In late 2004, a website called musiclab.columbia.edu offered its visitors a deal that was hard to refuse: 48 songs by bands nobody had ever heard of, free to download. Fourteen thousand three hundred and forty-one people came through, recruited mostly from a teen site called bolt.com. Rate a song from one star to five, download it if you like it, log off. The setup looked like a music page.
It was really nine music pages running in parallel. Twenty percent of arrivals were handed the songs with no numbers attached. The rest were split into eight social influence worlds, each shown live download counts from its own world only.
Matthew Salganik, Peter Dodds and Duncan Watts published what happened next in Science in February 2006. In the one world where every choice was made blind, the songs settled into a single honest order - the cleanest measure of what the crowd actually liked. In the eight worlds where every song wore its count, that same crowd produced eight different charts. The good songs almost never sank, the bad ones almost never broke through, and every song in between could be a hit in one world and a dud in the next. Make the signal more salient - sort the songs by their counts instead of scattering them across a grid - and the effect got stronger still. Popularity concentrated harder than quality alone could explain, and it scattered differently every time the tape was rerun.
That result raises a question most retellings skip. Was the count reporting demand, or creating it?
The lie they told once
Salganik and Watts came back for a third run, published in Social Psychology Quarterly in 2008, with 12,207 fresh participants and one surgical change. Let a social influence world settle into its natural order. Then, once, in two newly opened worlds, invert it.
"She Said" by Parker Theory sat at number 1 with 128 downloads. "Florence" by Post Break Tragedy sat last with 9. The researchers swapped those counts, and every other pair down the list, ties broken at random, so the next 9,996 participants - spread across four worlds, two of them running the inverted chart - arrived at a site where the least-listened-to song looked like the smash hit. From that moment on, the counts updated honestly. One lie, told once, at the door.
The lie came true. Download trajectories bent toward the numbers the visitors had been shown, and the paper's own summary still reads like a trick being performed on the reader: "most songs experienced self-fulfilling prophecies, in which perceived - but initially false - popularity became real over time." Extrapolate every trajectory to its finish and the projected final rankings in the inverted worlds bore almost no relationship to the rankings the songs had earned - a rank correlation of 0.16 in both, against 0.84 in the untouched world. The paper's authors expect the true number 1 song to claw all the way back to the top. Almost everything else kept whatever the lie handed it.
So here is my opinion, stated as plainly as I can. A rank line is not a thermometer; it is a thermostat. It reads the temperature, then drives it. Every bestseller list, download counter, follower count and "most popular" badge you have ever displayed was doing two jobs at once: reporting the past and manufacturing the future. Marketing keeps treating these numbers as verdicts, proof arriving after the game. The MusicLab experiments say they are inputs. Position does not just record demand. It writes the next line of it.
And the 2008 data carries a warning almost nobody quotes. Distorting the signal did not merely redistribute success - it shrank the market. After the inversion point, participants in the honest world made 2,898 downloads; in the two inverted worlds they made 2,197 and 2,160. People steered by fake counts toward songs they did not actually like listened to fewer of them, downloaded less of what they heard, and left with less. The authors frame it as a social dilemma: any one band gains by inflating its numbers, but once enough bands do it, the numbers stop meaning anything, the crowd tunes out, and every band loses together. Faked popularity pays the first seller who tries it and bills the entire market for the privilege.
What the label does and does not do
This is why I keep thinking about what it means to sell rank in the open, which is what Bid Spotlight does. Every placement on the board is labeled Sponsored. No rank exists before a payment confirms one. Nobody pretends the board's order is the organic verdict of the crowd; the number is an ad, priced in daylight.
The MusicLab result says the position works anyway. A subject in a social influence world was about six times more likely to listen to the song ranked first than to one in the middle of the list, and about three times more likely to try the very bottom of it. The number did the recruiting. Honesty about what a ranking is does not switch that mechanism off. Nothing does. Early listens became downloads, downloads became rank, and rank recruited the next listener. That loop is the thermostat running.
What follows for a buyer is the part worth writing down. A top slot is an amplifier, and an amplifier multiplies whatever signal it is fed. The inversion could not keep genuinely appealing songs down - the true number 1 came back, on the paper's own projection - because the top slot sped up discovery of exactly what was true. The same speed applies to what is weak. Position does not change the verdict your page receives; it delivers that verdict sooner, and to more people, and the verdict compounds either way.
Buy the position, and build the page that can back it up - then claim your spot at bidspotlight.com/submit.