It only describes the shape of the market ... It's a market definition, not a competition definition.
Not wholly accurate. The "long tail" phenomenon as used in this context describes the belief that the revenue gained from the long tail exceeds that gained from the top selling items. The idea encourages production and sales of items that appeal to niche markets.
This idea has not yet been born out by the marketplace. While it's true that those who frequent Slashdot (and let's be honest, most here are probably close to being outliers in any normal distribution) may enjoy the less popular media the populates the long tail, most of the media sold in the U.S. is still sold by Walmart, which keeps primarily the most recent stuff in stock.
Just a couple of days ago I listened to a presentation by researchers from Emory who came to the same conclusion with music album sales. In order for the long tail to work, prices have to be low enough that sampling behavior is enabled--but there is still a significant herd mentality when purchasing music. (So it's not much of a stretch to extrapolate this to video, to align w/ the article.) A quick peek at the seasonal distribution of sales explains this: for album sales, the Christmas season sells significantly more than at the rest of the year, suggesting album sales are mostly gifts, where purchases are likely to be risk-averse. (So, no long tail effect there.) The same research has not been done at the _track_ level, where sampling is more likely, but it the very un-bundling of tracks from albums that is leading to revenue declines in the music industry.
Other research suggests that Netflix would be *more* likely to display the long tail effect than other media outlets. Its very nature is sampling (rentals at very low cost), so it's a good place to look for the long tail. Decision psychologists refer to the "want self" and the "should self", the competing faces of our internal dialogue. A service like Netflix is more likely to trigger the "should self" because gratification (watching a DVD) is delayed from the initial decision, leading to a higher likelihood of "art film"-style selections than is found at, say, a $1-night DVD vending machine. Thus, the findings of these researchers showing no long-tail effect from Netflix data are additionally convincing.