Tim O'Reilly, Long Tailer

Chris Anderson
The Long Tail

May 15, 2006

Tim O'Reilly [ http://tim.oreilly.com/ ], the CEO of O'Reilly Media [ http://www.oreilly.com/ ], the technology book publisher, helped a lot (along with his research staff) with my book research. Now he's just posted a fascinating Long Tail analysis [ http://radar.oreilly.com/archives/2006/05/long_tail_evidence_from_safari_1.html ] of his own book sales compared to the industry at large.

You should go read the whole thing, but here's a taster. His team plotted Nielsen Bookscan data (major bookstores) for the top 10,000 titles, grouped by sales decile. They then did the same for the top 10,000 boot titles on Safari, their own online ebook service. The results are as follows:

What makes this a Long Tail analysis? After all, the tails are the same length (10,000) titles. The answer is the there are two scarcity functions that tend to make sales hit-heavy in tradition bricks-and-mortar retail:

Limited shelf space and thus limited inventory
Limited "findability" due to constraints of physical organization and the fact that hits get all the prime real estate.

Safari helps in both those dimensions, but in this instance I suspect we're mostly looking at the effect of the second. When books are easier to find, thanks to all the online advantages of search, recommendations, easy sampling and dynamic popularity lists, the niches are on an equal footing. And thus overall sales tend to be less hit-heavy.

There's a lot more analysis and charts in Tim's post. Read the whole thing [ http://radar.oreilly.com/archives/2006/05/long_tail_evidence_from_safari_1.html ].

UPDATE: In the comments to his post [ http://radar.oreilly.com/archives/2006/05/long_tail_evidence_from_safari_1.html ], Tim clarifies that the graph shown actually represents 2,100 books, not 10,000. "The data displayed is for the intersection of the two data sets. That is, for the Safari graph, it's for the 2100-odd books that appeared in both the Bookscan top 10,000 AND in Safari. Books that don't appear in both sources are not represented in the Bookscan graph."