Programs of experimentation and pivoting for (overconfident) entrepreneurs

How much should an entrepreneur experiment and when should they pivot? Do the answers to these questions change if the entrepreneur is subject to behavioral biases, such as overconfidence? In this paper, we develop a conceptual logic to address these questions. We focus on the design of a "program of experimentation," which we define as a sequentially interdependent set of experiments and pivot decisions an entrepreneur undertakes to develop a viable business idea. This conceptualization highlights two dimensions of program design: the number of experiments to run and the pivot threshold for evaluating experimental outcomes. Our computational model suggests that while experimenting and pivoting can improve new venture performance, it can also be taken too far. Programs of experimentation that generate frequent and early pivots may impede learning and underperform more conservative programs that generate fewer pivots. We also show that an effectively designed program of experimentation can partially remedy an entrepreneur's behavioral biases. Our work informs ongoing scholarly efforts to more fully understand the Lean Startup's strengths and limitations and advances the theoretical apparatus underlying a scientific approach to venturing.