Improving resource allocation with portfolio decision analysis
By Ahti Salo, Jeffrey Keisler and Alec Morton (LEFT TO RIGHT)
Practically all organizations achieve their objectives by building a portfolio of activities subject to budgetary and other relevant constraints. Industrial firms, for example, allocate resources to research and development projects (R&D), expecting that these projects lead to profit-generating products. Municipalities commit funds to initiatives that offer social and educational services to citizens. Regulatory agencies impose controls to counter threats to human safety and environment. Even individual decisions can be viewed analogously. For instance, students must decide which courses and recreational activities to take on, realizing that time is a limited resource when seeking to complete one’s studies successfully and on schedule while still having a rewarding social life.
Despite their differences, these decision problems share similarities. They all involve decision-makers faced with alternative courses of action, which, if selected, consume resources and lead to consequences with regard to multiple criteria. The actions are typically interdependent, if only because they compete for resources from the same resource pool. They may have direct interactions, too, as in the case of synergistic R&D projects. Uncertainties, too, can be important, because it may not be known ahead of time what consequences the actions will lead to or how much resources they will consume. And apart from the technical aspects of decision modeling, the social and political aspects of the decision process can be central, particularly when addressing strategic questions where the decision-makers need to develop an improved shared understanding of and a joint commitment to the appropriate way forward.
Portfolio Decision Analysis
In response to these challenges, decision analysts have developed approaches that bring greater rationality and transparency to resource allocation ( Kleinmuntz, 2007). These approaches – which we call “portfolio decision analysis” or PDA for short – are employed extensively in most industries and many areas of public decision-making, yet PDA has not attracted comparable attention in the operations research literature. This realization was one of our motivations for editing the book “Portfolio Decision Analysis – Improved Methods for Resources Allocation,” which has just been published by Springer . Written by leading researchers and consultants, the 15 chapters in the book present theoretical and methodological advances. They also illustrate how different problems can be approached with PDA.
In contrast to “conventional” decision analyses, a distinctive feature of PDA is that it explicitly recognizes interdependencies among the alternatives. This is not the case, for example, if decisions are taken a one-on-one basis by checking whether or not the alternatives exceed a pre-defined hurdle rate. By design, PDA recommendations are therefore based on a more comprehensive problem representation where resource constraints bear equally to all alternatives. Also, the portfolio approach ensures that decisions are taken following a well-structured process, based on a comparable set of information about each of the alternatives. Such a process improves the quality of decision-making because it ensures that all alternatives are treated similarly. For example, it eliminates the risk that some unattractive alternative would be chosen merely because it was fortunate enough to have been evaluated at the beginning of the financial year when the budget was not yet depleted.
PDA methods can be applied through a series of logical steps, proceeding from: (i) structuring the problem by clarifying what the decision objectives are and how important they are, to (ii) identifying what alternative courses of action will be analyzed together as a portfolio, to (iii) assessing what impacts these alternatives have on the objectives, to (iv) specifying what resource and other constraints must hold in portfolio choice. Depending on the problem, characterizing the interdependencies between the alternatives or characterizing the key uncertainties around costs or impacts may be necessary. Finally, synthesizing all these inputs within a PDA framework gives insights into which portfolios are “good” choices (i.e., non-dominated, meaning that no other portfolio is least as good on all criteria and strictly better on some), and what alternatives are contained in these non-dominated portfolios.
Taking an example from the pharmaceutical sector – where PDA methods are used extensively and whose portfolio management practices are covered by Jack Kloeber in our book – the alternatives can be R&D projects in different clinical phases in the product development pipeline, the constraints are implied by limited budgets and the availability of human resources, and the objective is to maximize the profitability of the product portfolio that builds on R&D. In this decision problem, PDA methods show which portfolios are likely to generate most profits and, moreover, which projects should be selected, subject to relevant constraints and the recognition of possible project interdependencies.
A key question in PDA is what alternatives should be treated as being part of the “same” portfolio. Expanding the set of alternatives may lead to more comprehensive (and hence “better”) solutions. Yet expanding the problem scope in terms of its planning horizons, amount of resources, span of organizational unit or number of participating stakeholders may lead to difficulties. For instance, more time may be needed to assess the impacts alternatives; these assessments may be less accurate due to longer planning horizons and management may not be able to get a good intuitive grasp of portfolios that contain hundreds of alternatives. Furthermore, it can be difficult to “operationalize” criteria that can be applied to alternatives that are very different from each other. For example, while the efficiency of fuel cars can usually be measured by fuel consumption, it is not immediately clear what a suitable shared metric would be for measuring the efficiency of a large pool of vehicles consisting of fuel cars, hybrid cars, motorcycles and electric buses. Thus, instead of seeking to determine a global “optimal” portfolio, it may be better to define multiple, possibly interlinked portfolio problems, for instance by making use of formal problem structuring methods discussed by Gilberto Montibeller and L. Alberto Franco in our book.
From the viewpoint of accountability, a benefit of a systematic PDA process is that it leaves an explicit track record that can be used later on. This track record can be used for justifying and communicating why the decisions were taken the way they were or for explaining what the likely results would have been, had a different portfolio been selected. The track record can also be sued for purposes of organizational learning, because organizations may work more efficiently if they adhere to a similar process design across departments and sections. These kinds of process benefits can be important particularly in public decision-making where the decision process is often subjected to at least as much scrutiny as the decision itself.
Using Tools and Software
PDA problems are computationally challenging: Even if the available set of alternatives is small, the number of different portfolios can be staggering. For instance, in the seemingly simple problem where five projects are to be selected from a set of 20 project proposals, 15,504 different portfolios can be considered. Although the large majority of these portfolios are uninteresting – in the sense that they are dominated by some other portfolio – the large number of possible portfolios makes it imperative to apply appropriate methodological and computational tools.
Thanks to active efforts in this area, several software tools now help identify all non-dominated portfolios a rational decision-maker would be interested in ( Lourenço et al., 2008). Some of these tools – such as RPM-Decisions, which is based on Robust Portfolio Modeling 9RPM (Liesiö et al., 2007) – admit incomplete information about the alternatives and the relative importance of evaluation criteria. Based on this information, they identify “core” alternatives that belong to all non-dominated portfolios. These core projects can be recommended to the decision-maker, because they would belong to the optimal portfolio, even if one were to obtain more information about the alternatives or the decision-makers’ preferences. Thus, the possibility to work with incomplete information means that PDA methods can be usefully applied before all the alternatives have been thoroughly evaluated in order to give decision recommendations based on information that is already available or which can be generated with a reasonable effort. The value of RPM-Decisions and similar tools has been demonstrated convincingly in numerous case studies. One of these is presented by Antti Toppila, Juuso Liesiö and Ahti Salo who report a successful application of RPM to technology management at a major telecommunication company.
Numerous methods exist that can be employed to support portfolio decisions. This leads to the question of what methods are “good” or perhaps even “better” than others. There is no definitive answer to this question, because the “goodness” of a given method in a given problem depends upon how adequately this method is able to capture the salient problem characteristics and how responsibly it is deployed to inform decision-making. Having said this, we believe that a strong case can be made for approaches that build on the logical foundations of decision theory, most notably in the spirit of multi-attribute value theory and multi-attribute utility theory which both have solid axiomatic foundations.
Many promising avenues exist for further research and applied work in PDA. On one hand, recent advances in PDA methods and tools provide tested but not yet widely adopted approaches for improving the quality of decision-making. This suggests that organizations can reap significant benefits by revisiting their current practices and replacing these by “state-of-the-art” approaches where appropriate. On the other hand, decision processes themselves need to evolve under many pressures. For instance, there is a need to: account for a growing range of incremental and disruptive uncertainties; to understand what portfolios perform acceptably even in the less desirable scenarios; to bring in knowledge from an even greater number of experts; to exploit technologies for social networking in portfolio decision-making; and to build capacities for reaching well-founded decisions more quickly. These and yet other challenges suggest new topics for research, too, leading us to believe that PDA will become an increasingly vibrant area of decision analysis and operations research at large.
Ahti Salo (email@example.com) is a professor and vice head of the Systems Analysis Laboratory at Aalto University School of Science in Aalto, Finland. The president of the Finnish Operations Research Society (FORS), Salo has published extensively in leading international journals (including Management Science and Operations Research) and received awards for his research from the Decision Analysis Society of INFORMS.
Jeffrey Keisler (Jeff.Keisler@umb.edu) is an associate professor at the University of Massachusetts-Boston. Keisler is president-elect of the INFORMS Decision Analysis Society and has been working in decision analysis for more than 20 years as a researcher and as a consultant to government and industry. He is a Fellow in the Society of Decision Professionals.
Alec Morton (A.Morton@lse.ac.uk) is a lecturer in operational research in the Department of Management at the London School of Economics and Political Science where he teaches courses in decision analysis, simulation and statistics. A graduate of the Universities of Manchester and Strathclyde, he worked at Singapore Airlines and the National University of Singapore before joining LSE.
This article was prompted by the INFORMS International Activities Committee, of which Salo is a member.
- D.N. Kleinmuntz, 2007, “Resource Allocation Decisions,” in W. Edwards, R.F. Miles, D. von Winterfeldt (eds.), “Advances in Decision Analysis,” Cambridge University Press, Cambridge, Mass.
- J. Liesiö, P. Mild and A. Salo, 2007, “Preference programming for robust portfolio modeling and project selection,” European Journal of Operational Research, Vol. 181, No. 3, pp. 1,488-1,505.
- J. Lourenço, C. Bana e Costa and A. Morton, 2008, “Software packages for multi-criteria resource allocation,” 2008 IEEE International Engineering Management Conference, IEEE Operations Center, Piscataway, N.J. (ISBN 978-1-4244-2288-3).
- A. Salo, J. Keisler and A. Morton (eds.), 2011, “Portfolio Decision Analysis: Improved Methods for Resource Allocation,” Springer International Series in Operations Research & Management Science, Vol. 162, Springer, New York, p. 409 (http://www.springer.com/business+%26+management/operations+research/book/978-1-4419-9942-9).