Dr Lawrence Phillips, Visiting Professor of Decision Science at the London School of Economics and a director of Facilitations Ltd, is a leading expert on ways in which organizations can improve their decision-making. He has long been fascinated by the challenges associated with deciding how best to deploy limited resources across a range of possible projects and getting people to buy into the outcome—the quintessential budgeting and indeed management problem. He teaches decision science to graduates and undergraduates at the LSE and conducts training courses on decision science and facilitation skills to external organizations. His expertize is in applying a wide variety of approaches, particularly decision and risk analysis, to issues of strategic and operational management, option evaluation, prioritization, resource allocation, and crisis management. In November 2005, Dr Phillips was awarded the Frank P. Ramsey Medal for distinguished contributions to decision analysis by the Decision Analysis Society of INFORMS.
Traditional Budgeting and Alternatives
Every organization has finite resources and must, therefore, assign priorities to a range of possible options when allocating its budget. However, in my experience, very few organizations do this particularly well.
The traditional approach to budgeting invariably leads to “silo decisions,” in which resources are allocated on a project-by-project basis. The individual judgments that are being made preclude any coherent analysis of the wider options available, often resulting in missed opportunity for the organization concerned. However, by applying the principles of decision analysis, it becomes possible to create a portfolio of options that really do make the best use of the available resources.
Budgeting is generally an exercise in balancing costs, benefits, and risks. It involves persuading a wide constituency of stakeholders to sign up to these decisions. In the process, multiple stakeholders with different agendas will compete for limited resources. A classic example is the UK government spending round, in which different spending departments slug it out for a slice of the Treasury’s pie.
Resource allocations that are optimal to the individual organizational units are rarely collectively optimal, and those who are dissatisfied with the outcome can become jaundiced and resistant to implementation. In this article, I explain three current approaches to resource allocation, taken from the worlds of corporate finance, operational research, and decision analysis; the latter is one of my expertizes. I draw heavily on an earlier paper coauthored with Carlos A. Bana e Costa in 2007, entitled “Transparent prioritisation, budgeting and resource allocation with multi-criteria decision analysis and decision conferencing.”
What I want to sketch out is a technical process, multi-criteria portfolio analysis, which makes it possible to balance the conflicting elements, and a social process, decision conferencing, which ensures that all relevant players are engaged in the modeling process, ensuring their ownership of the model and their satisfaction when it comes to implementation.
The essence of much decision-making, including budgeting, is that, when presented with a large number of opportunities, decision-makers have insufficient knowledge of each option to be able to make informed choices. Another problem is that the benefits associated with these opportunities are typically characterized by multiple objectives, which themselves often conflict.
A Delicate Balance
What is needed is an approach that enables decision-makers to balance costs, risks, and multiple benefits; to construct portfolios of investments across different areas to ensure that collective best use is made of the limited total resource; to consult the right people in a structured, coherent way, so that their multiple perspectives can be brought to bear; and to engage with key players to ensure they are aligned with the way forward, while preserving their individual differences of approach.
That may sound a tall order. But it can be accomplished by blending a technical solution that captures the differing perspectives with a social process that engages with the people concerned. The technical solution I am suggesting is multi-criteria decision analysis (MCDA), blended with decision conferencing.
We need to distinguish between resource allocation, which is done by a manager bearing the responsibility for this allocation (ultimately the responsibility is borne by the board in a public company) and resource prioritization, which can involve input from multiple line-management functions. It is also useful to distinguish between two prioritization tasks: the appraisal of options, and the construction of portfolios.
The former orders options within an area, the latter refers to the appraisal of options across multiple areas (a portfolio of options), with the aim of finding the best combination of options for a given level of resource.
The three main perspectives on portfolio resource allocation decisions are derived from the worlds of corporate finance, operations research optimization, and decision analysis. Each places a different emphasis on how benefits, costs, and risks can best be handled.
In the corporate finance perspective, it is assumed that benefits are expressed in monetary terms and that the appraisal of a project’s worth is determined by calculating its net present value (NPV). One of the commonest budgeting techniques is to select from a universe of options by ranking them in terms of highest to lowest NPV, and assigning the budget to options until the resource reaches zero.
It seems logical and it is widely deployed in the finance departments of major corporations. But this approach conceals a schoolboy howler. The essence of the howler is that the method uses a simple “screen” that masks out the possibility that other budget combinations, based on a ratio of NPV over cost, will deliver far better returns.
There is no substitute, in short, for carrying out a scoring and weighting analysis of all the options. In all cases with finite budgets, which effectively means in all budgets, the appropriate criterion is not just a positive NPV (or highest ranking NPV), but rather the ratio of NPV to the investment cost. This is a profitability index that represents value for money.
The Binary Knapsack
Unlike the corporate finance perspective, the optimization perspective of operations research takes the “binary knapsack” approach. This casts the problem as one of maximizing the sum of the benefits of all investments subject to the constraints of the budget. Each chosen project is put into the “knapsack” and when this is full, it’s full. The challenge is to fill it with the most valuable projects. The hidden “risk” in this approach is that in the real world, it is more realistic to think of the options as varying degrees of funding the projects, not as “go, no-go” alternatives for each project. In this way, allocating more resources to the more promising projects can be accomplished by spending less on the projects that are characterized by lesser opportunities.
The third perspective, decision analysis, comes in two flavors. In the first flavor, each project’s risks can be modeled using “decision trees,” as can the possible future decisions. Typically, NPV is used to provide a “score” with the discount rate set as “risk-free.”
Why? The answer is simple and straightforward. All the uncertainties about future events are intended to be fully modeled as probabilities in the decision tree (what else is it?). This point is missed again and again, even by academics who really ought to know better.
I have lost count of the number of times that I have found people building risk into NPV discount rates in decision trees. Uncertainty about the future is better modeled by probabilities of later events, followed by downstream decisions, a true options analysis that provides expected (weighted average) monetary values as the basis for valuing the options, and, when divided by costs, providing indices for constructing portfolios.
The second flavor relies on multi-criteria decision analysis for placing values on the consequences of the options. In this way, monetary and nonmonetary values can be incorporated in the model, with common units of added-value across all the criteria. Thus, it is not necessary to use money as the common unit, easing the process of valuation. Risks here often become criteria, which many people find congenial because they think of risks as negative values rather than as probabilities.
All three approaches, financial modeling, operations research, and decision analysis, conform to the principle that the correct basis for prioritization in budgeting, the one that ensures that the best value is obtained for the available resource, is risk-adjusted benefit divided by cost.
Weakness Remains Widespread
However, in practice, I and the coauthor of the original paper, who have amassed some 35 years of experience working with organizations between us, have not once come across an organization that really makes this principle work in practice.
What they actually do amounts to a variant of the following five steps:
List the projects they wish to support;
Determine the benefit that each project is expected to create;
Rank the projects from most to least benefit;
Associate a forward cost to each project;
Go down the list, choosing projects until the budget is exceeded.
In short, projects are prioritized on the basis of benefits only. It is easy to show that this does not make the best use of the budget and that choosing on the basis of the benefit-to-cost ratio is always a better way of maximizing the total benefit.
One further important point needs to be made. Insofar as future benefits are uncertain, then the benefits should be risk-adjusted. In decision theory, this is accomplished by multiplying the benefits by the probability of realizing them, a necessary step to ensure consistency of preference between projects with different benefits and probabilities of success.
To be useful to decision-makers, models need to be able to accommodate all of the following: financial and nonfinancial benefit criteria, risk and uncertainty, data, and judgment. They should also be transparent (in a way that real-option analysis, for example, most certainly is not), and should provide an audit trail which enables a review, after the fact, of how particular decisions were reached and what the underlying assumptions were. Decision analysis does all this exceptionally well.
Allow me to provide a perhaps grossly oversimplified example of decision analysis.
A tossed coin has a 50–50 probability of landing heads or tails up. If I offer you a ticket that entitles you to one toss of the coin to win £10,000, you will judge yourself to have a one in two chance of winning. A risk-adjusted view of the benefit conferred on your ticket would value it at £5,000. If we say that the ticket is transferable, it immediately has a value, probably some value less than £5,000, unless the buyer happens to be a Muggins.
How much would you sell that ticket for, trading off a potential win which you only have a 50–50 chance of securing, in return for real cash in your pocket now? The answer you give places a cash weighting on the decision, and indeed the difference between your minimum selling (or reserve) price and £5,000 gives a measure of your risk aversion. If we imagine a collection of people with a stake in the ticket, then their collective answer provides a group view of the value, which would probably be different from most of the individual views, and this gives us the basis for a decision conference. Note that the whole process builds in the possibility that the coin toss will not be favorable and so the consequence of the coin toss will turn out to be valueless in reality when that moment of truth arrives.
Today decision conferences, where all the key players in a decision process come together to work on an issue of concern to their organization, are becoming best practice in many organizations. They create a model of their decisions, which always involves assigning values to the consequences of the options and taking account of uncertainty that the consequences may or may not occur.
It is a highly effective way of ensuring that everyone buys into the final decisions that inform a budget. Assisted by a decision conference facilitator, who is a specialist in decision analysis, and whose job it is to help people in how to think about the issues, and not what to think, the process aims to get the whole group thinking more clearly about the issues involved.
The model represents the collective view of the group at any point during its generation and modification, and serves as a way of examining the impact of differences in perspective or vagueness in the data. Everyone should understand the formation of the model as it is constructed, so that they can see and understand the impact of their own participation in it. In practice, we have discovered that decision conferences are a tremendous way of getting organizations to think creatively and to move away from merely pumping resources into maintaining or slightly adjusting the status quo.
It generally takes around two weeks for senior staff to digest the results of the decision conference fully. For larger problems, a sustained workshop approach is often required, utilizing several decision conferences along with workshops, interviews, and individual meetings. But one outcome is certain—better decisions, and decisions in which the participants really believe, will be achieved.