified data? Maybe,

Sal:

honestly, I don't know. I don't know what the restrictions, but yeah, then they sell that information and then they can have, I start to build that as an income as well. That's it's really cool

Frankie:

Yeah, and so they're actually thinking this will start to be available by next year. Oh, that's crazy. Yeah and I just wonder like How expensive is it going to be and will you know your? Regular area clinic be able to afford something like that. i'm not so sure so that might be one of the downsides of that technology, but It is pretty amazing and I think that it'll be really interesting another potential downside to it is just what if that Data becomes identified it claims to be de identified data, as the samples are digitized, but you never know if is there a way that they can connect to your

Sal:

so the Just from i've worked in the past with de identified data and how it actually goes through that pipeline. That's actually you one of the main things that you have to do is actually send it to a third party That de identifies it and then sends it back. So you never act like you're not the one deep You know wherever you're getting from sends it directly to that third party then De identifies and then it gets sent to you. What if that

Frankie:

third party is hacked

Sal:

that there? that's the only place that it could be a risk, but they I would still say that there is very minimal because it's they don't do any like really human interaction in that aspect. So like maybe when you get the file initially, that's when it could get hacked. But like every point after that, it's pretty much identified. Not saying that it can't happen, right? I'm sure they'll find a way

Frankie:

if there's a way.

Sal:

Yeah,

Frankie:

I don't know why someone would want. To hack that I don't know.

Sal:

Yeah,

Frankie:

sometimes people just want to watch the

Colleen:

world burn They just want to mess things up

Sal:

Yeah, and the next article is around Apple's partnership with open AI And a little bit around the data concerns that come with that With an AI system kind of embedding on every one of our products slightly scary, but there's a couple of big risks that they identified within this, this article is around data scraping and what open AI has access to around your data. and then how do you opt in or out of that, and regulatory issues. So depending on the GDRP or if you're thinking through, Japan's regulations, are they. Is it incorporating all that, into it? And then also, Sam Altman's, I won't say neglect, but his kind of, as it said, I'll put it exactly the other way around, the criticism that he is facing for not implementing a robust ethical safety guide, for his data. the data is not very secure in that aspect and there might be some. Sensitive data and not putting that forward. So this article brings up a lot of those different points, but.

Colleen:

hopefully they go through some refinement stages or something before it becomes available to the general public.

Frankie:

Yeah. Are you guys going to enable that?

Colleen:

I feel back from my days when I used to have to upgrade like software, for example, we'd wait, cause these software patches for this product I used to support would come out quarterly. We'd always be one or two quarters behind so that other people could work out those bugs and that they could apply patches for them. So I'll probably do the same. And it sounds cool. And maybe it would be useful, but I'm going to probably let other people work out the bugs there first, before I turn on that option for myself.

Sal:

I probably would turn it on mostly.'cause it's okay, it's like the herd mentality or whatever it's called. everybody's doing it, but everybody does it. You don't wanna miss out the odds that I get hacked or Sure. yeah, they could, but like the odds that they actually end up using it, knock on wood, now I'm gonna get hacked. Or It's targeting you now sale? yeah, In pure numbers, like it's going to be hard to pinpoint me and the risk to me. I don't think it would be any different than the risk already.

Colleen:

I mean, there's a lot of truth to that for sure.

Sal:

Cool. being back to our main topic today, around, music and data analytics.

Colleen:

I guess earlier we were talking about our music preferences. So Sal and I are Spotify users. I'll have to say my probably number one played playlist is Chris Stapleton radio.

Sal:

That's a good one. I love Chris Stapleton.

Colleen:

Yep. He's great. I love anything that's bluesy infused rock. That's my, Yeah, he's good. What about you, Sal? What's your biggest played playlist?

Sal:

So the playlist that I typically go is there's two of them. So it's morning commute on Spotify. Okay. and then the other one is chill mix. and they're probably the same thing. It's like Noah Kahn, Jack Johnson, like way back then. Yeah.

Colleen:

folksy. Yeah, really

Sal:

laid back. and Wild River is another band that I like and Lumineers. And so a lot of that. And then my other one is if I'm working in the yard, it would be like, Morgan Wallen.

Colleen:

Morgan Wallen? Okay. Just

Sal:

Morgan Wallen. and. Yeah, that would probably be my main one. And then I also have So no

Colleen:

80s hair bands for you is what you're saying?

Sal:

No, but I'll go even farther back. Like my, my, CCR like playlist. Okay. Yeah. I've got like a 70s

Colleen:

classic rock kind of thing. What about you, Frankie?

Frankie:

Same. I listen to literally all music. Yeah, but

Colleen:

you don't use Spotify, right? I'm not a Spotify user. Maybe a generational thing. I don't know. But what's your number one platform?

Frankie:

So I'm using YouTube music and Purely because shout out to my sister. She added me on her account. Thank you. but they have like super mixes that they, it's built upon music that you like to listen to. And so for an example, I have one that is Taylor Swift, Greta Van Fleet, Miley Cyrus, and Linkin Park. And so that is a wild ride, but I have

Colleen:

themes or like names.

Frankie:

no, they're just like my mix one, my mix two, like they're really lame names. But, I do think like my husband's a Spotify user and I really like their day list. like if I click on his day list right now, it's like Old country, punchy, texas old country afternoon. Okay That paints a picture doesn't it? Yeah for sure. So I just think those are really funny. their day list names really crack me. Yeah

Sal:

Now i'm trying to find it but yeah, like I think I always wonder, like with these algorithms, do I just get pinholed into a certain type of music? Cause I do like to expand my music a lot and it gets us back to our articles and the things that we want to discuss today is like, how does this all be built? How is this being built? Second is like, how are they making money? So one of the main articles that we, and they're breeding looking through it is a Rolling Stone, Rolling Stones article. how big data, is the music industry's goldmine? Which, and then how they're creating it.

Colleen:

This article was really interesting to me for a couple reasons. And one of them, I think we can all think about these playlists that we like. And like I mentioned, I like my Chris Stapleton radio. There have been other artists I've been introduced to because they are similar to Chris Stapleton. So there's that whole side of things. and we can go into more of the details as to how Spotify uses. other information that it's collecting to make those suggestions. But then there's the other side of that too. And like how the, producers and the artists themselves can utilize big data to make changes to either their touring schedule or their promotional schedule and things like that, that I didn't even think of because that's not my world, but it's really cool to think about how they can get smarter about. Maybe where they're going, what locations they're going to.

Sal:

It really makes you feel like in, in my head music or artists are like artists. They just create something and then they like it, put it out, out in the world, and they want to like just share it whenever they get it done with. And now you're like, holy cow, it's actually real business. And like they have to plan it out.

Frankie:

Yeah. There's a lot more thought that goes into it.

Colleen:

they're even talking about how they're. Analyzing data as to what's popular to influence what the artists record going forward.

Sal:

I'm a little nervous on that. do they just now just go to the population? some of these artists are like, Oh, it doesn't make it. It honestly comes in my mind. I'm like Post Malone coming out with his country album, Beyonce coming out with their country album. Oh, is that the hot thing is so like all these artists that were not ever, and they want to expand into these markets. Is it just because of the algorithm is telling him to expand it? Or is it because they actually are like, they like this music, right? And they want to sell it to that music and they think they have, they're Creative mind in that area.

Frankie:

I guess in the end, like it is a business. They're doing what the business needs to do to be profitable,

Colleen:

right? And if it's not going to make money, they're not going to do it. Yeah. Go ahead with your thought. I

Frankie:

think one thing that's interesting is that all of us have, we've been a little bit negative over our podcasts and like talking about how our data is being used. And yes, we're very cautious, but. When we're talking about music, we're all like very optimistic and we love that it recommends new music and it's using our data and we're like, yeah, this is great. So it's funny to just recognizing how much, how much different it is when it comes to our music.

Colleen:

Cause it's a different level of like severity, right? Like my medical records versus what music I'm listening to. Like clearly I can be influenced through music, but

Sal:

I don't think it can hurt me. Exactly. Maybe you

Colleen:

can't, I like, so my mind goes in five different places with this. When we start talking about this and it's is everything going to become so homogenized that every artist is going to be really similar. But then my mind also thinks, okay, but at some point, I don't know, trends over time, ebb and flow. And so you're, You could get this big group of people all releasing country albums or whatever the next big trend is. There's going to be people who buck that. You're going to have the Macklemore's of the world that come out there with their completely independent thing that can't be categorized that then become really popular because people get sick of the norm or whatever the mainstream thing is.

Frankie:

Sure. That's really interesting. Like I didn't think about how much, like they're going to be similar to one another and, And not to

Colleen:

say they're the

Frankie:

same, right?

Sal:

It's there is a pop algorithm. Like a lot of artists and producers have said that, they couldn't make a song to hit the general population, and make, know that it's going to sell.

Colleen:

And I'm sure there's some sort of AI that could put together the right combination of notes that would make it super catchy. But we're not talking about AI. I know we're not talking about AI, but just want to put that out there that I'm sure, like in combination with other things. could you write the perfect song?

Sal:

Yeah, bro. Like I wonder if it would be like especially if ai is doing it Like yeah, how much it would take from older like music nowadays, they sample a lot of other artists right and then they take portions of it or they get they learn from it and it's taking some of those elements, but like How AI is able to do that and then can it come up with something new

Frankie:

this would be like something that I feel Wouldn't use a lot of historical data. Yeah, I feel like you'd want to use your most recent data Available because music changes so fast and what's popular changes Extremely fast as well. And I mean I think about The weather today could impact what music I'm going to listen to. Or my job today could impact my music. Like how your commute

Colleen:

was on the way into work. Yeah.

Frankie:

there's so many things that impact what kind of music you listen to. And I mean, most people listen to a variety of music. I guess, there are some people who are very like particular and like a single type of like genre of music, but for the most part, I think, you There's so much that can impact what you're listening to.

Sal:

Yeah. actually it's funny you say that, but there's an artist, James Brown drummer, Clyde Subbletfield, who's actually, I think he's from Madison, or he lives in Madison or lived in Madison. He

Colleen:

has some connection. Yeah. No,

Sal:

he's lived there for the last, he might've passed away already, but, he's like one of the biggest, sampled artists ever because his drum beats are carried in all hip hop.

Colleen:

Oh, yeah, and so

Sal:

using that in so like you'd be surprised how much music today is actually like Historically used and like they're just sampling it when in turning down or up or pitching up or down

Colleen:

and you see you know There's like artists who will sample artists from the past and then people who will sample them and you know I've seen over the last 20 years This content just get reused and recycled in, in over and over again to the point where is it even the same as the original and do people even realize where that originally came from?

Sal:

So get back to the data topic. There's a couple of kind of cool things that I thought were interesting in this Rolling Stone article. It's around like the data pyramid and there's four levels of data that they were collecting or managing through. So the first level is data, just the raw data, coming from, Facebook, just standard metadata coming from those about individuals that are listening to different music. the informational, section, so it's actually like where they're diving into trends and visualizing those trends. the knowledge, so informing, That knowledge based on that the data, and setting actual benchmarks for like where they think upcoming and mainstream Streams are artists are within those areas, which I thought was cool. I was like, oh now there's like benchmarking whether or not you're a Popular artist or not, like where is that? It's not where is that benchmark and then Finally, intelligence, which is, like predictive outcomes based on, like a confidence score, and trying to get the maximum, impact for release dates and tour markets. I thought that was all really interesting

Colleen:

how they broke that down. Yeah, you just seen that visualization really, painted a great picture for me. oh yeah, these things exist and this is how you break them down.

Frankie:

Yeah, something that kind of stuck out to me in the article was talking about, The ways that they're recommending and personalizing the user experience and they talk about collaborative filtering natural language processing and convolutional neural networks I thought like that part stuck out because I just I think I'm just naturally more drawn to what like the more in depth data so the collaborative filtering was the model analyzes and compares user profiles, listening patterns, and identifies similarities to make recommendations for similar songs. and I thought

Sal:

K means right there. Yeah.

Frankie:

But yeah, I thought that was really interesting. what goes into those models and like, how are they doing that? And they're taking the songs and they're Analyzing, the raw audio file, to see, what's the beats per minute, the loudness, the time signature, and the key, which are all relative to, the musical terms, terminology, and, if you're not a music person, just really analyzing. How the song is written basically

Colleen:

things help you to identify what genre the song is in, right? like your certain time signatures or keys are gonna be more identified as bluesy versus like Pop and I think that all those things go a long way toward Kind of lumping things together making like items.

Sal:

Yeah, it's crazy like These the system really is built To try to figure out what people like right and so as you think through it and you're like, all right This is our brain We're literally like saying all right our brain when we listen to music needs a collaborative filter a natural language processing just so we can understand the words and then a CCN, to understand what kind of music that I what's the pitch, what sounds good in my ears, right? these are all the things that the data scientists that are building these, that are working with these, these music group, or music, what would you call it? Music Platforms? Platforms or Yeah. Streaming services. are thinking through and it's crazy how you can think of Someone's likes like someone's behavior in that way

Colleen:

Very interesting stuff

Frankie:

so imagine If our data around music was connected to any other data about us Like our what about our facebook data or you know any social media or anything online

Colleen:

shopping habits online shopping. Yeah

Frankie:

I wonder like if they had the combination

Sal:

getting all this,

Frankie:

but if they had the combination, I mean, maybe they would know you're going through a breakup or something, like to get coupons from target for chocolate.

Sal:

Wait, we should not, we should sell this.

Frankie:

This is a really good way for free. It is like a really interesting thought though. Like we were talking about how Oh, our music data is not dangerous, but what if it was combined with other data? It could be very powerful. Yeah. I mean, my music says a lot about me.

Colleen:

Yeah. I mean, we immediately went to, could the music affect our mood or could people use music to affect your mood? But the point we're just getting to here is what are the things could you combine with your music listening habits to sell a product or to make your company more efficient or effective at their sales or their marketing?

Sal:

I absolutely think my music could. Paint itself a good picture of who I am.

Colleen:

I really though that your first thought was that you're a swifty. What makes you say that?

Sal:

I'm just kidding. I'm not. I probably,

Colleen:

I thought maybe you meant you were really mainstream or like I figured it would be

Sal:

a good joke. Okay. Okay.

Colleen:

I thought maybe there was a deeper thought there. So

Sal:

I do Taylor Swift though, so yeah, I'm proud of that. teardrops on my guitar back in the day. that's the original.

Frankie:

Okay. Yeah. Yeah. Yeah. Just, I'm more of the folklore Swiftie. Okay.

Sal:

I like that.

Frankie:

I like it all. I'm a fan. Okay. I would, I don't know if I'd call myself a Swiftie, but I do enjoy listening to her music.

Sal:

She did go, she had a song with a Bueno Ver, exhale. I really like that song. It's one of my favorites. Yeah. That's a good song. It's on rotation,

Colleen:

regular rotation. Yeah.

Sal:

And I can't tell if it's cause he's.

Frankie:

You like Bonnie Ver?

Sal:

Yeah. He is just so good. He's

Frankie:

great. Yeah. Yeah. ringing it back in. I think we can move on to the next article, which is from Yellow Brick. and it's unleashing the power of data analytics in the music industry. And so I felt like this article touched more on how a music company can use data analytics. and I thought it was really interesting, like how. And from a company's perspective they're looking to make money right and they they want to do whatever they need to do To make money and so they're looking to identify, you know who's an emerging artist and is that an artist that they can sign to their label if they're a label and or is it an artist that they can bring to the table more like if you're Spotify or youtube music can you add them to people's playlists more because you think they're going to be big

Colleen:

Somehow work the algorithm to have them suggest it more often than not. Yeah. That's interesting.

Sal:

Like with TikTok artists that have come up, right? Like people that have created this things and they use TikTok as their medium to get the music out. But, I think this is a really big thing is like they've identified ways to identify. Trendy top or trendy artists or, up and coming artists in from tick tock or Instagram and using it. And then like, all right, we can sign them, sign a deal with them. And then, and then it gets on the radio. It's just crazy that you can just

Colleen:

look here and it's Oh,

Sal:

wow.

Colleen:

I read an article and I wish I would have thought to share it with you guys, prior to this, but it was about Rihanna and how smart some of her recent tours were, she toured in China and I forget the name of the city that she visited, but they're known for this like street food. That's essentially a savory pancake. Please forgive me, everybody out there listening. If I've gotten some of that wrong, but the gist of it is her tour incorporated those things, knowing that she was going to this area where the street food was popular. They literally filmed her. Making this street food and that she was like working that into her show and working that into her visit to that area. And it was really popular and very well received. And I thought that's probably a prime example. Her producers and her, the, meet the, whatever record label supports her probably knew all of that and knew, Hey, Rihanna is big here. So we're going to go to this area. And also this street food is really big. And if we can combine them together, even bigger. So I just thought that was really super interesting. Like, how else would Rihanna have ever known about this particular element that's unique to that region of China?

Sal:

Yeah, it's just

Colleen:

cool.

Sal:

It's amazing. Cause they start to use data analytics for so many different things. an artist is not just the music is marketing. It is branding. It is, Yeah, like of clothing lines and all that and all that stuff needs to go into these algorithms and these analytics to build a full Suite for an artist to give to their fans I think it's really

Colleen:

and it's interesting to look at that and like what percentage of that is that person's talent and how much of It is all those other aspects around it, right?

Frankie:

do they just have a really great business team? Yeah

Sal:

being in analytics. Did you even think that there was music and like data analytics for music like Mentally, I just didn't even think that way.

Frankie:

No. And when we were talking about doing this topic, I was like, what are we going to talk about? we can't just make that a topic.

Colleen:

I think we can make anything a topic. Yeah. We can talk about everything. Sit down and start talking about it. We probably have interesting things to say.

Frankie:

So something like that's not a part of this article, but triggered a thought was, when people are creating like ads and, working on marketing. They pick like certain music based upon probably statistics as well. we don't have an article about it or anything, but I thought that would be like, that's interesting too. Like I think about commercials that I see where it's either a catchy song, or a annoying song or, like a very positive song or a very like negative one.

Sal:

Yeah. That is actually like thinking I might be wrong, but like thinking through it and like thinking. Like watching TV. I don't remember them like a lot of companies having jingles anymore. Now it's just like playing like clips of like popular songs or like things that you've seen on Like TikTok or Instagram that you song that was created there and they're using it. That's I honestly think it's shifted that way. I mean, I might be wrong. And again, there's something You

Colleen:

know, there's a couple that come to mind but they're

Sal:

way older I guarantee it. Oh, yeah. Yeah.

Frankie:

All right Radio the radio. I hear a lot more jingles from smaller companies or local companies but Yeah, and I think about the chevy commercials Always have super catchy songs. Yeah. Yeah. I can never get them out of my head.

Sal:

Like literally my head and thinking like imagine dragon sale.

Colleen:

That's funny. So my brain went back to the, like a rock commercials. Do you remember those? They're a little older, I'm just gonna say a little.

Sal:

One thing I did like in this is like NYU, is building a kind of career and education program for data music or data analytics and music. I thought that was interesting. I was wondering if like, I know Madison, when I went there didn't have it, they might have it now, but, if they end up starting to build out like in their music programs, data analytics,

Frankie:

that's interesting. And is that the direction that data analytics is moving? Is it going to be very industry specific?

Colleen:

I, because I do feel to a certain extent, it's like 20 years ago, technology became a part of everything, right? You couldn't say I'm not a technical person anymore. And I wonder if we're going to get that way with data analytics. Is data going to seep into every aspect? like you said, if you've got, if you're a music major, is there going to be then courses on music analytics? to, to work in that angle for the people who maybe don't create the music, but maybe we'll produce it or direct it.

Sal:

I mean, just being in this industry, like you, I feel like the most valuable people are the people that know the business. Like we were just talking

Colleen:

about the Celtics, right?

Sal:

Exactly. They know the business, but they can do the analytics,

Colleen:

do the analytics. Yeah.

Sal:

Yeah. I do think that I think there's going to be a move towards being specialized in a certain kind of. Vertical or whatever. Yeah. and then having the analytic skills and like you can learn to Python and use Python or Tableau or visualize things, right? and pull data like, but to know about the industry is way more important, I think.

Frankie:

Sure. so cycling back into like less data nerdy things that we can talk about. Do we want to move on to the next one? Yeah, sure. so we also read a an interesting medium article that was Talking about the inner workings of spotify AI powered music recommendations and how spotify shapes your playlist And so this one is a little bit more in depth. but there's some really interesting like metrics and things that they're looking at that they point out and So they first just touch on You how Spotify has evolved as a company and I thought this was really interesting that Spotify's number of active users has continued to increase quarter over quarter every quarter except for one since the beginning of 2015 Yeah,

Sal:

that is

Frankie:

it's a weird little outlier there. It's mind blowing in it. I think it was before the

Colleen:

pandemic.

Frankie:

Yeah it shows though like just how much the music industry is evolving and like how Data and spotify being known as one of those providers that makes really good recommendations and puts together great playlists like I think it's definitely impacting their business in a very positive way and then they start to go into the things that are more in depth like the collaborative filtering and the content based filtering and so Let's start diving into those the collaborative filtering is Talking about how Spotify connects the tracks and the users so that everything has a relationship. And so I just was thinking about a couple of examples that I can think of that are like, when a user is frequently putting a song in a playlist and then other users are doing that same exact thing and putting a song in a playlist. Then you know that song might be Closely related to the other songs in that playlist and then they start to map those together and create a music map

Sal:

so this is my recommended or my Recently played versus my recommended stations. Okay, that kind of ties into this. Yeah, so chill mix, which has Wild River if you ever heard of them We I have morgan wallen Noah khan post malone folk artist mix and sam hunt. Okay. Yeah, my recommended stations are camp head and heart which are both like mix between country and and I don't even call it maybe

Colleen:

pop kind of Or folk?

Sal:

Folk. Yeah. Folk is the word. Mumford and Sons, Death Cab for Cutie, Old Dominion, and, Teddy Swims. So they're all very similar, even though I probably haven't, I don't really listen to Death Cab, that often, but you could see like, all right, that's not that far of a dotted line.

Frankie:

Yeah, sure. And I think of it as If we're thinking about like the map of the United States, for example, like we start to cluster things together. Like the Midwest for example, is a cluster. and we like the tendencies of Midwesterners are all grouped together. And I think people kind of think of Wisconsin and Minnesota and Canada also as a cluster like. The way that we communicate and like the culture so thinking about our music in that kind of way All the music is going to relate somehow and like it will be it'll fall into multiple clusters. And so there's so many different things that they look at with the behavior the patterns the metadata and like They'll look at the description of a song and utilize that Language in on Natural language processing model.

Sal:

Yeah. So it says it looks at tempo, genre, mood, and so many more variables.

Colleen:

You probably even look at what time of day you're listening to that music and where you are and look at the weather.

Sal:

And I think like music is math in some aspects. It has a lot of math elements to it. and so like building out like those characteristics of probably actually not B2. Challenging because it already is structured in a way that you can work with. Yeah

Frankie:

Yeah, I think it's interesting too how they look at Users and they compare users to one another and map users. I

Colleen:

mean, that's really smart

Frankie:

I

Colleen:

think we saw a graphic on one of these articles We read that was like I like songs a b and c and the other person's like I like b d and e I recommend to you song a and I recommend to the other guy song d because clearly there are probably similar elements

Frankie:

Exactly. So that's very interesting. And then they also have a more in depth version, which is called content based filtering. And so that also dives into metadata, but it's at a deeper level and then it dives into the raw audio analysis and cultural context. So I thought the cultural context part was really interesting because, it's I mean you think about like all across the world. Everybody listens to different music and There is some component of culture in that and like music culture is a really strong thing, right?

Colleen:

a lot of people identify with their culture based on the music that they listen to

Frankie:

sure and then I also think like people, maybe your friend group is partially based upon the music you listen to. Yeah and I just think about like You know, there's certain friends that I would go to particular concerts with and then others where I'd never, ever invite them to that concert. I don't know. I'm thinking like Taylor Swift. I would never take my husband to a Taylor Swift concert. I bet you he'd

Colleen:

go

Frankie:

with you though. He's a good guy, but I don't know if he's that good. He might not have a friend to go with, but yeah. But yeah, it's just a very interesting piece of this. And.

Colleen:

I wonder what a map would look like, even if you just looked at the United States and you mapped out the percentage of users, whose number one genre on even just pick one of the streaming services is hip hop versus country versus whatever, what those demographics, what that would look like and how the bubbles would shift, like to show the, or is it more homogenous than we thought?

Sal:

I also wonder if, what if you moved? Okay. Do you like automatically pick up those regional, let's say the South or something, or into a multicultural,

Colleen:

do you ever go on a road trip and you're in a different area and different radio stations come in and you find yourself listening to something that you maybe wouldn't have at home. I did that this weekend.

Sal:

So speaking of, this is actually probably on point, is. So my, my son goes to a Spanish immersion school, right? And so my Spanish is not great. and he's going

Colleen:

to help you get it better. He's going to talk to you in Spanish. He's

Sal:

going to make me so much better, but at home I listen to a lot of Spanish music cause I like to have it and have him dance to it. But I also ask echo all the time. Like, how do you, like, how do you say this word? How you say this word. And now I get Spanish, iTunes and Spanish, Spotify all the time. So like it's telling you,

Colleen:

yeah,

Frankie:

that's

Colleen:

interesting. Can we also talk about the genius that is the Spotify wrapped? I know it's a little bit tangential to our data, but I feel like it's straight up analysis of your, at the end of the year. Yeah. and the fact that everybody shares that. I don't know if they could have predicted that it was going to be that popular, that people would share that, but that's straight up a data analysis of what you've listened to in the last year. And I think people are sometimes surprised by what shows up in that list and probably does affect or influence their listening habits going forward. And I think too, it's cool. I like to see friends. Spotify wrapped lists, because I usually find new music or things that I would have maybe not come across otherwise and their Spotify wrapped lists. So they tapped into something really cool there. And maybe it's just me that I love.

Sal:

I love data,

Colleen:

but I, you see so many people share it. You, you've got to think there's something there.

Sal:

And I think there's been like a competition built with it now. Like people like compete Oh, I'm the number one Taylor Swift.

Colleen:

Oh, that way. Sure. Wow. You're in the top 10 listeners. Thank you.

Sal:

A lot of people will be like, Oh, they want to

Colleen:

2024. Yeah.

Sal:

And so they'll listen to an artist over and over just for that reason. That's so funny. And just imagine the time and the ads that person has listened to. Sure. They have premium. Yeah. Oh yeah, that's true.

Colleen:

Yeah.

Sal:

So thinking through this, is there a way to beat the algorithm? what if we want to change it? But what does that

Colleen:

mean? Just to change it so that it recommends, Yeah,

Sal:

what if I don't want to be what you have put category you put me in, like, how do I find other things? yeah, it's like really this kind of maybe brings us to our next article, understanding a complete guide of the Spotify recommendation algorithm. this article talks about How Spotify in this case will recommend music and discover music. And it actually dives into, generating track content, base or the music content based on track titles, release dates, art, artists, name, different credentials that have different tagging. So what type of mood it might be tagged to styles, and then it builds in these algorithms. I don't know if it's exactly how you can hack the system, but in a way, if you're starting to think through Hey, these are the things that build these algorithms, how do I maybe throw in a wrench every once in a while to see something different, or maybe something with different energy levels or danceability or tone? How do I build that in, into my listening so that I can keep my music expanded?

Colleen:

Yeah. That's a really good point because if your goal is to just even Keep your mind open and you want to know what's I think they do have a what's new feature, right? I think most platforms do it'd be cool too If they just had a spin the wheel and like it just lands on something random and like awesome Yeah, all of a sudden now you're listening to like some folk music from some european country Like in

Sal:

college, I remember my brother. He studied in spain for a year But in europe they had different music, right? yeah, it wasn't like spanish i'm not saying but they had like their artists that were really good like the cooks, right? I didn't hear I heard the cooks well before they came to us and stuff like that but like The reason I liked them just because my brother was sharing with him. I was like, how do I get that? I don't know what artists are out there You don't know what you're missing in my own comfort level Yeah,

Colleen:

yeah,

Frankie:

go ahead I bet like the first thing that I would do if I was trying to do that was Switch up the language. Find music in a different language. You mean

Colleen:

the setting on the app itself?

Frankie:

No, I just think like finding, like start searching for music with French? Maybe Spanish music. Throat singing? but I think that would really throw off the algorithm and start to be like, it would probably be like, Oh, they're listening to different languages. Like maybe we should throw them some different songs and in this language. I

Colleen:

wonder if there's some like music of the world playlist or something that would combine there's gotta be like the number one hits or something from. There's a place for everything. And if there's not, let's create it.

Sal:

I know Apple music or iTunes has it, but I'm actually looking through this, the Spotify. I don't think Spotify has that, but yeah, they have like music. Yeah. Like music of the world and you can do I'm

Colleen:

going to hope that this doesn't start playing on me when I click on it. it is not very, I thought maybe I'd click on it and would have the countries labeled. They do not. some of them look French. Some look like African languages.

Sal:

now because you just clicked on that, you're yeah, I just, I'm looking at the playlist. That's cool. That's

Colleen:

fine. I'm for it. but some of the stuff is probably really interesting. And it'd be interesting to see, okay, we've got this playlist now. How many plays have these songs gotten? Like how popular are these in their countries? Or is this just a completely random list?

Frankie:

What about things like, like the type of song that it is? is it, does it have thoughtful lyrics or, does it tell a story, anything like that? do you guys listen to certain music?

Sal:

Yeah, I was actually gonna say with that, do you, do you listen to the music or do you listen to the lyrics?

Frankie:

I think for me, I'm listening to the music part. Yeah. Yeah.

Colleen:

I'd have to say the same.

Sal:

Yeah. I usually listen to music, but I know a lot of people like, they like the ri like the lyric part of it and the music is the second part. Yeah. Which like,

Colleen:

so they're more poetry lover lovers maybe, or, yeah,

Sal:

exactly. And so like, when you're thinking through this, like how do you. Like these models building into those elements of like where you listen to it. How you like what? What area of the song do you listen to?

Frankie:

Those are the people who can listen to a song one time and then they're singing it the second time And i'm just like how? And i'll be messing up the words and i've heard this song 70 000 times bathroom on the right exactly but yeah, and then The other like question that Spotify looks at that I thought was interesting is does the user like songs you can dance to and are they analyzing Like your behavior when listening to those songs. that's an interesting thought like oh, is this person having a party or is this person? Studying or like all these different things that you could be doing I wonder, do you think they

Colleen:

have any sort of data around how people are listening to this, whether they're listening, like through the app? Or whether they're listening through like headphones or on a car radio. do you think, cause that could tell you a lot of information about where people are listening and maybe what they're doing.

Sal:

I don't have access into the Spotify one, but based on this podcast, we can actually see every, all of our listeners. So anybody that's listening to it, we can actually see, we can't see you as an individual, but we can see where you're listening to it. So like actually to the city. So if you're listening in Waukesha, you're You can, we'll see that. we can also then see Oh, they're listening to it on their iPad or they're listening to it on their iPhone on Spotify or iTunes, right? Like it could tell us exactly all that information so that we can better understand. Our listeners But like it's crazy. I would imagine spotify can because people who

Colleen:

listen to again time of day if it's 10 p. m Where you are and you're listening on like airpods You're probably going to sleep or you're trying to wind down from your day versus if it's a speaker in your home And it's a certain type of music probably indicates you're having a party

Frankie:

Yeah, that's a really good thought

Colleen:

interesting stuff

Frankie:

Alright, that's a wrap on today's episode around how data is enhancing the music industry. The purpose of this discussion is to demonstrate just a really cool way that data is being utilized and help you understand where your recommended music is coming from. Thank you again to our sponsor, Continuous Technologies, for providing us use of their space and technology. If you loved today's episode, make sure to subscribe to stay up to date on other topics related to data. Next time we'll be talking about beer data. Looking forward to that one too. Thank you for listening to Cream City Calculations. Until next time, keep calculating.