Fostering breakthrough technologies — How do optimal funding decisions depend on evaluation accuracy?

• We develop a model to form guidelines for fostering breakthrough technology projects.
• The model accounts for the option to abandon projects based on interim evaluation.
• A large number of projects should be launched but many of them later abandoned.
• This policy differs notably from maximizing the expected project portfolio value.
• We show how optimal policies depend on evaluation accuracy.

There is a growing interest in fostering breakthrough technologies that offer exceptionally high value to society. However, when starting technology projects, it is impossible to know which of them have the potential to lead to breakthroughs. Therefore, organizations have adopted funding policies in which on-going projects are subjected to interim evaluations based on which some projects may be abandoned to release resources for seizing new opportunities. In this paper, we study which funding policies are optimal when the objective is either (i) to maximize the expected value of the project portfolio, or (ii) to maximize the expected number of exceptionally excellent projects that may lead to breakthrough technologies. We show that the optimal policy for funding exceptionally excellent projects is to start a large number of projects and abandon a high proportion of them later, whereas the optimal policy for maximizing the expected value of the project portfolio is to grant long-term funding to a smaller set of projects based on initial evaluation. Furthermore, we show how the trade-off between these two objectives depends on the initial project evaluation accuracy and the rate at which this accuracy improves. Our results suggest that this trade-off is particularly significant when the initial project evaluations are very uncertain but become more accurate soon after the projects have been launched. In such a setting, policies that seek to maximize the expected portfolio value may fail to promote breakthrough technologies.