What I can do for you!
SF Music Tech
I moderated a panel on building
music related web applications at the last SF Music Tech Summit at the Hotel Kabuki in San Francisco. Thanks for coming out everyone!
Recent blog posts
Let me open with the big news: I’ve taken an engineering gig at Twitter and I couldn’t be more excited!
I’ve always been one to choose my own adventure. In the past, this has led me to found and work at a string of small scrappy startups, of which I am immensely proud. But new challenges are also important - they help us to grow and hone our craft.
A couple of months ago, I went down to Chirp - Twitter’s developer conference - to promote a Twitter app I developed called BirdHerd. While BirdHerd received some great press during the conference, what really happened was I was struck by two things:
- The quality and passion of the team at Twitter
- The scale of the problems the team are tackling
Basically, I realized that I wanted to play too. All of those users. All of that data. The scale - It was all just too irresistible. I was sold.
I believe that Twitter is a truly transformative product - so much so that I think we have trouble even imagining where it will take us in the years to come. Will Twitter be able to tell us who our next President will be, before voting even starts? Could Twitter help spread revolutions in far away countries? Perhaps Twitter will fundamentally change how businesses interact with their customers… Who knows! But I *do* know I want to be there front and center to find out.
Twitter is a product that I use every day, and I’m incredibly stoked to be working on making it better.
So in the immortal words of James Brown:
Using the CBC R3 listener data I was able to build a quick little tool to see the most popular tracks in each province, and how those top tracks change from region to region. Verdict: There are some interesting differences in taste across the country!
Click on the image below to play around with the tool.
For the technically inclined, the tool was built using the JS library Raphael and the data was manipulated using Ruby.
While digging around the CBC R3Labs data, a question came up - What does it mean for a track to be “popular” on R3?
Fortunately, it’s pretty easy to find the number of times the top 10% or 20% of tracks are played, but we also thought it would be interesting to compare some of this “popularity” data from the R3 website with that of music site last.fm. We found the comparison to be actually quite interesting, in a geeky/push-the-glasses-back-up-on-nose kinda way.
We looked at plays of the top 100 tracks on both services for a given week, and found that “popularity” is noticeably skewed towards the mega hits on last.fm, in comparison to R3. For example, the most popular track on last.fm accounts for well over twice as many of top 100 plays as its R3 counterpart. Also, the top 20 tracks on last.fm account for almost 40% of the plays of the top 100 songs. This is in contrast to less then 30% for the top 20 on R3. Check out the chart below to see the differences.
While we don’t pretend to know all the reasons for the difference in the popularity curves between these two services, it’s certainly fun to speculate! Perhaps CBC R3 visitors are more exploratory then last.fm users, often venturing out past the obvious tracks on the website. Or maybe Canadian audiences are not as influenced by the massive music marketing machine as the predominately US based last.fm audience. In a perfect world, I would like to imagine that Canada’s history of providing recording and tour grants for artists has helped fuel both the creation of this large back catalog of interesting music, while at the same time, helping build demand.
What do you think is behind difference in “popularity” between R3 and last.fm?
(This is a repost from the R3Labs blog over at CBC R3)
As mentioned earlier, Jer Thorp and I are digging through mountains of CBC Radio3 data looking for interesting tidbits. We noticed that there is a feature of the Radio3 website that lets a band list all their various members and also mark down what instrument/role that member plays. Since bands are able to change the order of the members on the list, I was able to find out what types of musicians list themselves as first, and more importantly, last!
As expected, we were able to confirm that Drummers, always the butt of jokes, were indeed often last on the list, while ego loving, spotlight hogging guitar players typically received top billing.
Check out the graphic for full breakdowns.
Recently, Jer Thorp and myself have embarked on an exciting new project for CBC Radio3. Entitled R3Labs, we’ve been tasked with sifting through a vast amount of data looking for interesting tidbits, with the goals of the project being to help listeners discover more music, bands to find new fans, and for R3 programmers to identify interesting new listener trends. You can read the official R3Labs introduction on the CBC blog.
The data we have been given access to is very comprehensive! For example, we have data on over 40 million website track plays, going back to 2001 when “indie” still meant something, Last.fm and Gmail didn’t exist yet, and I was still touring with my own band.
One of the first things we did with this data was to geocode all 40 million of the listens, and plot it on a map using Processing. Check it out below:
I’m going to be doing some regular posts here with updates from this project, so stay tuned!