Bits: July 20, 2004
iTunes 100 million song contest
Apple recently had a contest to give away a 17" laptop and a coupon worth 10,000 song downloads to the person who bought the 100 millionth song from the iTunes Music Store. They also gave away 50 iPods (1 at every 100,000 from 95 million to 100 million). They displayed the current number of songs purchased (updated periodically) and I started keeping track of it to see if I could try buying some songs at the right time to give me a chance at winning. Below is a plot of the data I recorded over a few days, including a line fit based on the data before 99.9 million that gave me a general idea when it would hit 100 million. I knew this would be off, but at least I knew that I would need to stay up Sunday night. The graph also shows how song buying changes over the course of a day (the vertical dotted lines mark midnight eastern time). Looking at the slope, it shows that the iTunes Music Store sells about 6.3 songs per second on average.
Of course, the above plot wouldn't let me know when I should actually buy songs, since the buying accelerated at the end. Below is a zoom in on the end of the graph above. It has another line fit that includes the data after the last iPod was won at 99.9 million until I started buying music. It shows that there was a 10,000 song jump around 99.9 million when the iPod was won. Accounting for this, the red star on the line fit shows when the song count should hit 100 million - 10,000 = 99.99 million. This analysis predicted that the 100 millionth song would be purchased at 1:33 am. According to Apple, it was actually purchased at 1:21 am. From my graph, it looks like the winning song was purchased at 1:23, but all my data points are delayed a bit.
Regardless, I was purchasing songs around the time of the winner, so I guess I did everything I could. Given that they were selling almost 100 songs per second at that point, I shouldn't be disappointed that I didn't win. Oh well, at least I got some good music out of the deal.