Improvement accelerates when leaders stop asking how to optimise everything and start asking what actually governs the result.
Operational systems contain hundreds of variables. That does not mean hundreds of variables deserve equal management attention.
The strategic challenge is to identify the small number of conditions that determine whether the system succeeds, where additional effort stops producing value, and which interfaces can destroy otherwise strong component performance.
Without that discipline, optimisation becomes expensive activity. Teams improve variables that are not limiting the outcome, pursue local maxima that weaken another objective, or keep pushing a parameter after the system has already reached a practical plateau.
The question is not simply, "What can be improved?"
It is:
What variable, threshold or interface currently governs the outcome we care about?
The Strategic Context
Several papers in the supplied research set illustrate this principle from different engineering systems.
Peng and colleagues developed a mathematical model for arsenic and iron immobilisation in a denitrifying granular biofilm process. Their modelling indicated that granule size had limited impact over the conditions studied, while hydraulic retention time had a strong effect as it increased from one to twelve hours. Beyond that range, further increases produced little additional improvement in removal.
This is a classic saturation pattern. If a leader looked only at the fact that more retention time improved performance, the intuitive response could be "more is better". The more useful conclusion is that retention time matters until it stops mattering.
Jusoh and colleagues provide a different example in composite membranes for CO2/CH4 separation. The study was not simply about choosing a stronger polymer or a more selective zeolite. Performance depended heavily on the interface between the two. Some surface treatments improved interfacial adhesion and separation behaviour; another did not. The system outcome depended on compatibility across the boundary between components.
Lai and colleagues studied microbial fuel cells for treating the azo dye Acid Orange 7 while generating electricity. Their work adjusted membrane composition and biological cathode conditions to improve proton transfer, water retention, pollutant removal and power generation. Again, the result emerged from interaction among components rather than one isolated material property.
Ashok and colleagues tested two additives in a biodiesel-fuelled compression-ignition engine. Titanium dioxide nanoparticles improved brake thermal efficiency and reduced some emissions, while the antioxidant BHT reduced NOx more strongly but increased hydrocarbon and carbon-monoxide emissions under the reported conditions. There was no single additive that was "best" independent of the objective.
Together these studies support a broader management principle:
The governing constraint is defined by the outcome, the operating condition and the interfaces in the system. It can move.
What Leaders Commonly Misread
The first error is optimising the most visible variable rather than the limiting variable.
If an operation is late, management may increase labour. If a machine produces defects, engineering may tighten tolerances. If a project is delayed, leadership may add reporting. If an IT service is slow, infrastructure may be expanded.
Each action can be rational in isolation. None is guaranteed to touch the true constraint.
The second error is treating averages as sufficient evidence.
A process with acceptable average performance can fail at a particular interface, load condition or transition state. The membrane study is a useful physical analogy: excellent ingredients do not automatically combine into an excellent composite. The interface can dominate.
The third error is assuming the optimum is a point rather than a region.
In real systems, exact optima are often unstable because inputs vary. Raw materials change. Demand changes. Weather changes. Operators intervene. Equipment ages. A solution that performs brilliantly at one laboratory or modelled point may be fragile in service.
Leadership therefore needs to understand the performance envelope, not merely the best observed value.
Reframing the Issue
A governing constraint can take at least four forms.
A physical constraint
A hard limit imposed by geometry, heat transfer, flow, material strength, residence time, storage, power or another physical mechanism.
An interface constraint
A loss created where two components, functions, organisations or systems interact. Examples include filler-polymer adhesion, supplier-customer data quality, handover between project and operations, or incompatible incentives across business units.
A policy constraint
A rule that restricts the system even when the equipment could perform differently. Batch sizes, approval levels, shift patterns, procurement thresholds and scheduling conventions can all become policy constraints.
An information constraint
A decision is delayed or degraded because the right state information is unavailable, inaccurate or arrives too late.
These categories matter because the remedies differ. Adding equipment to solve an information constraint is rarely effective. Training alone will not remove a hard physical limit. Improving one component may not fix a weak interface.
The Value of a Saturation Curve
Executives do not need to become process modellers, but they should ask teams to show how the system responds as a key variable changes.
The shape of that response contains strategic information.
If performance rises steeply with a small increase in a variable, investment there may have high leverage.
If the curve is flat, further effort may be waste.
If the curve peaks and then falls, aggressive optimisation may damage the system.
If the curve changes under different operating conditions, the organisation may need adaptive control rather than one fixed target.
The arsenic study's hydraulic-retention-time result is useful because it demonstrates the principle clearly. The important management insight is not the specific twelve-hour value, which is study-specific. It is the need to locate the point after which additional input no longer creates a proportionate outcome.
Every major operational improvement program should be asking some version of:
Where does the response flatten?
Interfaces Often Matter More Than Components
Organisations are naturally structured around components: departments, assets, projects, vendors and accountabilities.
Failures often emerge between them.
The composite-membrane study demonstrates this physically. A filler may have attractive separation properties and a polymer may have attractive manufacturing properties, but poor interfacial adhesion can create voids and undermine the combined membrane.
The organisational equivalent is common.
A strong sales function can create poor enterprise performance if its incentives create unmanageable delivery variability. A well-designed project can fail to realise benefits if the operational handover is weak. A sophisticated AI model can create poor decisions if the surrounding data, workflow and escalation logic are immature.
This leads to a practical rule:
When all the individual parts appear capable but the system still underperforms, inspect the interfaces before replacing the parts.
Multi-Objective Constraints
Sometimes the governing constraint is not a physical bottleneck at all. It is a trade-off between objectives.
The biodiesel-additive study is a good example. BHT reduced NOx more than the tested TiO2 formulation, but the reported effect on HC and CO moved in the wrong direction. If the objective were only NOx reduction, one decision might follow. If the objective included efficiency, multiple emissions and durability, the choice could change.
The lesson for executives is that optimisation requires an explicit hierarchy of outcomes.
If leaders have not agreed what must be protected, teams will optimise whichever metric they own.
That is how organisations become locally efficient and globally incoherent.
Decision Framework
ERANORTH recommends a Constraint and Threshold Review for complex improvement decisions.
1. Define the outcome variable
Specify what success means in measurable terms. Avoid vague targets such as "improve performance".
2. List plausible governing variables
Use process knowledge, data and frontline observation. Include physical parameters, policy rules, information delays and interfaces.
3. Test sensitivity
Ask how much the outcome moves when each candidate variable changes. A variable that can change significantly without changing the outcome is unlikely to be the governing constraint at that operating point.
4. Locate thresholds and plateaus
Identify minimum viable levels, saturation points and failure regions. These are often more useful for investment decisions than a theoretical optimum.
5. Test interfaces
Examine handovers, coupling, adhesion, compatibility, dependencies and feedback loops. A weak connection can nullify high component performance.
6. Optimise across objectives
Define which outcomes are mandatory, which are tradeable and which are merely desirable. Do not allow one metric to silently dominate.
7. Re-test under different conditions
Change demand, feedstock, temperature, user behaviour, supplier availability or other relevant conditions. A constraint that moves should be managed dynamically.
From Strategy to Execution
Immediate action: require improvement teams to state the hypothesised governing constraint and the evidence supporting it before major spending is approved.
Medium-term capability building: strengthen sensitivity analysis, experimental design, process modelling and interface ownership. Give operational teams access to data that reveals response curves rather than only monthly averages.
Long-term strategic positioning: design systems with margin around critical constraints. Avoid architectures that perform well only at one narrow operating point.
Related article: Before You Build More Capacity, Change the Control Logic
Related article: Environmental Decisions Need Confidence Ranges, Not Just Precise Scores
Signals to Monitor
Look for improvement spending that rises while the enterprise outcome remains flat; teams optimising metrics with weak causal links to value; large performance differences between apparently similar sites; recurring failures at handovers; operating targets that remain fixed despite changing conditions; repeated attempts to improve component performance when the system bottleneck is elsewhere; and diminishing returns from successive improvement projects.
Another warning signal is the phrase "we have already optimised that" without evidence of the operating range under which the claim is true.
Questions for the Leadership Team
- What variable currently governs the outcome, and what evidence supports that conclusion?
- Where does additional investment stop producing a material improvement?
- Are we managing a hard physical limit, a policy limit, an information limit or an interface failure?
- Which system interfaces receive less management attention than the components they connect?
- What outcome are we sacrificing when we optimise the metric currently receiving the most attention?
- Does the identified constraint remain the same under peak demand, degraded inputs or failure conditions?
- How much operating margin exists between normal performance and the point where the system becomes unstable?
Closing Perspective
Complexity tempts organisations to spread attention across everything. Good systems thinking does the opposite. It concentrates attention where the system is most sensitive.
The governing constraint is not always the largest machine, the most expensive component or the most visible problem. It may be a threshold, an interface, a rule or a variable that matters only under certain conditions.
Leaders create leverage when they identify that point before they spend heavily trying to optimise the rest.
Source basis: This article is an original ERANORTH synthesis principally informed by Peng et al. (2017), Enhancing immobilization of arsenic in groundwater: A model-based evaluation; Jusoh et al. (2017), Fabrication of silanated zeolite T/6FDA-durene composite membranes for CO2/CH4 separation; Lai et al. (2017), Enhanced bio-decolorization of acid orange 7 and electricity generation in microbial fuel cells with superabsorbent-containing membrane and laccase-based bio-cathode; and Ashok et al. (2017), Experimental studies on the effect of metal oxide and antioxidant additives with Calophyllum Inophyllum Methyl ester in compression ignition engine, all published in Journal of Cleaner Production, volume 166.
About EraNorth Insights
EraNorth Insights publishes practical analysis on strategy, projects, operations, transformation and decision intelligence for professional and organisational use. About EraNorth.