Bi-criteria risk analysis of domain-specific and cross-domain changes in complex systems

• Model the risk that life cycle cost exceeds β and system availability falls below α.
• Develop a procedure to rank and select proposals to reduce risk.
• Ranking involves test for stochastic dominance when criteria are correlated.
• Examine linear combinations of proposals to compare single vs. cross-domain.
• Cross domain solutions may not be better: need a model and procedure to compare.

Government and not-for-profit organizations measure success in terms of their ability to promote an organizational mission. Complex assets in such organizations are acquired in a budget-allocation process which reflects mission priorities. So, complex assets in such an environment must be managed so that availability of the asset is sufficient to support mission objectives as planned. But cost must also be contained within the budget plan, or other mission objectives may suffer. Hence, an objective in such environments is to simultaneously control (1) the risk that percent-availability will fall below a minimum planning threshold α, and control (2) the risk that cost will exceed the planned budget β. This problem is especially difficult because the two risks are negatively correlated.In this paper we examine this bi-criteria risk minimization problem, for an organization in which the departments (domains) of the organization must compete for scarce resources to achieve organizational objectives. We develop a model that can be used to assess bi-criteria risk of single-domain proposals, and a ranking-and-selection procedure which can be used to choose between those proposals. We then conduct a limited search of solutions which involve linear combinations of the proposals, in order to investigate the potential benefits of ‘breaking silos’ and ‘cooperation’ across domains. Results suggest that for complex systems at least, cross-domain solutions are not always superior to single-domain solutions, and that integrated system models are needed to properly evaluate single-domain or cross-domain solutions.