Demand — Growth

February 3, 2020

Share on FacebookShare on TwitterShare on LinkedIn

Demand — Growth

Click here for more info on this project.

Exponential Growth

Postmates has been growing like a rocketship since the first delivery was completed in late 2011. The chart below shows weekly completed deliveries since first launching, not cumulative. This curve can be kind of tricky to really understand, because if we looked at this same chart at the same point in time last year, the shape would be incredibly similar. We start to get used to this shape.

Our brains aren’t very good at understanding the concept of exponential growth, and getting accustomed to charts like this are why this chart doesn’t seem as impactful as it should feel. That’s why I thought it would be a great candidate to turn this into a sonification. The sound you heard was based on the exact same data used to create the above chart.

What You’re Hearing

The way this piece is set up is that every time our total delivery count reaches a new power of 10, a new note is held by a string instrument. The volume at which those instruments are being heard is determined by the weekly deliveries. It’s best experienced with headphones or good speakers, turned up very loudly. No processing was done to the audio that would qualify as “mastering”. The intent here is to let the data create the impact that volume has on the listener.

This brings up an interesting point. Dynamics in music is a bit of a controversial topic. During the 90’s and aughts, there was something going on called the loudness war. It’s worth taking a look at, or a listen to. TL;DR: loud sounds better, everyone tried to make their music the maximum volume as much as possible. This issue has mostly been dealt with thanks to normalization techniques championed by streaming platforms these days. With all of that said, there’s still a massive range in how dynamic certain genres of music can be. If you’ve ever been to performance of orchestral music, you probably heard a range from the softest solo instrument to the loudest that every instrument can play together. That’s why I thought it would be a great match for this particular sonification. The beginning is supposed to feel whisper quiet, and the end is supposed to feel loud. It’ll depend on the listener cranking it up, but it’s worth it.

The useful thing about a sonification is that the listener experiences the entire dataset in the length of time determined by the creator. When trying to illustrate a concept like exponential growth, feeling how long it takes to start becoming noticeable, and then how quickly it becomes overwhelmingly loud really helps demonstrate the power of this curve, so to speak.

Since we just hit 100,000,000 deliveries when I made this sonification, you notice that the very end is punctuated with a huge pizzicato pluck of a double bass section to mark the peak of the crescendo. At the beginning, it’s almost hard to detect the different layers since they come in very close together, but also quietly.

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.

In this case there were two separate datasets that I grabbed, one that had the timestamp of each delivery that reached a new power of 10, and another with weekly total deliveries.

With that, I built two MIDI files, one that creates a note for each time we crossed a power of 10 in cumulative deliveries, and a second one that converts the weekly delivery numbers into CC values that will control the dynamics of the instruments.

Once those MIDI files are done, I loaded them up into a Digital Audio Workstation (DAW). I pasted the CC values into the notes midi file, since that was just a lot faster than figuring out how to get them to line up in the same file. Then I loaded up the track with a sampler. I used sample libraries created by Spitfire Audio, and the reason I picked their string libraries is because they have controls for dynamics that behave more like the way real string instruments behave, it’s not just a simple volume knob.

The data performs the rest.

Conclusion

Exponential growth is a pretty crazy concept. There are a lot of contexts where it starts to become “normal”, particularly in the tech world. The same idea is true for numbers like “million” and “billion”, they sound similar, and they’re both big. But again, humans just aren’t great at understanding such massive numbers, we lose a bit of context on what these things actually mean.

Recontextualizing a million and a billion into time really helps drive home how different these numbers are. A million seconds from now is halfway through next week. A billion seconds from now is almost 32 years.

Transforming this exponential growth curve from a line on a page into a sound that you experience for a full minute (again, with your speakers blasted) can help recontextualize this curve to demonstrate just how powerful it is.

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.

Engineering

More from Engineering

View All

Machine Learning Based Personalization and Discovery at Postmates

Our passion is making it as easy as possible for our customers to find what they’re looking for.

November 24, 2020