A model can identify the best answer to the problem it was given; leadership remains accountable for whether it was given the right problem.
Executives increasingly encounter optimisation in capital planning, logistics, energy, infrastructure, AI, supply chains and sustainability. The appeal is understandable. Complex decisions contain too many variables for intuition alone, and mathematical methods can test combinations that no committee could examine manually.
But optimisation creates a subtle governance risk. Precision in the answer can make the assumptions behind the answer disappear from view.
A model may minimise cost, emissions or risk. It may balance several objectives. It may account for uncertainty. Yet it cannot independently decide whether cost deserves twice the weight of environmental performance, whether a community impact is an acceptable constraint, whether a ten-year payback is strategically reasonable, or whether a supposedly fixed requirement should be challenged.
Those are management decisions.
The Strategic Context
The 2017 research set provides several examples of increasingly sophisticated decision models.
Perez and colleagues designed a multi-objective approach to locating water-quality monitoring stations. The model did not simply minimise the number of stations. It also sought to improve detection of lower-compliance areas, coverage of affected population and representation of important river stretches. The problem was inherently multi-objective because public value could not be reduced safely to one measure.
Habibi and colleagues' municipal-solid-waste model similarly considered economic, environmental and social objectives while addressing facility location, capacity allocation, transport and technology under uncertainty. The point was not merely to find the cheapest network.
Yousefi and colleagues developed a robust fuzzy model for sustainable supply-chain evaluation that accounted for internal network relationships, management goals and ambiguity. Their work is technically specialised, but strategically it highlights a simple truth: the “best” unit depends on the goals against which performance is judged.
Mohammad Rozali and colleagues' hybrid-power framework adds another critical dimension. Technical performance was screened against economic requirements, including an owner-defined payback target. The model could support a cost-effective configuration, but the acceptable investment threshold still came from human decision-makers.
What Leaders Commonly Misread
The first misreading is that optimisation eliminates subjectivity. It often relocates subjectivity into model design. Objective functions, weights, constraints, scenarios and thresholds encode choices about what matters.
The second is that more objectives necessarily create a more balanced decision. A model with environmental, economic and social criteria can still be poor if the indicators are weak proxies for real outcomes.
The third is that a single optimum should be preferred to a range of good solutions. When assumptions are uncertain or objectives conflict, a portfolio of near-optimal options can be more decision-useful than one mathematically superior point.
The fourth is that models should replace executive judgement. The strongest use of optimisation is usually to expose trade-offs, identify dominated alternatives, reveal sensitivity and test constraints. It should improve judgement, not conceal it.
Reframing the Issue
The strategic question is not “What does the model recommend?”
It is “What value system, assumptions and constraints must be true for the model's recommendation to deserve capital?”
This reframing separates three layers that are often confused.
Analytical optimisation determines the best result against specified objectives.
Managerial choice determines which objectives, thresholds and trade-offs are legitimate.
Governance determines who has the authority to make those choices and how the assumptions are reviewed when conditions change.
The model belongs inside this system, not above it.
The Model Cannot Choose the Objective
A city designing a waste network might minimise cost and emissions while also limiting local visual impact. But leadership must still decide whether the social indicator adequately captures community consequences, whether particular locations are politically or ethically unacceptable, and whether resilience deserves an additional objective.
A company designing a renewable-power system might minimise lifecycle cost subject to a reliability threshold. Yet an executive team may value energy independence, strategic learning or future expansion capacity that the original model does not include.
A water authority can optimise monitoring coverage, but it must first decide what constitutes important coverage. Population, regulatory compliance, ecosystem sensitivity and detection speed are different value propositions.
No algorithm escapes this problem because optimisation is conditional: best, given X.
The governance task is to make X explicit.
Multi-Objective Decisions Reveal Trade-Offs
The strongest value of multi-objective modelling is often not the final ranking. It is the visibility of trade-offs.
If a small increase in investment produces a large environmental improvement, leaders can see that leverage. If reducing emissions further causes cost to rise sharply, they can identify the point at which marginal improvement becomes strategically unattractive. If one facility-location option is slightly more expensive but materially reduces community exposure, the decision becomes a visible value choice rather than a hidden modelling assumption.
This is where Pareto-style reasoning is useful even without presenting mathematical frontiers to a board. Executives can ask which alternatives are clearly dominated and where genuine trade-offs begin.
A decision committee should spend less time debating options that are worse on almost every relevant dimension and more time examining the small number of options that represent different value choices.
Decision Framework
Use a six-part governance test before accepting an optimisation result.
1. Objective legitimacy
What is the model trying to maximise or minimise, and why is that objective strategically legitimate?
2. Metric validity
Do the selected metrics represent the outcomes leadership actually cares about, or are they merely easy to quantify?
3. Constraint integrity
Which constraints are genuinely fixed by law, safety, capacity or policy, and which are inherited assumptions that could be redesigned?
4. Trade-off visibility
Can decision-makers see what is sacrificed when one objective improves?
5. Sensitivity
Which assumptions, weights or thresholds change the ranking materially?
6. Decision ownership
Who has authority to choose among near-optimal alternatives when the difference is value-based rather than mathematical?
The result should be a short decision record that states not only the recommended option, but also the values and assumptions embedded in the recommendation.
From Strategy to Execution
Immediate action is to stop presenting optimisation outputs as self-justifying. Any major model used for investment approval should include a management-readable summary of objectives, constraints, exclusions and sensitivity.
Medium-term capability building requires executives, analysts and domain experts to work together before modelling begins. If the modeller receives a poorly framed objective, technical sophistication will simply optimise the wrong problem more efficiently. Decision design should therefore precede model design.
This is particularly relevant to AI-enabled optimisation. As organisations use machine learning, simulation and automated decision support more widely, the risk of objective-function blindness increases. A model can optimise a proxy with extraordinary efficiency while degrading the real outcome the proxy was meant to represent.
Long-term strategic positioning means building a culture in which models are used to challenge assumptions. The most valuable question may be, “What would the model recommend if we changed the constraint?” rather than, “What is the optimal answer?” This turns analytics into strategic exploration.
Related article: A Sustainable Decision Begins With the Boundary: Why Whole-Life Thinking Changes the Answer
Related article: Business Cases Are Investment Hypotheses, Not Permission Slips
Related article: Decision Quality Depends on How the Portfolio Is Seen
Signals to Monitor
Watch for models whose objective is easier to explain than the business outcome. If the organisation says it is optimising “sustainability” but the model only minimises carbon and cost, the language is broader than the mathematics.
Monitor sensitivity to expert weights. A ranking that changes dramatically when reasonable weights move slightly should not be treated as robust.
Look for fixed constraints that have survived from earlier designs without challenge. Sometimes the most valuable strategic move is not to optimise within a constraint but to redesign the system so the constraint no longer binds.
Also monitor the gap between model performance and realised performance. A model that repeatedly produces technically optimal plans which operations cannot implement is revealing a missing variable, often capability, behaviour, maintainability or organisational capacity.
Questions for the Leadership Team
- What value judgement is hidden inside the objective function?
- Which important outcome is represented only by a weak proxy, or not represented at all?
- Are the model's constraints truly fixed, or simply familiar?
- What trade-off does the recommended option require leadership to own explicitly?
- How sensitive is the ranking to reasonable changes in assumptions or weights?
- Would a near-optimal option create more strategic flexibility, resilience or learning value?
Closing Perspective
Optimisation is one of the most powerful tools available to modern decision-makers, precisely because it can make complex trade-offs visible.
It becomes dangerous only when mathematical precision is mistaken for strategic authority.
The model can tell leaders what follows from the objectives they set. It cannot decide what the enterprise should value, what risks it should accept, or which consequences it has a responsibility to consider. Those decisions remain human, governable and ultimately accountable.
About EraNorth Insights
EraNorth Insights publishes practical analysis on strategy, projects, operations, transformation and decision intelligence for professional and organisational use. About EraNorth.