Conducting parallel new product development efforts in a competitive environment can increase the expected profit from the developed product. As more development efforts are conducted, the probability of finding a promising product amongst several uncertain alternatives increases. The pitfall is that only one effort ends up as a product so the costs of the other efforts are sunk. Therefore, there is an obvious tradeoff between the investment in initiated efforts and the expected profit from this investment. Dahan and Mendelson (2001) modeled this tradeoff for equal cost parallel efforts and profits distributed according to extreme value distributions. In their settings, each initiated effort is fully paid and must be completed.
We were motivated by the facts that: 1) parallel development policies are common in industries such as Pharmaceutics, Hardware and Software and Aerospace and Defense, and 2) in real settings the parallel efforts durations can be different. Thus the decision maker can decide, when an effort completes, whether to continue the other on-going efforts or to abandon them and save the future investment. To the best of our knowledge there are no previous works that models the continue/abandon situation in parallel development efforts so there is an opportunity to improve the existing models.
We formulate a stylized model to provide answers for questions such as: under what circumstances to terminate the remaining efforts or to continue them? Is there an optimal decision policy and what is its structure? And, what are the parameters that affect these decisions?
The objective of our model is to maximize the expected net profit from parallel development efforts. Since all efforts are alternatives (e.g., concept tests) for the same type of product it is reasonable to assume that their gross profits are identically independent distributed random variables. When an effort completes its profit is realized and the decision maker decides if to continue the other on-going efforts or to terminate them.
The decision problem amounts to defining, at each effort’s completion time, a stopping policy that relies on the performance so far and future forecasts. We transform the problem into a sequential decision problem for which the literature provides an optimal stopping rule when the costs of the efforts are non-decreasing. Our first result relies on this optimal stopping rule to determine conditions under which the sequential optimal stopping rule would be optimal for the parallel development problem. Next we extend the results and provide policies for scenarios when the costs of the efforts are decreasing. We use illustrative examples to present intuitions and demonstrate the efficiency of the suggested policies against other alternatives and against the best possible results achieved by an oracle.
