foresight

As an organisational leader, it’s inevitable you’ll need to make tough decisions where the stakes are high and there’s no clear best option — just trade-offs in all directions. The chances are that right now you’re mulling over a few tough decisions in your professional or personal life.

Hard-to-make decisions can be a source of anguish — not knowing what to do when the stakes are high is stressful. Most of us will admit to experiencing regular moments of uncertainty about the best option and, from time to time, having no idea whatsoever. This is why decision theory exists. It’s a field of research designed to help people, as individuals and groups, make wiser choices. Decision trees, risk analysis, benefit cost analysis and scenario modelling are among its tools, procedures and techniques.

A key technique is multiple criteria decision analysis (MCDA) and the model is much simpler than it sounds. MCDA is as much a human organisational process as an analytic tool. An MCDA model can be used to rank or score the desirability of a finite set of decision options against multiple criteria. Typically, the criteria are weighted by decision-makers to reflect relative importance and can be measured in different quantitative units (dollars, hectares, tonnes) or qualitative scores (for example, a five-star rating system). The criteria are converted into commensurate units using transformation functions and adjusted by the weights to give each option an overall score and/or ranking.

The construction and application of an MCDA model happens within the context of MCDA decision-making process. This begins with the all-important phase of problem structuring — working out the options, objectives and decision-makers. If these things aren’t done well, no amount of analytics can save you down the track. Studies of organisational decision-making often identify a failure to define and structure the problem as the main shortcoming. It’s worthwhile front-loading your decisions and putting more effort into these early tasks.

Decision analytics

The next phase of the MCDA process is decision analytics. The analyst obtains weights and performance data from the decision-makers, builds the evaluation matrix and transforms the criteria. This is the analytic and data-intensive part of the problem. Often, obtaining, analysing, interpreting and formatting the data are the hardest and most time-consuming tasks within this phase. Data scientists often say, “Garbage in, garbage out” — and it certainly applies to MCDA models. Good data will give you good results.

The final phase of the MCDA process is decision-making. In this phase, ranks or scores are generated for each decision option and sensitivity analysis can be applied. This involves running the model thousands of times with different — sometimes randomly generated — weights and criteria scores to see whether the solution is stable under alternative specifications and identify where better data might be needed. In the final stages, decision-makers review the MCDA model outputs, consider other issues and make a decision.

It is important to see MCDA as a support tool, not a decision-making tool. I’ve built countless MCDA models in my career at CSIRO to help industry and government leaders make better choices — but we have never yet built a model that was perfect. The decision-maker’s intuition and gut feel always play an important role. And there are always factors above and beyond what’s captured in the model. We need to ensure decision-makers have room to disagree with the MCDA model.

Why it’s a good idea

There are thousands of well-documented case studies of MCDA applications within Australia and overseas to assist in decision-making. These range from identifying investment priorities, selecting building sites for new facilities and evaluating alternative acquisition and merger options. It was used in Italy to evaluate the COVID-19 lockdown priority of urban districts. Intelligent application of a MCDA model within a well-designed process will substantially uplift the quality of organisational decision-making and help achieve better outcomes. Some of the benefits include:

  • Transparency People are more likely to accept a decision — even one with negative outcomes if they understand why it was taken.
  • Auditability An MCDA model can be replicated for auditors to quickly show why one option was selected over others. Even if the outcomes aren’t good, the model allows you to defend the integrity and quality of the decision process.
  • Analytic rigour The approach brings data and analytics into the heart of decision-making. This helps you remove red herrings and conjecture, and focus on substantive issues.
  • Improved understanding Building a MCDA model will help you convert ambiguity into structure to understand the decision problem and what you’re trying to achieve.
  • Better outcomes Better decision processes lead to better outcomes. The approach is all about helping you and your company make a rational choice consistent with your objectives.