Supply-demand planning in high-tech industries: Models, algorithms, and case studies

• Main Author: Atan, Mehmet Oguz
• Format: Dissertation
• Language: English
• Subjects: engineering, industrial; operations research
• ISBN: 9781124005997; 1124005994

This dissertation is organized as a collection of three independent essays that considers separate critical supply-demand planning issues in high-tech industries. In the first essay, we study a capacity allocation problem with asymmetric information. We present a decentralized multi-unit capacity auction that elicits truthful information to optimally solve the coordination problem. This mechanism is strategyproof, individually rational, and efficient. We also investigate the value of improved demand information to guide planners toward a more educated decision about under what conditions forecast improvement techniques should be pursued.
In our second research, we address the demand forecasting problem in dynamic, volatile high-tech markets, and develop a general leading indicator forecasting framework that perpetually reduces forecast variance through information updates, model combinations, and new demand observations over time. We incorporate a vast variety of dynamically evolving market information, which include indications of future demand trends, and become available throughout a product's lifecycle. We present a comprehensive theoretical setting that guarantees variance reduction when relevant data are filtered from these information sources and utilized as leading indicator signals in a consistent, systematic manner to update prior beliefs. The improvement that our framework provides in forecast accuracy is demonstrated through a case study.
In the third essay, we aim to establish high-level tradeoffs between forecast accuracy and operational costs. We model a high-tech factory to investigate the optimal operational strategies that should be employed under diverse scenarios of capacity and demand realizations. We derive the relationship between forecast variance and operational costs in closed form and show that expected operational costs increase with the increase of forecast variance. Findings suggest that variance reduction techniques should be used in forecasting to obtain operational cost savings; especially when cost parameters for supply overage and underage are higher.
Although the issues that we investigate are motivated by operational and decision making challenges observed in high-tech environments, we introduce mechanisms and offer managerial insights that are general in scope and applicable to similar problems in different industries.