Free PDF available here

Premise

R&D productivity can be measured well enough to: 1) establish which companies are more or less productive than their peers; and 2) give some indication of changes that underperforming companies can make to substantially improve their performance

Managements and investors have long been aware that R&D productivity is weak; left largely un-addressed, all indications are that the average company’s economic returns to R&D spending now are below costs of capital. Yet there is still no consensus on how R&D productivity should be measured. Lacking credible metrics, managements fail to set operational goals and hold their employees accountable for meeting these; and, investors have no objective productivity criteria for holding managements responsible

Subjective estimates of R&D productivity tend to fill the void; as incomplete metrics, these tend to raise false hopes, and in so doing ease the pressure on managements just when efforts to improve productivity should be mounted far more aggressively. A rising number of NME/NBE[1]’s filed or approved is a favorite. Increased filings and approvals are plainly a good thing, but without an understanding of what was spent to generate these outcomes, how these investments compare to the earnings potential of the corresponding new products, and whether a given level of increase reflects anything other than the sporadic nature of research output, the filings/approvals number can be grossly misleading

Perhaps worst of all, sizable contingencies argue that R&D productivity simply cannot be measured – it’s too complex, the timeframes are too long, and so forth. This emphatically is not true, and often rings of simple self-preservation

We make no assertion that our metrics are millimeter precise, or that we’ve captured all of R&D productivity’s moving parts. What we have done is develop a workable empiric foundation for determining which companies are more or less productive, and for providing actionable levels of insight as to what underperforming companies can do to improve. There is no reason for managements not to act, or for investors not to hold managements more immediately accountable for raising their organizations’ R&D productivity

Internal bias

Despite the fact that the large R&D based companies each generate a very small share of global innovation (1.3% on average), most dedicate very large percentages (70% on average) of their development spending to their own discoveries (Exhibit 1). In effect, the companies can be described as generally ignoring externally-generated discoveries in order to focus development spending on internally-generated discoveries. If the internally-generated discoveries were of higher average quality than those available externally this might make sense, but that’s not the case (see next). We calculate a simple ‘internal bias index’ by dividing the percentage of development spending dedicated to the company’s own discoveries by the same company’s share of global biomedical innovation. Companies with relatively low shares of global innovation but whose development pipelines consist largely of their own discoveries have high (unfavorable) scores, and vice versa. Among the large cap names, LLY, AMGN, and AZN have the highest internal bias scores; BMY, PFE, and Roche have the lowest internal bias scores (Exhibit 1, again). We believe that high rates of internal bias – even for the companies with the most favorable scores – are a major addressable contributor to poor R&D productivity

Lack of focus

Not even the largest R&D spender can fund active world class discovery efforts in all possible research areas. As such, companies focus their efforts in a select group of research areas; however most fail to achieve world class ranks (e.g. 1^st, 2^nd, or 3^rd) in the subset of research areas they’ve targeted

For each of the 22 largest US-listed companies (by R&D spending) we identified the research areas accounting for at least 80 percent of each company’s pre-phase III (i.e. from discovery to phase II) innovation. We then calculated each company’s rank in each of its key research areas, and use this as an index of whether companies are or are not actively managing their early- to mid-stage research such that they achieve high average rankings across the research areas they’ve targeted

The companies’ average rank in their key research areas is 8^th (Exhibit 2).  I.e., most companies are actively conducting research in areas where they have little or no hope of producing a commercially viable result, an unforced error that many companies compound by prioritizing these non-competitive discoveries in their development programs (Exhibit 1, again)

The best in class company is BMY, who has an average rank of 1.9 (Exhibit 3, column ‘f’) in the 11 research areas that account for 82 pct (column ‘c’) of its pre-phase 3 innovation. BMY is closely followed by Roche and PFE, who have average ranks in their targeted research areas of 2.0 and 2.5, respectively. ALXN is worst in class among the 22-largest R&D spenders with an average rank of 21.4; among the large caps LLY is worst in class with an average rank of 11.7 (Exhibit 4, column ‘f’) across the 16 research areas that account for 80 pct (column ‘c’) of the company’s pre-phase 3 innovation. LLY ranks first in none of its targeted research areas (column ‘g’) and among the top 3 in only 2 research areas (columns ‘h’ and ‘i’)

Simple ‘dollars per idea’ inefficiency

The number of quality-adjusted ideas produced per dollar of R&D spending varies widely across companies; some companies simply spend too much to generate a given amount of innovation as compared to their peers

Among the larger cap names, JNJ, GSK, NVS, LLY and SNY all consistently produce fewer quality-adjusted ‘units’ of innovation per R&D dollar spent than peers (Exhibits 5a thru 5e); AZN was dramatically below the peer average for most of its history, and has only recently reached the (falling) peer average (Exhibit 5f). BMY (Exhibit 6) is best in class among the larger cap names

Low average quality

As is the case with R&D spending per idea, the average apparent quality[2] of ideas generated varies widely across companies, and some companies have consistently produced innovation of lower average quality than peers’.  Bayer, GSK, LLY, MRK, and SNY (Exhibits 7a thru 7e) all have average qualities of innovation that are consistently below the peer average. Conversely BMY, CELG, and VRTX all have average quality levels that consistently exceed the peer average (Exhibit 8a thru 8c)

Size

Larger firms tend to have lower R&D productivity than smaller firms; this is true both across firms (larger less productive than smaller) and within firms (a given firm becomes less productive as it grows)

Exhibit 9 demonstrates the principle across firms, comparing cumulative 1993 – 2013 R&D spending for a given company (x-axis) to that company’s economic returns to R&D spending (y-axis).  Exhibits 10a thru 10c demonstrate the principle within firms for AMGN, BIIB, and GILD, respectively. As the real value of each firm’s R&D spending grew (grey line, left y-axis), the firm’s R&D advantage eroded (average quality of innovation, right y-axis). VRTX is the only remaining firm to have kept its R&D quality advantage as it grew (Exhibit 11)

We appreciate that large size can hardly be described as an unforced error – however allowing (or committing) the behaviors that lead to falling productivity as firms grow is an unforced error. For example, we believe that as firms grow, flows of power, information, and compensation tend to shift from meritocratic (information and money flow to best ideas) toward bureaucratic (information and money flows determined by relatively rigid organizational processes). And, we see a pattern in which innovative breakthroughs form the basis for new companies (at birth, the firm is a great idea seeking capital), who once successful, generate significant cash flows whose primary use is to perpetuate the firm through further discovery (now maturing, the firm becomes capital seeking great ideas). With the exception of VRTX (Exhibit 11, again) and DNA (not shown), we’re unaware of a firm that has maintained a significant R&D productivity advantage during and after a phase (or phases) of significant growth