Consider these three industries: pharmaceuticals, semiconductors, and software. They all have very high R&D intensities, but each invests in R&D in its own way, shaped by its own risks, time to market, industrial organization, regulatory regimes, and business models.

Pharmaceuticals companies live and die on R&D: their R&D Intensity averages 16.4 percent. Because the vast majority of apparently promising compounds end up as failures, a firm must sink billions of dollars over many years just to get one or two successes. That's why a drug company's fortunes can turn on the result of a single patent trial; it's also why pharmaceutical analysts work ferociously to track R&D projects as they snake their way through the many stages of the pipeline.

Herman Saftlas, who covers some major pharmaceutical companies for Standard & Poor's, says most firms highlight their pipelines much as a manufacturer might account for back orders. Pfizer, for example, documents its drug pipeline in detail in a special report (http://www.pfizer.com/pipeline). The transparency of pipelines allows analysts an opportunity to evaluate the productivity of a firm's R&D, and Saftlas has concluded that some firms are simply better at getting more for their R&D buck than others. (He notes, for instance, that Pfizer's R&D performance has been below average for its sector, in part because its 2003 acquisition of Pharmacia Corp. hasn't worked out as well as it expected; Merck, by contrast, has bettered the average.) Firms have adjusted their R&D strategies to fit the changing marketplace for drugs. The attrition rate for candidate drugs has risen so high that firms are beginning to give up hope of bringing in billions with blockbuster drugs, and are instead settling for mere tens of millions in niche markets. It's the difference between, say, a Viagra and a longer-acting antihistamine pill. The development timeline for such niches is shorter and less risky.

Semiconductor manufacturers are putting a high and rising share of their resources into R&D, for a research intensity averaging 17.1 percent. Still, Clyde Montevirgen, who covers the sector for Standard & Poor's, says R&D is not his single most important metric—sales are. He reasons that R&D still constitutes a small part of overall costs, and one that is hard to gauge because most firms jealously guard their data, rarely talking at all about developments that are more than a year away from the market. Montevirgen therefore measures R&D effectiveness by “design wins,” in which a firm announces a partnership with a hardware device maker. For example, in August, STMicroelectronics, No. 60 on the leaderboard, announced that Garmin had selected STM's chips for its new range of portable and handheld GPS and navigation devices.

Market changes since 2000 have pushed semiconductor firms to ratchet up their R&D budgets, Montevirgen says. Because an ever-broadening array of products incorporate semiconductors—think PDAs, smart phones, GPS, iPods—many in the industry hope that the market will grow even faster than it already has. However, Montevirgen cautions that the semiconductor industry is a highly cyclic market and R&D spending will surely follow the cycle. If the market turns downward, expect R&D spending to drop back.

Jim Yin, who covers major software firms for Standard & Poor's, puts less emphasis on R&D spending than his two colleagues do, even though the sector's average R&D Intensity of 17.9 percent puts it higher than those of semiconductors and pharmaceuticals. He looks, instead, mainly at sales.

Because R&D spending varies widely, he says, it's hard to compare numbers across the industry. Investments increase rapidly before a product launch and then fall quickly afterward. A young firm spends all it has to make its first launch, while an established firm can get sales out of existing lines with far less R&D investment. Software projects are notoriously prone to late delivery, and that makes it even harder to forecast the payoffs of R&D.

To limit the confusion, Yin focuses not on total R&D but just on the portion budgeted for a specific product launch he is following. He also divides software into categories requiring differing R&D investment. For example, a game company may need to spend more per product because each game has features that cannot be shared easily with others. Yin adds that Microsoft's immense profitability allows it to place bigger R&D bets than other firms do and to wait longer for those bets to pay off.

So, keep the following in mind as you review our R&D leaderboard. R&D spending and intensity do matter, but different industries require different R&D investments, as do companies in different niches within an industry and at different stages of development within a niche. Whether a firm emphasizes quick or long-term payoffs depends on its country, its industry, its maturity, and its strategy. Finally, what matters isn't how much money you spend but how wisely you spend it. Unfortunately, our leaderboard doesn't have a column for wisdom.