Depressed? Merck’s research neuroscientists would have reason to be. The pharma giant bet a fortune over the past decade that a novel compound that blocks a neurotransmitter called “Substance P” would become an effective new treatment for depression-a multibillion-dollar global market profitably dominated by Prozac, Paxil, and Zoloft. Merck bet wrong. Its aspiring antidepressant performed well in early experiments but flunked the phase III clinical trials required by the U.S. Food and Drug Administration. Not good.
Over the course of the multimillion-dollar tests and trials, however, researchers outside Merck observed a curious digestive detail signaling medical potential. The scientists noticed that ailing lab ferrets-yes, ferrets; they’re the new rats-ingesting the Substance P blocker vomited much less than expected. Ferret vomit thus became the leading indicator that Merck’s new compound enjoyed an unexpected effect on the brain. The drug apparently blocked neuroreceptors located in regions associated with both emotion and nausea-failing on the emotional front but helping curtail vomiting.
Even as Merck feared the drug would fail as an antidepressant, the ferret results held the promise that the compound could be the basis for an antinausea drug for humans. This proved a good bet. Merck won approval for Emend in 2003 as an antinausea/antivomiting medication for patients enduring chemotherapy. No, Emend is not a billion-dollar blockbuster. But it might well be a better-than-niche business for a patient population that drug companies consider a growing market-those craving sanctuary from the wrenching side effects of cancer therapy.
Mere serendipity? That’s the lazy rationale of least resistance. The better explanation simultaneously upgrades and inverts Pasteur’s famous aphorism “Chance favors the prepared mind.” Indeed it does. But now more than ever, “The prepared mind favors chance.”
That’s not mere wordplay. It’s the essence of a new generation of data-driven strategic innovation. It’s no longer enough for innovators to be sensitive to potentially provocative correlations; today’s innovators must explicitly generate them en masse. While spotting ferrets that weren’t puking their little guts out may seem like a case of pure luck, the reality is profoundly different: capital-intensive innovators like Merck increasingly structure their research initiatives to ensure that such startling correlations trigger recognition and review.
The discontinuity emerges from the vast breadth and scope of data that a Merck, a GE, an Airbus, a Wal-Mart, or a GM can reliably generate. There’s an extraordinary clash and convergence of opportunity and intent. On the one hand, innovators are seeking laserlike precision in the focus and specificity of their innovation initiatives. On the other, it’s become so cheap and easy to collect data on every aspect of an experiment’s progress that the question has become, Why not? Data diversity that would once have been dismissed as chaotic noise is now understood to contain meaningful signals. Correlation becomes the crucible for innovation and insight.
The ongoing explosion in genomic and proteomic sequencing, for example, guarantees that the Mercks, Pfizers, and GlaxoSmithKlines of 2014 will be exploring statistical correlates in dataspheres better measured in exabytes than in gigabytes. A simple mutation in a bacterial colony may prove as medically significant as a nauseous ferret. This is a tale of data-driven scale.
So Merck and Wal-Mart won’t merely explore provocative correlations; they’ll be exploring the provocative correlations of correlations. These meta-analyses will become how prepared minds cultivate chance as well as exploit it. Innovators will spend less time designing clever experiments to generate data and more time scouring the data to generate hypotheses.
Exploring correlation and causality between multiple monitoring modes should spawn even greater opportunities. We won’t be measuring ferret nausea as a function of vomit; we’ll be giving those rodents hourly positron emission scans to see how nausea manifests itself inside their brains. Organizations like GE’s aircraft engines division already rely upon these kinds of data-driven techniques to optimize their products and processes (see “If It Ain’t Broke, Fix It,” TR September 2001).
Of course, correlation isn’t causality. Remember: there are lies, damned lies, and statistics. Then again, the economics of exploring correlation for innovation are irresistibly tempting.
The future of innovation will increasingly be determined by the future of data-driven statistical techniques. The future of data-driven statistical techniques, however, depends on innovators who grasp that ferret vomit can be a source of inspiration. That’s not lucky serendipity; that’s good design.