To return to the example that opened the chapter, at Vertex, each of the four possible drug development programs entailed varying degrees of uncertainty in both the underlying science and the markets. Each program had accumulated only fairly limited data from small samples of human patients at the time. Anyone familiar with drug R&D knows that preliminary clinical studies are poor predictors of program success. The vast majority of drug R&D programs eventually fail either because of safety concerns or lack of therapeutic efficacy. Additionally, market sizes are notoriously difficult to forecast in pharmaceuticals (because the attractiveness of the drug on the market depends on the data generated from clinical studies that have not yet been conducted). Thus, it was natural for Vertex management to frame their discussions around hypotheses concerning issues like these: Do we understand the underlying biology of the disease? Do we understand how the particular molecule intervenes in the disease process? What are the risks concerning the molecule itself (e.g., safety or side effect issues)? What might be the potential challenges of commercializing this drug? None of these, of course, were answerable up front. There were some preliminary data and some previous scientific literature that provided clues. Different people in the organization had different judgments about which hypotheses were more plausible than others. But being clear about the hypotheses motivating each program helped to focus management’s discussion and thinking around critical issues and trade-offs. It enabled the senior leadership to ask relevant questions, consider the risks, and formulate alternatives.
Use Analytics to Drive Questions Rather Than to Provide Answers
In traditional resource allocation processes, financial analytics (and other quantitative analysis) are used to provide answers. They become key tools in the advocacy process. In a learning-centric approach to selection, analytics are not used so much to provide answers (“we estimate the rate of return on the project will be 30 percent”) as they are used to suggest questions. The power of quantitative approaches is that they provide a structured and logical way to think through a problem. Putting together a good financial analysis is really hard. You have to think through many details like future prices, future competitive conditions, future customer tastes, future costs, and alternative technologies. Forcing deep thinking about these issues, and forcing management to recognize the uncertainties around these issues, is the valuable part of undertaking structured analyses. That is, good analytics is a process, not an outcome.
Vertex utilized relatively sophisticated analytical tools (real option analysis, combined with Monte Carlo simulation) in its process of evaluating its four candidate programs. However, Vertex did not use these tools—or, more specifically, the numerical outputs of these tools—to decide what to do. The company used financial analytics as a tool for inquiry. Analyses would be conducted and presented, but, rather than triggering a decision, they triggered questions about assumptions and alternatives. The analysis would then be iterated based on more up-to-date views (and with different assumptions). One program for a molecule hypothesized to target the hepatitis C virus initially had significantly lower commercial value than the other three programs being considered. In many organizations I have seen, this would have been enough to terminate further consideration. Economically, the project looked like a loser. But, instead, the senior management team asked more questions. Why was the commercial valuation so low? Was it the size of the market? Was it due to expected competition? Was it due to the potential therapeutic benefits of the drug? Digging into the model they learned that the molecule in question had a very high expected manufacturing cost. Normally, in the drug industry, manufacturing costs are such a small part of the revenue (less than 10 percent) that they are not a major consideration as to whether to undertake a project. But, in this particular case, the molecule was so complex that it was expected to require a very expensive manufacturing process. Based on assumptions about the market and the expected price, the high manufacturing cost would eat up a big chunk of the potential profits. Again, many companies would stop there.
But the team at Vertex asked more questions. Were there alternative processes for making the drug? How much would it cost and how long would it take to find out? Some additional exploration with external experts suggested there could be some alternative paths forward (not definite) that could lead to lower manufacturing costs (again, not definite). Further analysis ensued that indicated that even moderate reductions in the manufacturing cost would have a large impact on future commercial value. Nothing was definite, but the analysis helped Vertex senior management understand the trade-offs and the risks.
Using financial analytics as a tool of inquiry is an iterative dialogue between quantitative analysis and management judgment. Unlike traditional approaches to financial analysis, the learning-oriented approach does not let the “numbers” drive the decision. But neither does it toss out financial analysis in favor of gut feelings. Analytics are used to focus the conversation, drive deeper investigation into hypotheses and assumptions, simulate alternative scenarios, and structure the project plan to address critical uncertainties.
Foster Vigorous Debate
In an advocacy approach to project selection, debate centers on winning the argument, not on learning. My experience with some companies is that debate is tacitly discouraged for running counter to norms of “teamwork” and “cooperation.” This represents a deep misunderstanding about the nature of debate. Debate is a form of conflict (intellectual), and conflict makes some of us uncomfortable. But if properly managed, debate and the associated conflict can be very productive. It can serve as a critical tool of inquiry and learning.
Vigorous debate is essential to a rigorous selection process. Through such debate we scrutinize facts, assumptions, and logic. Vigorous debate also helps to surface additional information and alternative perspectives. For any proposed program of reasonable complexity, lack of debate is actually a worrisome sign. It suggests that either people are not engaged intellectually or they do not feel safe enough to engage in constructive disagreements. Research on decision making suggests that leaders make better judgments when they utilize debate among peers and team members as a means of deepening their own understanding of a problem. A classic example is how President Kennedy handled the Cuban missile crisis in comparison to how he (mis)handled the Bay of Pigs invasion.13 With the failed Bay of Pigs invasion, a fairly small group of advisors dominated the formulation of plans. There was little debate about the plan or exploration of potential alternatives to invasion. With the Cuban missile crisis, Kennedy insisted on debate among representatives of the State Department, the CIA, the Department of Defense, and the Joint Chiefs of Staff. Subgroups were established to examine various options, and he pushed the group to think through alternatives to solutions that were on the table (e.g., an airstrike).
Creating vigorous debate does not happen naturally in organizations. Leadership must drive the process. There are several steps that leaders can take to help foster healthy debate. The first is actively and publicly soliciting input from people and parts of the organization known to have conflicting views. In many organizations, some functions (say R&D or finance) tend to dominate the project selection process. This is not healthy. It shuts out valuable perspectives from others and leads to worse decisions. The leader’s job is to make sure the relevant perspectives enter the debate in a serious way. Actively asking for input from groups or people who have tended to be shut out of the process is a good step toward changing the tenor of the discussion. As a leader, you need to make it clear that you sincerely want and need input from different sources, and that reticence to engage in the process is not acceptable. At Vertex, Boger made it clear that input on the R&D programs could come from any source. In my interview with him, he told me that, as companies grow, employees who join later are often considered (and feel) less important than the original founding team. He tried to make Vertex work differently: “We believe the last person in the door is just as important as the
first.… We want the guy on the loading dock to be thinking about clinical programs.”
Good debate also requires a willingness to criticize. Criticism, again, is one of those potential conflict-inducing behaviors that many of us prefer to avoid. It can be particularly difficult for people to criticize ideas or proposals from people more senior to them in the organization. Leaders can counteract this by demanding criticism of proposals, even those they put forward themselves. Approximately three weeks before the planned invasion of Normandy (Operation Overlord) during World War II, General Dwight Eisenhower conducted a briefing of the battle plan to the top one hundred officers of the Allied forces. Military organizations are extremely hierarchical, and chains of command are sacred. Failing to follow the orders of a superior can lead to a court martial or other sanctions. And this makes it all the more remarkable how Eisenhower started that meeting:
I consider it the duty of anyone who sees a flaw in the plan not to hesitate to say so. I have no sympathy with anyone, whatever his station, who will not brook criticism. We are here to get the best possible results.14
Notice here that Eisenhower is not just suggesting that it is okay to criticize the plan. Criticism is not just to be tolerated. It is a duty. And then he crystalizes the whole reason they are in that room: to get the best possible results. Eisenhower was indeed open to criticism. The plan did evolve and actually changed up until the final hours of the invasion. There were, however, boundaries. The strategic decision itself to invade Normandy was not up for discussion. That decision had already been debated, analyzed, and made, and this was not the time to revisit that discussion (Great Britain’s prime minister, Winston Churchill, had strong reservations about Operation Overlord). Criticism becomes most productive to learning when it is focused on the relevant issues.
A third characteristic of healthy debate is transparency. Debates should be held in forums that enable the relevant parties to hear the others’ arguments and perspectives and to respond. However, because of the political climate in many organizations, too often real debate does not happen in the open. In the meeting room, everyone is sufficiently “polite,” and there is a lot nodding. Everyone seems to agree. But, once the meeting is over, the real debate—in the hallway, behind closed doors, on the walk to the parking lot—begins. Here, different players pull various leaders aside and begin to make their case. They will point out that “Joe’s data was a bit suspect” or that “Mary’s analysis was making some questionable assumptions.” This sort of behavior is troubling in many respects, but it is particularly damaging for learning. Maybe Joe’s data really is suspect. If this were true, then wouldn’t it be better for everyone (including Joe) to know? If this issue were surfaced, rather than buried, the team could think about what new data might be needed to make a more informed decision. What if, though, the skeptic is wrong? What if Mary’s assumptions are actually quite reasonable? Wouldn’t it then be much better for Mary to have a chance to defend her assumptions and allow the management team to feel confidence in them? Of course, for practical reasons, not every discussion can take place at the assigned time in the assigned format. Two people might run into each other in the hall or be having lunch and start to talk about the program. There is no harm in that. Or a senior leader might call someone into her office to ask some additional questions about the project. Side discussions will inevitably happen. It is important, though, that the substance of those discussions be brought into the room or made available to everyone at the appropriate time. Finally, with transparency, the senior leader’s task of explaining the rationale behind his decision will be much easier. Everyone—whether they support the final decision or not—will have had access to the same basic facts, analyses, arguments, and counterarguments.
Finally, vigorous debate should be about the ideas, assumptions, hypotheses, data, analyses, and judgments and not about the people who bear them. Debate should never be personal. If debate degenerates into personal attacks on someone’s credibility, competence, or motives, the atmosphere will become so poisoned that it will be impossible for productive learning.
Keep Your Mind Open as Long as Possible
If you want to create a learning environment, then you, as the leader, must truly be prepared to learn. Learning, by definition, means evolving your thinking and, at times, changing your mind. We all enter decisions with opinions, sometimes strong ones. You may be really excited about a project and see it as definite winner (or vice versa). This is natural and fine as long as you are willing to keep your mind open to evidence, logic, and perspectives that conflict with your own. If not, you are not setting yourself up to learn. As a side point, others in the organization will quickly sniff out your bias. The pretense of “learning” and “exploration” will correctly be viewed as a farce. Your colleagues will either disengage, or they will simply tell you what they think you already want to hear.
At some point in the process, as you listen to the various arguments about a project, your mind may start to become set. Again, this is natural. But, as a leader, the second you vocalize your point of view, you may begin to distort the process. You may essentially end the debate. This may be warranted if you believe all the available information, arguments, and viewpoints have been presented and opportunities for further learning are small. Then, by all means, decide. There is no point in prolonging a debate beyond the point of diminishing returns. However, you need to be careful that you do not end the debate prematurely. If you believe there is still some information that could cause you to change your mind, there is value in keeping the debate alive. This suggests an exception to the full transparency principle mentioned above. You may not want to “tip your cards” if you want to keep the debate alive. Signal, instead, that you still have questions or are still not completely convinced about a particular option. At Vertex, Boger was so concerned about prematurely shutting down the debate that he would not disclose to the team where he stood. He commented:
We [Boger and Vertex president Vicki Sato] have talked a lot about which project to select [for the past several months], and I have pretty much made up my mind. However, I want to keep the channels of information open and to keep the discussion going.… I could change my mind tomorrow if someone comes to me with new information.
Keeping your mind open—being willing to change your mind—is difficult in practice. We all take comfort in the certainty of our intentions. Given the choice of knowing what to do and not knowing what to do, I am pretty sure most of us would choose the former. Uncertainty about our decisions makes us anxious. Moreover, leaders often feel pressures to decide now. So much of the writing on leadership highlights that leaders are “decisive.” They have the psychological stuff to make “courageous decisions” in the face of uncertainty. They are not afraid to commit. All of this sounds wonderful, but, from a quality of decision point of view, it is pretty bad advice. Good leaders make decisions when they have to, but not before, and they hold out for as much information as possible, right up until the last minute. As Boger explained to me, he tried not to make premature decisions: “I have an incredible tolerance for ambiguity… not indecision, but openness to contradictory points of view. As a result, it takes Vertex longer to make a decision.”
Another leader I interviewed, Sergio Marchionne, the CEO who saved both Fiat and Chrysler from the brink of extinction (and later lead the merger of the two) explained to me, “I have often changed my mind about the changes I want to make to an organization. That’s ok. I am willing to wait. Action does not always mean resolution.”15 Changing your mind is not a sign of weakness. It is a sign of an agile mind.
Conclusion: Building a Learning Innovation System
This chapter concludes Part II and its focus on the key components of an innovation system: search, synthesis, and selection. By necessity, the chapters were written in a sequential order. As I warned at the outset, innovation systems do not work in a linear fashion. Search, synthesis, and selection are highly interactive activities. A key
theme in this chapter on selection is that it is not a one-shot event. Selecting which projects to fund is not like choosing dishes from a restaurant menu where you are presented with a set of well-defined choices. With the exception of some minor modifications that a good restaurant might make to your dish, you are pretty constrained to choose what they offer. Project selection does not work that way. The menu evolves, and you can actually change the dishes pretty substantially. Selection triggers additional search and synthesis. Organizationally, this implies that the common practice of dedicating groups to project selection (e.g., R&D portfolio committees) is flawed because it isolates selection from other innovation activities. The management involved with selection should also be capable of search and synthesis. They are not “selectors”—they need to be innovators.
Learning was a critical theme echoing throughout the three chapters of Part II. Innovation—particularly transformative innovation—necessarily involves learning. You never have all the right answers formulated in advance. You search in a complex, incomplete, and evolving landscape of opportunities. You conduct small (inexpensive) experiments where you can. You focus on having a high “learn-to-burn” ratio. Selection itself is a learning process. A learning approach to project selection probably feels a lot “messier” than many of the resource allocation processes you may have experienced. It has some familiar elements (financial analysis, review meetings, etc.), but they are implemented in some unfamiliar ways. Financial analyses become input used to generate additional search and experimentation—they are not the arbiters of which projects are good or bad. Decisions may be postponed until better information is available. Projects can be reshaped. There is debate, criticism, and even conflict. Yet the process is not a free-for-all. It can be structured around specific meetings and specific types of information that are expected to be available at different points in the process. Rules of engagement can be set to ensure that debate is focused on the right issues and undertaken in a way that avoids personal attacks. Senior leaders need to signal their personal willingness to learn from the process.
Creative Construction Page 20