Click here for more info on this project.
Groups of Four
Four is a pretty common number in music and in life in general. There are four cardinal directions, there were four Beatles, I bet the building you’re sitting in has four sides, the most common time signature in popular music is 4/4 (it’s even called common time), there are four horsemen of the apocalypse, but my favorite is that there are four orchestral string instruments. If we’re going to nitpick, the sections of strings don’t line up exactly with the different string instruments, the four sections are Violin I, Violin II, Viola, Cello + Bass. But since we’re all about breaking conventions here, I’m going to divide them into Violin, Viola, Cello, and Bass for this sonification, since they sound the most distinct from each other.
Let’s divide up the US into four groups so we can get each group to play its own string instrument.
This might look like a bit of an odd way to divide up the US, but chopping it up like this gives us four relatively similar groups in terms of total Postmates deliveries in the course of a week. We’ve got four regions, and we’re going to assign a string instrument to each one.
- Southern California: Violin
- West: Viola
- Central & South: Cello
- East: Bass
Now that we’ve got our performers, what are they going to play? I wanted the behaviors of each region to really stand on their own, so I want the data to inform when they’ll be playing, what they’ll be playing, and how they’ll be playing it.
The Score
Let’s start with what notes they’ll be playing, and the timing of those notes. Since the groups are relatively close in size, I decided to have each group play a note each time there are an additional X number of deliveries (that number is held constant between all groups, but I won’t say what that number is in case there are any analysts out there that have really good ears). So that will take care of the timing of the notes.
The pitches are determined by the average distance that Postmates travel between pickup and dropoff of those X deliveries. With that, we can make a “score”.
Obviously a conductor isn’t going to have a great time if this is what you put in front of them, but you get the idea. The higher the average distance for a given X deliveries in a region, the higher the note will be. You can see the different sections have different ranges, and the data has been scaled to the min/max of each instrument’s specific range. In this case, I decided not to limit the notes to be in a particular scale.
Now that we know what and when the instruments will be playing, now we need to find out how they’ll be played.
Articulation
How hard each note is struck is determined by the average price paid during those X deliveries. If you think about how string instruments are played, the attack is just one element of the dynamics. String players (and my sample library) can control the dynamics of the sustained note after the note is struck, and in this case it’s determined by the total number of deliveries bucketed by minute.
Even this is still not the whole picture of how string players can use their instruments. There are many different types of articulations that can be played to create different effects. Tremolo (playing rapidly back-and-forth with the bow), spiccato (the bow bounces lightly off the strings), col legno (the strings are played with the wooden part of the bow), the list goes on. The sample library I’m using has even more complex articulations than this, and I’ll primarily be using the more colorful ones.
These different articulations don’t really make sense as a scale, they all have very different characteristics and are used for different types of color in orchestral music. I figured that’s actually a pretty good parallel for different food categories. Mexican and Italian food aren’t really on a scale, they’re just different groups that have different flavors and cultural history. I grouped all of our categories into seven different parent categories (in alphabetical order).
- American
- Asian
- Coffee
- European
- Fast Food
- Grocery
- Latin American
Each of those groups are tied to a specific articulation. They differ slightly between the instruments to take advantage of that particular instrument, so they won’t be exactly the same between the different regions.
Let’s zoom in on just Southern California, and we can see what this articulation data looks like:
In this chart, the attack of the note is represented by the size of the circle, the color represents the category of that X numbered delivery (and the articulation of how the note is played), the X axis is time, and the Y axis is pitch (average distance).
Tracks
I know, this is a lot all going on at once. I’m not going to lie either, it’s hard to hear it all going on at the same time, with each instrument is playing its own atonal lines overlapping with each other.
So, if you really want to dig in here, here are the individual tracks all isolated so you can hear one at a time.
What You’re Hearing
This is probably my favorite one of these sonifications, but I’m a huge fan of unenjoyable music. There’s so much to unpack here, it’s almost hard to really focus on one thing.
One thing that stands out, is you can hear that the notes change more rapidly at some points, and hold steady for others. And it makes sense, they change rapidly at our lunch/dinner peaks, and the hold generally pretty steady late at night through early in the morning. The individual notes are a bit harder to focus on between instruments, but you can still feel that overall motion happening between the instruments. After the long sustained notes, you generally hear the lower instruments start moving again before you hear the higher instruments doing it, which makes sense since most of the Eastern Time is in the basses, and the Eastern and Central time zones are represented in the cellos.
How it Works
I made a simplified tutorial on how to create these from start to finish. Check out the Colab notebook if you want to make one yourself.
Since there’s so much going on in this one, I made three different MIDI files per region, so 12 total. Here are the three different files:
Pitches and attack values
- Avg distance and avg price for every Xth delivery
- Note on/Note off values
Dynamics
- Total number of jobs grouped by minute
- Control Change (CC) values only
Articulations
- Category of Xth job
- Note on/Note off values
- This one is unique in that articulations are switched by hitting a specific range of keys on the keyboard
- So each category needs to be tied to a specific note, and those notes need to be played BEFORE the actual note is struck so that it has time to switch to that articulation
Multiply that by four to get everything.
From there, I loaded it into my trusty DAW, REAPER. And this time I used one of my favorite sample libraries from Spitfire Audio, the London Contemporary Orchestra. It has non-standard articulations and playing styles that you really can’t get with sample libraries ordinarily.
Conclusion
This one is both the least useful, and the most interesting piece that I’ve made in my opinion. It sounds to me like a serialist piece, but it’s not guided by the same rules at all. It still does stick to very strict rules, as any visualization/sonification should, so I feel like it counts.
Aspects of this piece feel very human, the ebb and flow of the notes changing, the way the articulations flow into each other, the way the different regions seem to move together and pause together. But at the same time, you don’t walk away humming a melody or anything like that. That’s ok though, music doesn’t need to do that.
Postmates is always looking for creative data-focused people to join our team. If you want to make things like this, check out https://careers.postmates.com/ and say that Alex sent you.
