For those who don't know, Paul Lamere, the author, is a really cool guy that works over at The Echo Nest (recently acquired by Spotify). I had the pleasure of meeting him and chatting with him a few times last summer while I was interning there and he's basically a music+tech genius. Lots of the echo nests coolest api demos are all his doing. I'd link to stuff but I'm on my phone, but definitely browse through his blog. If this is your area of interest, Paul is the guy to read.
“It’s not enough to have one hook anymore,” Jay Brown, the president of Roc Nation, and Dean’s manager, told me recently. “You’ve got to have a hook in the intro, a hook in the pre-chorus, a hook in the chorus, and a hook in the bridge.” The reason, he explained, is that “people on average give a song seven seconds on the radio before they change the channel, and you got to hook them.”
>> "I wonder how this affects the writing of songs."
It depends on why you're writing the song I guess. If you're writing a pop song you want to get to the top of the charts following that formula is probably essentially (although it obviously doesn't guarantee success). Just listen to some of the top pop songs - they clearly have a formula. The problem is that when everything follows the formula you can't listen to much of it because it gets boring.
I'd say if you're writing a song you want to last, something that will potentially go down as a classic, forget about formulas a write what you think sounds good.
"The problem is that when everything follows the formula you can't listen to much of it because it gets boring."
This is why I can't stand pop music, and can't understand why it's so .... popular. I've already heard the C-major scale in 4/4 time, layered with false emotion by a 20 year old who's seen it all.
That's the exact reason that most pop music is marketed at teenagers. For them, it's the first time they've heard it (relatively speaking). As you get older, you've heard all the 4/4 C-major stuff you can stand. Even if you're not particularly musical, you will still subconciously recognise that pop songs are repetitive and fairly unoriginal after a while. But it's an easy sale for the record labels, so that why they do it.
That doesn't mean that there doesn't occasionally come along a good song or two. It's still perfectly possible to write a 4/4 song in C and have it be interesting/fun/good to listen to. Unfortunately you'll have to wade through a whole load of crap in the pop charts before you find it ;)
Speaking only for myself, I'm a jazz lover and musician, but yet I enjoy a good pop tune when hear it. But I don't listen to anything enough to get sick of it.
On episode 112 of Penn's Sunday School [1] Jay Frank talked a bit about this - that the intro to songs are getting shorter because people can skip easier.
A fellow congregation member! That discussion was the first thing I thought about when I saw this post. It's worth a listen if you're interested in this stuff.
A hook-heavy attitude toward composition is a style of writing songs in itself. However, we can't extrapolate to the graphs presented because Mr. Brown's attitude assumes a predictable reaction by the audience, as well as the graphs assuming a constant quality of music presented.
I think the graphs would be more interesting if they weighted the values according to e.g. the number of playlists a skipping user listens to regularly (or not, if they are a listener who seeks out new combinations/playlists).
Another analysis I'd like to see is how many songs people skip in a row. When I skip a song I end up skipping a bunch of songs until I find just the right one that I want to listen to at that moment, and I speculate that most of the quick skips come in these sorts of runs. If that's true, then it'd be interesting to try and figure out what properties the songs users end up on have as compared to the ones that they skip over.
At first I was surprised skip rate was so high, because I couldn't imagine somebody using a service where they skip every other song, but your observation explains this nicely. Even for my carefully selected music library, I sometimes burst-skip a bunch of songs, which would undoubtedly taint my skip statistics.
Very tangential but I used to skip songs a lot and make a lot of custom playlists but have moved towards listening to albums front-to-back and not skipping songs. I still have a couple small playlists that are a sort of "greatest hits" of my collection that I shuffle every once in a while.
It has greatly improved my listening experiences in general, and has made me discover songs I forgot I had/ I didn't listen to much before.
Interestingly, Last.fm introducing a skip limit of 6, with no way to remove the limit even for paying subscribers was what finally pushed me away from Last.fm and on to Spotify.
Especially with electronic music becoming so popular. The last 30 seconds of most EDM tracks is just a repeating beat to enable a DJ to mix it into the next track easier.
I think the constant non-zero skip rate throughout the entire song length is explained by people who would have skipped the song at the beginning but who weren't present when the song started (e.g. they went to the kitchen to get a drink). These people would skip the song as soon as they get back, which could be at any point within the song.
Paul works at The Echo Nest [0], which was acquired by Spotify. The Echo Nest has an API for music data but I'm not sure if skips are part of the public one.
I'd be interested in how often skipping is used on new songs vs ones we've heard many times before. Is it possible that a song can only become popular if it hooks the listener in the 1st 10-20 seconds? Important for songwriters to know.
Skipping is not a new 'iPhone' phenomenon. CD player in your car has a next-track button. Even the car radio - changing the channel was common. So no, music today isn't getting some sort of 'raw deal', or at least no more than ever.
> "get out of your chair, walk over to the turntable, carefully pick up the tone arm and advance the needle to the next track. That was a lot of work "
One possible explanation for why user's skip behavior varies by time of day/week is that they're more likely to be playing songs around friends. A lot of times you'll want to show just one part of a song to someone, or you want to manage several requests for songs when you're out.
Going a bit further... I wonder if anybody has thought of using the skip rate as a way of A/B testing pop songs, for instance similar arrangements of the same song.
edit: on my comp now, so I can add this link to some cool EN API demos, many of which are Paul's if I recall correctly: http://static.echonest.com/labs/demo.html