State of the Computer Book Market

by Tim O'Reilly, originally published in O'Reilly's Spring 2004 Catalog
April 2004

For a long time, I've been interested in redressing the poor job that traditional market research firms do of reporting on new developments in technology. Users knew about the importance of Perl, Linux, and other open source projects long before they were covered by industry analysts. As a result of my frustration, we've launched a new market research group at O'Reilly, whose mission is to develop quantifiable metrics about the state of technology adoption.

One of the rich sources of data we're mining is, quite naturally, book sales. In 2002, we subscribed to Nielsen BookScan, a service that provides continuous market measurement of retail book sales based upon electronic sales data (actual copies sold at the cash register to customers) rather than the traditional publisher measure of sell-in (copies sold to retailers). We now receive weekly information on what books are sold, in what quantity, and at what price.

BookScan doesn't yet report on every market outlet; Nielsen estimates the data represents perhaps 65% of the U.S. domestic market. It covers major national accounts, including Amazon, Barnes & Noble, and Borders, plus many regional chains and major independents. The report also covers only the top 3000 books sold each week. (Books that fall below this Top 3000 title list sell only nine or ten copies per week nationwide, so they are minor contenders and don't change the shape of the trend line.) So, though the report doesn't include every account or every title, it's still a very good tool for understanding the trends in what people are buying.

Of course, it's left as an exercise for the subscriber to do detailed analysis on the meaning of the raw numbers. And it's some of that analysis that I want to share with you.

Weekly Averages Trending Up

First, the bad news: the U.S. computer book market shrank by 10.8% last year in the BookScan Top 3000 year-end totals.

2002: 10,060,026 units
2003: 8,971,774 units

The good news for O'Reilly was that we bucked the trend and were the only major publisher that had an increase in unit sales. But more important than O'Reilly's ability to grow market share in hard times is what I see about the market as a whole. We're seeing a definite uptick in recent months, with the overall weekly averages trending up significantly. January 2004 has already beaten the same period last year (and for 2002) in terms of units sold. The one risk factor that might choke off the rebound is that some bookstores have reduced their shelf space for computer books just as sell-through is increasing. Alert retailers will be watching the rebound and increasing inventories of at least the faster-moving categories.

Consumer Categories Lead Growth

Figure 1 gives a rough snapshot of the overall market trend through 2003, divided into six major categories: applications, platforms (operating systems), software development, system and network administration (including certification), other, and unknown. Growth in the applications category was driven principally by Photoshop and other digital-media applications, while the year-end surge of operating-system books was driven by the release of Mac OS X Panther and holiday buying of low-end Windows books to go with new computers. The software development and system/network administration declines are perhaps related to outsourcing, though there may well also be a significant categorization artifact there, since many development and administration books are niche titles that don't fall into obvious categories, and so are listed as "other" or "unknown." We expect to improve our categorization of these titles in the near future.

Figure 1. 2003 U.S. Computer Book Sales by Category
(source : Nielsen BookScan Top 3000 / O'Reilly Research)

Enthusiasm vs. Raw Market Share

Another point that I want to bring to your attention is that book sales tell a very different story about technology adoption than do many other statistics. For example, while we hear from technology analysts that the Mac has only 3% of the operating system market, and Linux only 4-5%, members of these communities show far more enthusiasm for books on their operating systems than do Windows users. As shown in Figure 2, Windows represented only 56% of the operating system book market, while Mac OS X represented 23%, and Linux/Unix 18%. (In fairness, in addition to the innovation and enthusiasm in the Mac and Linux markets, the seasonality of product-release cycles also affects sales. Windows XP was released in mid-2001 while both Mac OS X and Linux have had more frequent and recent releases.)

Figure 2. 2003 Operating System Book Sales by Platform
(source : Nielsen BookScan Top 3000 / O'Reilly Research)

Over the coming year, I expect to do more reporting on the technology trends that are revealed through book sales and other related indicators, such as online job postings, newsgroup and mailing list traffic, and blogging statistics. These "faint signals," when properly correlated and interpreted, can do a much better job of forecasting technology growth patterns than traditional market research.