OR PRACTICE—R&D Project Portfolio Analysis for the Semiconductor Industry

Abstract

We introduce a decision-support framework for the research and development (R&D) portfolio selection problem faced by a major U.S. semiconductor manufacturer. R&D portfolio selection is of critical importance to high-tech operations such as semiconductors and pharmaceuticals, because it determines the blend of technological development the firm must invest in its R&D resources. This R&D investment leads to differentiating technologies that drive the firm's market position. We developed a general, three-phase decision-support structure for the R&D portfolio selection problem. First is the scenario generation phase, where we transform qualitative assessment and market foresight from senior executives and market analysts into quantitative data. This is combined with the company's financial data (e.g., revenue projections) to generate scenarios of potential project revenue outcomes. This is followed by the optimization phase, where a multistage stochastic program (SP) is solved to maximize expected operating income (OI) subject to risk, product interdependency, capacity, and resource allocation constraints. The optimization procedure generates an efficient frontier of portfolios at different OI (return) and risk levels. The refinement phase offers managerial insights through a variety of analysis tools that utilize the optimization results. For example, the robustness of the optimal portfolio with respect to the risk level, the variability of a portfolio's OI, and the resource level usage as a function of the optimal portfolio can be analyzed and compared to any qualitatively suggested portfolio of projects. The decision-support structure is implemented, tested, and validated with various real-world cases and managerial recommendations. We discuss our implementation experience using a case example, and we explain how the system is incorporated into the corporate R&D investment decisions.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 22 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.