Swinging on the branch of a tree: Project selection applications

Selecting research and development projects is always difficult because data for assessment are seldom available. ICI Americas used simple decision trees successfully to screen new product ideas and decide on a major effort in new process research. The new process research model incorporated only 10 variables, making estimation feasible; a visual sensitivity analysis gave management confidence in the model conclusions. The screening model incorporated qualitative selection criteria (for example, quick development, low risk, patentable, low capital, high sales) into a single quantitative expression of expected net present values. ICI Americas used it to select projects from 53 new product ideas. These examples illustrate the value of a logical, simple model even when data are sparse.