Donella Meadows spent her career watching well-meaning interventions miss. Pollution caps were tightened, subsidies were tweaked, taxes raised — and the system shrugged. By the late 1990s she had compiled a hierarchy of places to push, ordered roughly from least to most effective. The list is uncomfortable because the leverage almost everyone reaches for first sits near the bottom.
The hierarchy, bottom to top
12. Constants, parameters, numbers. Tax rates, minimum wages, interest rates. Tuning a number rarely changes a system’s behavior; it only shifts where the same dynamic plays out. A tighter speed limit doesn’t redesign the road.
11. The size of buffers. A larger inventory absorbs shock but slows response. Shrink the buffer and the system gets nimble — and brittle. Trade-off, not improvement.
10. The structure of stocks and flows. Pipes, plumbing, supply chains. Hard to change after the fact (a city’s road grid outlives every administration that paves it), so the leverage is mostly in the original design.
9. Lengths of delays. Information delays warp every loop. A thermostat that reads the temperature five minutes late will overshoot every time. Faster sensing rarely hurts; slower acting often helps.
8. Strength of negative (balancing) feedback loops. A balancing loop can only counteract a disturbance if it’s faster and stronger than the disturbance itself. Every safety system, every regulator, every compensator lives here.
7. Gain of positive (reinforcing) feedback loops. Reduce the gain and runaway growth slows. The leverage is in the gain, not the loop’s existence — you usually can’t remove a reinforcing loop, only damp it.
6. Information flows. Who can see what, when. Holland’s classic example: posting electricity meters on the outside of houses (where the resident sees them daily) instead of the basement reduced consumption ~30 percent without any new policy or price change.
5. The rules of the system. Incentives, constraints, who gets to vote, who counts as a stakeholder. Rules shape what the players do; changing rules changes the game without touching any individual.
4. The power to add, change, evolve, or self-organize system structure. A system that can rewire itself outpaces every system that can’t. Resist the urge to over-specify; preserve the capacity for local adaptation.
3. The goals of the system. Goals dominate rules. Change a corporation’s goal from “maximize quarterly earnings” to “maximize stakeholder welfare over fifty years” and watch every rule below it bend.
2. The mindset or paradigm out of which the system arises. What the culture takes as obvious. Growth is good. Nature is property. Time is money. Paradigms are sticky; they shift not by argument but by accumulated anomaly until the old frame collapses under its own contradictions.
1. The power to transcend paradigms. Holding any paradigm — including the one you currently believe — lightly enough that you can step out of it. The most powerful and the rarest leverage point.
Why this matters for engineers
Most engineering interventions are 12 through 8. We adjust thresholds, add buffers, tighten loops. These changes are tractable; they fit in a sprint. The higher-leverage moves — changing information visibility (6), changing the rules (5), changing the goal (3) — usually aren’t even in the engineer’s authority. They belong to product, leadership, or culture. So engineers reach for what they can touch, and the system stays the same.
Two practical implications:
- Before tuning a parameter, ask which loop you’re acting on and whether the loop’s structure is the actual constraint. Half the time, the answer is no.
- When you do have authority over information flow or rules — a dashboard, an alert threshold, an escalation policy, an SLO — take it seriously. That’s higher leverage than weeks of code optimization.
Caveats
Meadows herself listed the points “in increasing order of effectiveness” and then immediately undercut the ordering: “The higher the leverage point, the more the system will resist changing it; that’s why societies often have to overturn paradigms.” Higher leverage is not strictly better — it is more powerful and harder to apply. A paradigm shift attempted clumsily produces only resistance.
The list is also a thinking tool, not a checklist. Real systems sit in many of these levels at once; an intervention rarely lands cleanly at one number.
Sources
- Meadows, Donella. Leverage Points: Places to Intervene in a System. The Donella Meadows Project, 1999. https://donellameadows.org/archives/leverage-points-places-to-intervene-in-a-system/