One of the movies that will receive attention and possibly some awards on Oscar night is Moneyball, based on the true story of baseball's Oakland Athletics. Now it appears that some people in pharma are trying to copy what Oakland and other baseball teams did a decade ago. The likely result is that in pharma, as in baseball, not much will change.
The essence of the Moneyball story is that Oakland, as a small-market team, could not compete with teams in New York, Boston, Los Angeles and other major markets by paying exorbitant prices for free agent ballplayers. To help overcome that handicap, their general manager hired three econometricians (reduced to one character for the movie) to develop stochastic models (complex statistical and mathematical formulas) that could identify unappreciated performance indicators. By seeing which ballplayers excelled on these arcane measures, the A's were able to buy their contracts for relatively inexpensive amounts because wealthier teams didn't bid up the cost.
The A's enjoyed remarkable success in 2002, the first season in which they used quantitative modeling. Their results have been far less successful in subsequent years. Over the nine seasons beginning with 2003, the A's made the playoffs twice. In the last five years their mediocre records failed to even qualify them for the playoffs and they are now what used to be called a "second division" team.
In baseball, the A's haven't produced winners in recent years because the big-market teams also started to use the arcane percentages and formulas, but mainly for selecting the role players to round out their rosters. The stars are still chosen by competitive bidding, and in that contest big money is everything. So even as the mathematical models no longer gave Oakland a competitive advantage, perhaps more important, the wealthier teams were also able to acquire two or more players for various team roles. That allowed them to keep winning when the inevitable injuries or slumps occurred.
Now, in an effort to go forward from what is widely recognized as a failed business model, some segments in pharma are trying their own version of Moneyball. Most likely, the result will be what it was for baseball. While stochastic modeling will turn up an occasional gem for one company or another, the extent to which it can substantially increase productivity at new drug development or routinely create marketing successes appears doubtful.
Thomas Davenport, a business professor at Babson College, strongly encourages companies to rely on quantitative modeling for their management decisions. In his book, Competing on Analytics, he actually studied three pharmas to see how they use stochastic methods and what differences they make. Those companies were Novartis, Vertex and Millenium.
Basically, Davenport found that the greatest potential for quantitative models is in R&D, where the approaches can help analyze the enormous amounts of data generated by the "high-throughput screening" of chemical compounds. He readily admits that no company has mastered the complex process. Drug development remains an inherently high-risk enterprise, and while stochastic models can help inform the risk and accelerate the process, they cannot fundamentally alter its basic nature.
On the marketing side, Davenport saw that the pharmas he studied basically disdained the modeling approaches. Senior business executives at those companies saw little to be gained from bringing IT and other managers of quantitative modeling into the decision-making process. Occasional circumstances did give the modelers access and, in a few cases, an unexpected boost did occur. But as far as basing the business end of pharma on Freakonomics, fugettaboutit.
It is interesting that pharma is not the only health-care segment weighing whether it should base its future on quantitative modeling. Some health insurers, such as the Optum division of United Health Care, seek to earn their revenues by assessing the patient databases generated from electronic medical records and advising Accountable Care Organization providers on optimal practices.
And at this point, one encounters the true limits of stochastic modeling. Business as a type of social organization operates with highly quantified, discrete factors: dollars, finished units, work hours and hundreds more. For that reason, any process deserves encouragement if it can help make sense of each factor's role and the inter-relationships of many factors. In that way an analytical approach can help to inform decisions. On the other hand, looking to have the models decide existential matters such as, "what business should we really be in," is basically an effort by executives to shirk the responsibilities for which companies pay them enormous compensations.
If it is true, as Penn professor and former White House advisor Ezekiel Emanuel recently wrote, that health insurers will cease to exist and instead become data analyzers, stochastic modeling will not enable them to survive. Other entities (GE/Intel, IBM, Wal-Mart, many others) possess several advantages for competing in that space. Insurers and banks have historically made their money by aggregating capital and getting investment returns on the float. If they can no longer do that, then mathematical modeling will not save them. The same reasoning applies to pharma and its historic mission of developing new therapies to advance the standards of care.
The securities industry is one sector that did base substantial segments of its business on quantitative modeling. Journalist Scott Patterson described how well that worked in his book, The Quants: How a New Breed of Math Whizzes Conquered Wall Street and Nearly Destroyed It. As health care usually arrives at least a decade late for the business fashion parties, one can only hope they intend to learn from failure, rather than emulate it.
To check out more Check Up items, go to www.philly.com/checkup.