What can theory and tests do? Consider the following stylized representation of the strategic decision. Managers make a prediction, which we may call V̂, of the true value, which we may call V, and decide to proceed with an investment that is to commit resources. If the prediction V̂ is greater than a threshold V*. To fix ideas: V̂ is our prediction, while V* is our decision. Given the prediction V̂, the decision is more likely to be positive, if V* is smaller. and it is more likely to be negative if V* is higher. Thus, given the prediction, V* governs our decision, a point that we will clarify further in our lectures. A more conservative decision implies a higher V*, while a lower V* implies a more adventurous decision. What is the problem then? We hope to enjoy a true value V greater or equal to zero, and we make a positive decision if V̂ is greater or equal to the threshold V*. We can then make two types of errors: a type I and a type II error. In type I error, we predict V̂ higher than the threshold V* to find out in the future that the true value V is actually negative. In type II error, we predict V̂ lower than the threshold V*, but if we pursue the idea, the true value V would not be negative. Heuristics, biases, and imprecise interpretation of information imply that we predict V̂ too high or too low given V*, or a threshold V* too low or too high given V̂. This increases type I or type II errors. Theory and tests using data help to correct this mismatch. They reduce type I and type II errors by overlapping to a greater extent the decisions implied by the inequality V̂ greater or equal than V* to the event that the true value V is greater or equal than zero. Similarly, the decisions implied by the inequality V̂ smaller than V* overlap to a greater extent with the event that the true value V is smaller than zero. This leads to our view of strategy as theory, as emphasized by Teppo Felin and Todd Zenger cited in the previous lecture. Strategy as theory means that first, like scientists, managers and entrepreneurs have to define well their problem and the underlying question. Then, they have to make propositions that move from antecedents to consequences through logical links. A proposition P implies a proposition Q, which implies a proposition Z, which implies a proposition S, and so on. You start with P and conclude S, and each proposition has to be simple, clear, and in principle falsifiable, very much like scientists do. The benefits of this approach are that you understand the line of causation, you can test whether the proposition is true or false, you can stick to or change theory accordingly. It is really a way of thinking. All this shapes what we would like to call the Galilean manager. Like Galileo, the Galilean manager starts with theory. S/he first tries to understand the problems in general and abstract forms. Then, s/he defines the relevant antecedents and the logical steps that lead to consequences. Finally, s/he uses data to test the hypothesized mechanisms. If the theory fails the check of the data, or it is found to be logically flawed, the Galilean manager develops and tests a new theory using signals acquired during the process. In the language of entrepreneurship: s/he pivots. Starting with data cannot be ruled out. However, you always need a theory to interpret your data. With blind explorations, you can only be lucky. Data are most effective when you interpret them through the lens of your theory. We believe that the Galilean manager who uses theory and data can do better than the Baconian and Copernican manager. In our stylized representation, the Baconian manager only uses data like many firms whose decisions depends so much on what the books tell them. The Copernican manager develops deep theories. But like Copernicus, s/he does not have the telescope, and more generally, s/he does not have the data to test his theories. The telescope, very much like Artificial Intelligence today, enabled Galileo to make observations that Copernicus could not make. However, Galileo did not replace observation with theory. He complemented theory with observation, which we believe is what today's managers and entrepreneurs should do.
