I am pleased to see that my recent article in the Harvard Business Review, “Should You Invest in the Long Tail?”, has stirred up a debate among long-tail enthusiasts and critics alike.

In his response, “Debating the Long Tail,” Chris Anderson certainly makes a valid point about the need to look at the long-tail phenomenon both in a relative and an absolute sense. Astute readers will have noticed that this is exactly the position I advocate in my discussion of the customer transactions data for Rhapsody, when I state:

The top 10% of titles accounted for 78% of all plays, and the top 1% of titles for 32% of all plays. Pause for a moment, though, to reflect on those numbers. One percent of a million is still 10,000 – far more than the number of titles a U.S. radio station plays in a given year, and when translated into album terms, equal to the entire music inventory of a typical Wal-Mart store.

Based on the Rhapsody and Quickflix data, Anderson again makes the argument that online markets exhibit a long tail. I agree with that assessment, and have not claimed the opposite. However, I argue the data reveal two other important patterns. First, the tail is long but extremely flat—and, as online retailers expand their assortments, increasingly so. Second, compared with heavy users, light users have a disproportionately strong preference for the more popular offerings, while both groups appreciate hit products more than they like those in the tail.

I illustrate the second point using the Quickflix and Rhapsody data, and the first using the Nielsen VideoScan and SoundScan data. In his response, Anderson devotes all his attention to the Rhapsody and Quickflix results, thereby ignoring the bigger-picture findings on the changing sales distributions in the video (VideoScan) and music (SoundScan) markets as a whole. As I note in my article, looking at snapshots is not enough—strategists need to understand how markets are changing. The Nielsen data cover multiple retailers, multiple channels, and multiple years, offering a wealth of material to test aspects of Anderson’s long-tail theory. What emerges is not a rosy picture of the fate of long-tail products: the tail increasingly consists of titles that rarely sell and that are produced by smaller-scale players.

In his response, Anderson suggests that our divergent conclusions may stem from different definitions of the “head” and “tail.” That seems odd, as I have tried to steer away from drawing a sharp line between “head” and “tail”—those are Anderson’s interpretations in his review of my Rhapsody and Quickflix results, not mine—and instead focus on describing the distribution of transactions. I do so not only because I believe studying distributions is ultimately more insightful and relevant to managers faced with product portfolio decisions; any line one draws between the “head” and “tail” is also unavoidably arbitrary.

This is evident even from Anderson’s own writing on the long tail: he sometimes implies the “head” to be the assortment offered by all bricks-and-mortar stores (and, on his blog, points to the demise of a retailer like Tower Records as evidence of the growing power of the “tail”), and at other times, as in his response to my article, suggests the “head” to consist of all products offered by the largest bricks-and-mortar retailer. The latter, incidentally, strikes me as a rather peculiar definition—if one executive at Walmart decides to cut the company’s assortment of DVDs, then all of a sudden the “tail” can grow by leaps and bounds?

Arbitrary notions of the “head” and “tail” lead to other puzzling conclusions. For instance, Anderson argues that the transactions for Quickflix are more “concentrated on the head” than those for Rhapsody. I don’t quite understand what he means by this term, but I do know that the graphs clearly show that the distribution of Quickflix’s rentals across titles is in fact less concentrated than Rhapsody’s plays.

One final comment on Anderson’s response: Why would I assume that the Quickflix finding of lower ratings for more obscure content “extrapolate[s]… to all Internet commerce,” as he puts it? The “double jeopardy” phenomenon is one of the strongest empirical generalizations in the field of marketing. It has been demonstrated in a variety of offline settings, and in a number of markets for creative goods. Having now seen the pattern in the Quickflix data, I think it is safe to bet it holds for many other online retailers, too.

I applaud Anderson for his ground-breaking work on the long-tail theory. His efforts have led to a surge in academic research on how digital technology might be changing markets and, by extension, business principles. I believe it is crucial that managerial decisions are grounded not in romantic notions of the impact of technology, but are based on empirical evidence of what is actually taking place. That is what I set out to uncover in my research, and I hope readers of the Harvard Business Review have found it useful.