Numlock Sunday: Chris Dalla Riva on what Phish and Taylor Swift have in common
By Walt Hickey
Welcome to the Numlock Sunday edition.
This week, I spoke to Chris Dalla Riva, who wrote Swifties vs. Deadheads: A Meditation on Live Music for his newsletter Can't Get Much Higher . Here's what I wrote about it:
Swifties love to find out which special songs Taylor Swift will play at a given show on the Eras Tour, the songs that deviate from the main setlist and serve as a nice little surprise. In terms of unique songs performed at concerts — based on an analysis of 113,565 song performances at 5,029 concerts by 98 artists from Setlist.fm — she’s not too shabby. Swift played 66 unique songs at fewer than 20 percent of her shows on the Eras Tour, eighth place among tours, averaging a consistent 2.0 unique songs per concert. A few acts dominate these metrics: Phish’s Summer Tour 2022 averaged 7.58 special songs per night and 197 different songs played at fewer than 20 percent of shows. Dead & Company, Dave Matthews Band and Pearl Jam also crush it, placing Taylor Swift in the company of a bunch of jam bands.
Chris is doing some of my favorite pop culture data journalism right now, and I’m incredibly fond of his music newsletter. We talked about some of his recent hits.
Dalla Riva can be found at his newsletter, Can't Get Much Higher.
This interview has been condensed and edited.
You have a newsletter that I am very, very into. You started it almost a year ago.
It was October last year, so we're closing in on a year.
You're doing some of the best music data journalism out there right now, and we'll get into that in a little bit. But I guess I just wanted to take a little step back and ask, what drew you to music and data journalism?
First, I am a musician. I played music for a long time now, played in bands, write my own stuff, play my own stuff. But I've always had a twin interest in analytics, studied math and economics in college, and I ended up working for a music streaming service called Audiomack. I do, very broadly, data analytics for them. It's just a nice way to combine these two interests that I've always had with music and data and analytics.
The first story that I wanted to talk about was you investigated a mystery that was basically, what is it that Taylor Swift and the Grateful Dead truly have the most in common?
There was an article that struck me as strange from Fast Company, and it was called “Why Taylor Swift Is the New Grateful Dead.” Which, when I see something like that I'm like, "Okay, that is a bizarre claim," but I figured worth investigating to see if there's any overlap. The article made a couple nice points, but the one comparison that I found most odd in the article was that their live shows were similar, because the claim was that Taylor Swift, every night of her current Eras tour, most of the show is highly choreographed the same every night, but she always plays two surprise songs. The author was claiming that this was similar to how the Grateful Dead changes up their set list basically every night, which I honestly thought was sort of a ridiculous claim.
To investigate this, I went on Setlist.fm, which is just a publicly maintained repository of set lists. I pulled down set lists from a few hundred tours, which amounted to, let's see, over 100,000 song performances. I just tried to look at, across all of these tours — of course I included the Grateful Dead and Taylor Swift among many others — how do we measure set list uniqueness?
The first thing I looked at was just total songs performed across the tour and then unique songs performed. This methodology was pretty good. At the top of the list you see Phish, you see the Grateful Dead, Springsteen, Dave Matthews, the people you expect, Pearl Jam.
But no Taylor Swift; she was much further down the list.
My girlfriend, who is a Swiftie, complained that this methodology wasn't that great because Taylor Swift, her set list is so long that it very quickly will drown out any uniqueness, because she's playing like 40 songs every night.
Sure.
If you play two unique songs, those are going to get washed out. So I came up with a secondary methodology, which I liked a little bit better, which was, on the average night of a tour you go to, how many songs are going to be performed for you that are only performed at less than 20 percent of shows?
I called these “special songs.” So when you went out to see Phish on their summer 2022 tour, how many songs were they going to play that they played at less than 20 percent of shows on that tour? And what you actually see here is that Taylor Swift and the Grateful Dead are closer than you would expect.
Phish is at the top here. On an average night you're getting around eight songs that they perform at less than 20 percent of shows on that tour. Grateful Dead, it's like between four and five, and Taylor Swift, it's two. The conclusion really here was that for most tours, there's no variety. The show is what it is. So Taylor Swift injecting two songs in every night actually makes her a bit of an outlier. Now, she is not as much of an outlier as the jam bands of the world, your Phishes, Pearl Jams, and Dead & Company. But the comparison wasn't as far-fetched as I originally thought.
I understand exactly why each of these artists played a couple rarities each show, and it's different for each. You have bands like Phish and Dead & Company, which are jam bands. You have folks like Billy Joel, who has just got such a deep catalog that he can't just play the same hits every month. Then Taylor Swift, who does go out of her way to play at least two things special each time, pretty much on the nose.
To me the most interesting thing really was that on your average night going to a concert, you're seeing the same show as everybody else on the tour for almost every single concert that I pulled into this, and it was like a few hundred tours.
So even if you play two unique songs, it sets you apart. You're not a jam band; Taylor Swift is probably, at least on this tour, never going to give you the surprises that Phish provides every night. But it's still enough of a surprise that it stands out from most of the rest.
You actually had another story a few months ago that I also really, really dug that is less about the live performance of songs and more about how albums are actually constructed. Do you want to talk a little bit about that project?
I've had this debate with a few friends in my life where we always look at an album and I'm always like, "Why is track three always the best song, always the most popular song?"
I talked to other people like, "Oh no, it's always track one," or it's always near the end, blah, blah, blah. So I was like, "All right, this is something that's super measurable." I grabbed the top 50 most popular albums on Spotify of all time, and it was track four, closely followed by track three, that’s typically the most popular song.
But like I said, I work for Audiomack, which is a popular streaming service. We have tens of millions of monthly users around the world. I have access to a much larger repository of data there. Fifty super popular albums might not be illustrative of the more general trends.
I went and used Audiomack's data and pulled the millions of albums that are on the platform. What was interesting is that where the most popular song on an album is really depends on how popular the artist is.
When I look at less popular artists, it was typically track three. Then for artists who were getting 100,000 to 1 million streams per album, it shifts to track one. But then when you get to the most popular artists, it's more like track two or track six. There was less of a pattern.
What I found with this is that for the least popular and the most popular artists, the most popular track is typically whatever the single is. For artists who aren't that popular, artists getting a few thousand streams on an album, for whatever reason, track three is most often the single.
For the most popular artists, it's all over the place where their single is, but the single tends to be the most popular track.
This makes sense. If you're not a popular artist, a random person coming to listen to you who has no familiarity, they probably click play on the single. For the most popular artists, these are bona fide hits, so of course the singles are going to get more plays.
What's interesting is that there's this middle range of popularity, people getting, like I said, like 100,000 to 1 million streams per album. These people have fan bases, but they're not superstars. And most often track one is the most popular. For this one sliver of artists, it's actually when people listen to their albums, they're just clicking track one and listening. Maybe they're not listening straight through, but that's the place they're starting.
I find this interesting because it's telling you that for artists of that caliber, these artists are popular enough where they have some name recognition, but they're not pop stars. In this middle class of artists, people have some familiarity with them, and it seems like the way people listen to those albums is less singles-driven. They're most likely just to start with track one. Maybe because they might not listen to the whole album, track one is necessarily advantaged over the others. So pretty interesting.
The other cool takeaway I thought from this was that no matter what, it's usually a song that's in the earlier half of an album. Because again, those tracks are advantaged. If you're going to listen to an album, you probably start with track one, and if you don't make it all the way through, track one has an advantage over track 15 or what have you.
I noticed that with the bands that I follow, where it's just like the folks who really kicked off with a great single have a very different look about their view counts than the folks who are very album-based rock.
Yeah, exactly. It's pretty interesting. Just the way we listen to artists seems to be different based on how popular the artist is.
Again, I really like the way that you go about your newsletter. Questions that you have, arguments that you have, and then you actually have the means to settle them, which I think is the best way to discover things about data. I'm a big fan of it. Any articles in particular that we haven't mentioned yet that you'd like to call out as one that you learned something cool in?
My favorite things, like I said, I play music and I do data, and as a musician I'm like, a pop song can save your life. In terms of data analytics, I'm skeptical of that notion. So there's this tension that I find underpins a lot of my work. I like to just look over a big historical timescale at a lot of things. Because many things that seem common, if you look back maybe 50 years, they weren't common, things that seem natural.
One recently that I liked was an article called “Pop Songs Don't Need to Have a Chorus.” This just looked at this idea of when we think of pop music today, we just assume the structure is the song's verse, chorus, maybe a bridge.
But if you go back 50 years, that structure was not that common. If you think of a song like “Somewhere Over the Rainbow” or any of those old Broadway, Tin Pan Alley songs, it's a different song structure. I'm able to track that from a different project I work through, where I listen to every Billboard Hot 100 number one hit and I tag the song structure of each song.
Oh, cool.
Yeah, I actually use that data set for a lot of things I work on. But that just allows me to see that again, things that feel natural or things that feel like they are inherently true, it's not always the case, that things change over time. Working on that particular article gave me an appreciation for different types of song forms and just this idea that even if everything, surely pop songs, seem to always be one way today, in 50 years it'll probably be different than it is now.
Yeah, that's the thing about music, right? Popularity is never assured in music, and what's popular now is almost guaranteed to not be popular in 15 years, yeah?
Yeah, exactly. That's sort of the fun of it. It's constantly changing and I think it is just, as they say, it's only rock and roll, but people, even when I talk about something that I consider just to be trivial, like what's the most popular song on an album? People get pretty worked up sometimes, and I think often reflect something deeper, because songs are important to people.
Chris, where can folks find you? Where can they find the work?
You can find me basically all over the place. My handle is usually @CDallaRivaMusic. I post most regularly on Substack; my newsletter's called Can't Get Much Higher. The tagline is, the intersection of music and data.
Though branching off and doing some interviews, too; just talked with Billy Steinberg, who's a songwriter, who wrote “Like a Virgin” and a ton of other hits. That's one thing I'm trying. TikTok is the other place I post. I have 100,000 followers over there and I try to post regularly. The newsletter's the real passion, I would say. I prefer the written word. So either of those places.
If you have anything you’d like to see in this Sunday special, shoot me an email. Comment below! Thanks for reading, and thanks so much for supporting Numlock.
Thank you so much for becoming a paid subscriber!
Send links to me on Twitter at @WaltHickey or email me with numbers, tips or feedback at walt@numlock.news.