Optimal Parallel and Sequential Prototyping in Product Design

When designing a new product, developers must balance the expected profit of improving the design against the cost and time requirements of doing so. We investigate this tradeoff by modeling the prototyping stage of product development as a probabilistic search process. We study optimal policies for single-period and infinitehorizon problems and derive closed-form solutions for the case of profits distributed according to the three extreme-value distributions. We demonstrate the role played by the tail-shape parameter of the distribution in determining the level and form of experimentation. The effect of declining prototype costs on expected profits, number of experiments, and spending is shown to depend on the form of the profit distribution and its tail-shape parameter. The paper compares the performance of pure parallel, pure sequential and hybrid parallel-sequential experimentation policies.