Rank intervention options by structural depth before choosing

Before deciding where to intervene, map the options from "change a number" to "change the paradigm."

Why it works

Meadows found that interventions divide into levels: numbers/parameters (weakest), stocks and flows, feedback loop structures, rules, information flows, goals, self-organization, paradigms, and the ability to transcend paradigms (most powerful). Most default interventions target numbers — tightening a budget, adjusting a quota, adding a resource — which rarely produces lasting change because the system’s structure generates the behavior regardless of the specific parameter values. Explicitly ranking options by structural depth reveals whether you are solving the problem or just adjusting a number until the next crisis.

How to do it

1. List all the interventions you could make to address the problem.
2. Classify each by what it changes: a number, a flow rate, a feedback loop, the rules, the goal, or the paradigm.
3. Default toward the highest structural level you have access to, while testing feasibility.
4. Ask: "If I could change only one structural thing here, what would make the system produce different behavior automatically?"

Evidence

Meadows’ hierarchy of leverage points is a practitioner synthesis from system dynamics work and organizational consulting; it is widely cited and treated as canonical in the systems thinking community. Empirical support comes primarily from case studies rather than controlled comparisons of intervention types. (mechanistic)

The hierarchy is a heuristic, not a law; some parameter changes in the right place (a tipping point) can be high-leverage, while some paradigm changes are easier to achieve than they appear in theory.

Sources

• Meadows (1999), "Leverage Points: Places to Intervene in a System," Whole Earth magazine / Sustainability Institute

Common mistake

Choosing the most visible intervention rather than the most structural one — visible interventions are easy to defend to others ("we did something") but rarely address the underlying structure.