Reading notes: leverage points, and why the obvious ones are the weakest
Working through Donella Meadows on where to intervene in a system — and where I have been pushing on the wrong lever in my own.
Meadows ranks the places you can intervene in a system, weakest to strongest. The counterintuitive part is that the levers we reach for first — numbers, budgets, targets — sit near the bottom.
Roughly, in my own words, weakest first:
- Numbers. Prices, quotas, headcount. Easy to change, almost never changes behaviour.
- Buffers and delays. How much slack the system holds and how fast feedback arrives.
- Feedback loops. What the system notices about itself, and how hard it reacts.
- Information flows. Who gets to see what. Cheap to change, hugely powerful.
- Rules. Incentives, constraints, who is allowed to decide.
- Goals. What the system actually optimises for, whatever the mission statement says.
- Paradigms. The assumptions everyone shares so completely they can’t see them.
Where I recognised myself
I spent a quarter tuning a number — a conversion target — and got nowhere. The real problem was an information flow: the people who could fix the drop-off never saw the drop-off. They got a monthly summary. Users were leaving on a Tuesday and the team found out three weeks later.
Changing a dashboard did more than changing the target ever did.
The least effective place to push is usually where everyone is already pushing.
Two things I’m sitting with
Goals hide. Every organisation has a stated goal and a revealed one. The revealed goal is whatever survives when the two conflict under pressure. If you want to know what a company optimises for, watch what it protects during a bad month.
Delays are underrated. Much of what looks like bad judgment is a system reacting to information that was already stale. Before blaming the decision, check the lag.
Still unresolved: how much of this transfers to a five-person company, where the “system” is mostly five people and their habits. My working guess is that it transfers almost completely, and the paradigm level is simply the founder’s unexamined beliefs. Which is uncomfortable, which is probably the point.
