In 2004 and 2005 Matthew Salganik, Peter Dodds and Duncan Watts at Columbia University ran a website called Music Lab: 48 songs by unknown bands, and 14,341 visitors who could listen, rate and download. Each visitor was sent at random into one of several separate worlds. In eight of them you could see how often each song had been downloaded in your own world; in one you could not. Every world started at zero.

The charts came out different. Watts later gave the example of a song called "Lockdown" by 52metro: 26th of 48 where nobody saw the counts, number 1 in one world and 40th in another. Where the counts were visible, the hits were bigger and the flops deeper: an early lead, much of it luck, pulls in listeners who pull in more.

Quality still set the limits. In the paper's words: "The best songs rarely did poorly, and the worst rarely did well, but any other result was possible."

The small charts on screen are our own simulation of the set-up, not the experiment's data.

Which hit can you still not explain?

Sources: Salganik, Dodds and Watts, "Experimental Study of Inequality and Unpredictability in an Artificial Cultural Market", Science 311, 2006; Duncan Watts, "Is Justin Timberlake a Product of Cumulative Advantage?", New York Times Magazine, 15 April 2007; Salganik and Watts, Topics in Cognitive Science 1, 2009.
Sounds: our own, made in code (earshot, CC0).