Systems thinking becomes politically useful only when it helps answer a practical question: where should we intervene? The answer is not always where the problem is largest or most visible. A severe problem may be the downstream consequence of pressures located elsewhere, while a modest constraint may be limiting the effectiveness of several programs at once. This is why complex systems can absorb enormous amounts of effort without changing direction, then respond significantly when a comparatively small but influential relationship changes.
Two ideas are particularly useful and should not be confused. A bottleneck limits what a system can achieve. A leverage point is a place where intervention can produce a disproportionately large change in system behaviour. Sometimes the same feature is both; often it is not. A shortage of transport may be a bottleneck preventing people from reaching employment, education and health services, while the rules determining access, the information available to users or the authority controlling transport decisions may offer stronger leverage because changing them alters how the system behaves rather than merely increasing one input.
Donella Meadows made this distinction influential through her work on leverage points. She argued that changing numerical parameters such as subsidies, taxes, standards or targets can matter, but interventions affecting information flows, rules, institutional power and system goals may exert considerably greater influence. She also warned against turning the idea into a formula, describing her list as an invitation to think more broadly rather than a recipe for controlling complex systems.
The distinction is easy to recognise in public administration. Governments can keep adding funding to a service whose operating rules prevent it responding effectively. They can create more programs in a community that does not trust the institutions delivering them. They can set performance targets that improve reported numbers without changing the outcome those numbers were meant to represent. Sometimes resources are clearly the constraint, and systems thinking should never become an excuse for pretending a material shortage is really a problem of psychology or coordination. If a regional hospital lacks enough nurses, it needs nurses. But in other situations, additional resources enter an unchanged structure and therefore produce less than expected.
IEP’s Positive Peace research approaches part of this problem through constraints analysis. Rather than assuming every factor matters equally in every context, it asks which conditions limit progress elsewhere. The report finds that the relative importance of its Positive Peace Pillars changes according to the state of the wider system. For countries with low Positive Peace, Sound Business Environment and Good Relations with Neighbours tend to be more influential; in the middle range, Well-Functioning Government, Acceptance of the Rights of Others and Low Levels of Corruption become especially important; among high-Positive Peace countries, Equitable Distribution of Resources and Well-Functioning Government carry greater relative importance.
The important finding is not the ranking itself but the fact that constraints change with context. Imagine a region with good schools, reasonable health services and employers seeking workers, but unreliable transport. Education and employment programs may both be worthwhile, yet transport may continue limiting participation across several areas. Removing that bottleneck could increase the value of investments already made elsewhere. Institutional systems behave similarly. A government can invest in infrastructure, business development and skills while corruption or administrative unreliability prevents those investments delivering their expected benefits.
A bottleneck tells us what is constraining the system. It does not automatically tell us how to change it. This is where systems analysis requires intellectual discipline. A diagram connecting housing, health, trust, employment, education and political attitudes can look impressive while proving very little. The existence of a possible relationship does not establish its direction, strength or significance. A system map is a hypothesis.
If poor transport is thought to suppress employment, does improving transport increase sustained workforce participation? If distrust is believed to reduce service uptake, does delivering the same service through a trusted local organisation alter behaviour? If an eligibility rule appears to produce repeated exclusion, does changing the rule reduce downstream demand elsewhere? Intervention should test our understanding of the system rather than merely enact it.
That distinction separates systems thinking from systems storytelling. The first is willing to discover that the map was wrong.
The World Health Organization has reached a compatible conclusion in health systems. Its systems-thinking work argues that financing, workforce, medicines, information, governance and service delivery cannot be adequately improved as independent building blocks. Instead it asks how the whole system will respond, placing particular emphasis on stakeholder knowledge, unintended consequences and iterative learning.
Administrative data can reveal where people are not using a service without explaining why. Employment statistics can show vacancies alongside unemployment without revealing that shift times and public transport are incompatible. Education data can record chronic absence without capturing unstable housing, caring responsibilities or bullying. People inside a system often possess knowledge unavailable to the institution observing it. The Halo process therefore combines quantitative information with stakeholder deliberation, integrating local knowledge, theory, data and modelling to test assumptions about how the parts interact.
The answer may be another program. It may also be a rule change, better information, coordination between existing services, a transport link or a transfer of authority to an institution that already possesses local legitimacy. Elinor Ostrom’s work on polycentric governance is relevant because it challenged the assumption that complex collective problems must always be solved through either central government control or markets. Her research showed the importance of institutional diversity and decision-making at multiple scales. A potential leverage point may therefore involve giving government greater authority, but it may equally involve government relinquishing some.
Power matters because systems are not neutral mechanisms. They contain interests and unequal capacity to influence rules. Identifying a bottleneck does not mean everyone has an equal interest in removing it. Some actors may benefit from the existing arrangement, while others may understand the problem clearly but lack authority to alter the rule responsible for it. Meadows’ attention to information flows, rules and the power to change system structure is important for precisely this reason. A serious systems analysis therefore asks not only what maintains the current outcome, but who has the authority to change it and who has incentives to resist.
No amount of mapping removes uncertainty. Human societies adapt, reinterpret policy and generate responses that cannot be fully predicted. IEP acknowledges this directly, describing Halo as a means of improving understanding and testing assumptions rather than extracting immutable social laws. A realistic model of intervention is therefore iterative: diagnose the system as carefully as possible, identify a plausible bottleneck or leverage point, intervene at a scale proportionate to the evidence, monitor effects beyond the immediate target and revise the model as the system responds.
The purpose is not to discover a magical lever capable of solving complex social problems. It is to improve judgement about where intervention is likely to matter while remaining disciplined enough to discover when our assumptions were wrong.
