Donella Meadows: you don't control a system, you dance with it

Some time in the 1990s, Donella Meadows was sitting in a meeting about the new global trade regime: NAFTA, GATT, the World Trade Organization. The people around the table were discussing how those rules would change the world, and the longer they talked, the more uneasy she grew. They were arguing over details while the forces that would actually decide the outcome went unnamed. So she got up, went to the flip chart, turned to a clean page and wrote a list: places to intervene in a system, ordered from the weakest to the strongest.

That list became one of the most cited essays in systems thinking. Its fame comes from the order. At the bottom, where nearly everyone pushes, are the numbers. At the top, where almost nobody looks, are the rules, the goals and the way a system thinks about itself.

The list still describes a lot of meetings. A metric will not move, and the answer is another instrument: another dashboard, another OKR, another reorganisation. Underneath sits a belief nobody states: that a complex enough system can be run from a control panel, as long as we find the right lever and pull hard.

A modeller who distrusted models

Meadows (1941–2001) trained as a chemist and a biophysicist, and came to systems by the least romantic door: the computer models that, in the early 1970s, tried to simulate the whole planet. In 1972 she was lead author of The Limits to Growth, the report for the Club of Rome that projected what would happen if population, consumption and pollution kept growing on a finite planet. It sold millions of copies and earned as much scorn as praise. For decades it was the punching bag of people who confused a model with a prophecy.

The work taught her, from the inside, how far our tools for prediction and control reach and where they break. She came from the system dynamics school Jay Forrester had founded at MIT, and ended up distrusting its technological enthusiasm. Her best-known warning is aimed at her own people: whoever falls in love with systems thinking after growing up in the industrial world risks believing that interconnection and computing power are, at last, the key to prediction and control. They are not. Much of what she wrote afterwards is a patient working-out of that no.

She died in 2001, at fifty-nine, with the primer she wanted to write still unfinished. Her colleagues completed it and published it in 2008 as Thinking in Systems: A Primer. It is short and generous, and it is still the best way in.

The primer Meadows left unfinished and her students and colleagues completed. Edited by Diana Wright (Chelsea Green, 2008).

Structure, not culprits

The first thing Meadows teaches is a change of question. When a behaviour keeps repeating (an indicator that will not improve, a team that trips over the same thing, a decision that keeps going wrong), the reflex is to look for whoever is responsible. Who did this? Whom do we replace? She asks something else: what structure produces this behaviour?

Her grammar has three pieces. Stocks are things that fill and empty: cash in the bank, the trust of users, technical debt. Flows feed and drain them. Feedback loops are the circuits through which the level of a stock changes the flows that change it. A reinforcing loop amplifies; a balancing loop steadies. Most of the behaviour that surprises us in a system, whether it shoots up, stalls or oscillates, comes from the way those loops combine, not from anyone's malice or clumsiness.

The consequence is a relief and a burden at once. Changing who sits in a chair rarely changes the outcome if the structure that produces it is intact. In product, a metric that keeps getting gamed is almost never a problem of dishonest people; it is a structure that rewards gaming it. A team that always comes up short on discovery seldom needs more pressure. It needs a different loop, one in which learning early has visible consequences. The question "who failed?" produces reorganisations; the question "what loop sustains this?" produces changes that last.

Numbers are the weakest lever

The flip-chart list grew into Leverage Points: Places to Intervene in a System. A leverage point is a place in a complex system (a company, an economy, a body, a city) where a small shift can produce large changes in everything else. Every intervention is looking for one. Meadows's point is that we almost always look in the wrong place.

At the bottom of her twelve rungs are the parameters: budgets, prices, quarterly targets. They are the most visible points and the weakest; moving a number rarely changes the character of a system. Higher up come the structure of stocks and flows, the strength of the feedback loops, and the flows of information: who knows what, and when.

At the top are the rules of the system (incentives, punishments, constraints), the goals it pursues and, above everything, its paradigm: the shared assumptions from which the system thinks about itself.

The twelve leverage points, from weakest to strongest: the higher the rung, the more force and the more resistance. Based on Donella Meadows, "Leverage Points: Places to Intervene in a System" (1999).

We spend almost all our attention on the lower rungs because they are the ones that let themselves be touched. One more dashboard is, in her hierarchy, one more number. And there is a reason the upper rungs stay untouched, which she put bluntly:

"The higher the leverage point, the more the system will resist changing it — that's why societies have to rub out truly enlightened beings."

— Donella Meadows, Leverage Points: Places to Intervene in a System (1999)

The points that matter most are the ones the system defends hardest, because touching them threatens whoever does well under the current order. Changing a number is easy, and that is why it is weak. Changing the goal an organisation optimises for is hard because it would change many things at once.

Then comes the twist I find most useful. Following an intuition of Forrester's, Meadows noticed that people often find the right lever and push it the wrong way. They subsidise the very thing that causes the harm. They add control to a system that suffers from too much control. Anyone who has watched an organisation answer a coordination problem with another layer of oversight knows the gesture. The lever was right. The direction was wrong.

Optimising a part degrades the whole

Every organisation is tempted to improve the part in front of it. The team tunes its metric, the department defends its budget, each function polishes its stretch of the process. Meadows insisted that optimising a part hard almost always degrades the whole, because the parts are joined by flows and loops that sooner or later send back the bill.

Delays make it worse. Between an intervention and its effect there is usually a lag, and the lag tempts us to overcorrect. The indicator does not react, so we insist and double the dose. When the effect finally arrives, added to every correction we piled up while waiting, the system overshoots and swings. It is the shower with a badly set tap, jumping between freezing and scalding. In digital product, where any change can set off effects nobody was watching, overreacting to a dip in a metric is one of the commonest and most expensive versions of the mistake. Software behaves like a system of systems, with its own stocks, flows and loops, and it punishes the belief that the whole can be run by improving one part at a time.

Dancing is not a method

If the highest point is the paradigm, the most powerful intervention is also the quietest: changing how a group sees. Meadows knew it first-hand. The Limits to Growth was less a forecast than an attempt to shift the paradigm of growth on a finite planet. She also knew that this rung resists most, because a paradigm is not argued. It is lived in.

Her conclusion was not the engineer's, who finds the master lever and reaches for it. Above changing a paradigm, she put the capacity not to hold on to any: to keep the mind loose enough to see that every paradigm, one's own included, is a partial way of looking. That humility did not lead her to sit still. It led her to an image:

"We can't control systems or figure them out. But we can dance with them!"

— Donella Meadows, Dancing With Systems (2001)

It is easy to turn the image into a checklist (observe first, start small, listen to the response) and lose what it says. A dance has no plan you can execute. You lead and you are led, and you only learn the next step from how your partner moves. Good product teams work like that: they do not run a closed plan and they do not control the result. They probe it. What Meadows adds is that this is not a lack of rigour. It is the only rigour available when the system answers back.

A diagnosis without a lever

She had just asked us not to cling to any framework, hers included. Taking her at her word means naming the two places where her thinking strains.

The first is that the framework is superb for diagnosis and much weaker for action. Knowing that a structure produces a behaviour does not tell you which leverage point is within reach, at what cost or how fast. And there is an irony in the hierarchy itself: the highest points (rules, goals, paradigms) are the ones that demand a political and organisational power her analysis nearly always leaves out of the frame. The theory shows where to push and, in the same move, that pushing there is the hardest thing in the world.

The second is a blind spot. Meadows tends to treat systems as objects to understand rather than ground under dispute. But inside almost every system that matters there are people with interests, and they resist a structural analysis when it points at them. It is not only that the system resists through inertia. Some people benefit from its not being understood. The clean vocabulary of stocks and loops can turn oddly innocent in front of that grubby, human part of organisations. Not everything a system hides, it hides through complexity. Sometimes it hides it because it pays someone to.

The page on the flip chart

The same hierarchy explains much of what happens in the companies that make digital products. Changing rituals, tools or the org chart are parameter moves: visible, comfortable, short-lived. Changing the incentives a team works under is a change of rule, and it lasts after any ceremony has evaporated. Telling one from the other is the difference between a reform that takes and one that decorates. And the world full of information and instruments her models anticipated is the same one in which Herbert Simon saw that a wealth of information creates a poverty of attention.

Back to that meeting in the 1990s. Meadows could have taken part in the discussion of details and added her own numbers to the pile. Instead she stood up and named the levels nobody was discussing. The most quoted list in systems thinking was born as an interruption.

The next time a metric refuses to move, it is worth asking which rung the proposed fix sits on. Usually it is the bottom one.