How the world works
Thinking in Systems
Leverage
Thinking in Systems (2008 · Donella Meadows) — Leverage.
What actually happened?
She used to open a talk by holding a Slinky. One hand cups it from below, the other holds the top; take the lower hand away and the spring drops, bounces and oscillates. Then she asks what made it bounce. Not the hand: the hand only let go. The behaviour was latent in the spring the whole time, and the hand merely released it. Every diagnosis in the book runs on that distinction, and most organisational post-mortems get it exactly backwards by investigating the hand.
Stocks, flows and delay
A stock is whatever accumulates: cash, users, trust, inventory. Flows are the rates in and out. Intuition watches the stock while behaviour is set by the flows and by how long the system takes to respond. Delay is what makes you add more, overshoot, then correct too hard.
Headcount is a stock; hiring rate is a flow; the three to six months before a new person is productive is the delay. Ignore it and you will overhire.
Two kinds of loop, and only two
Reinforcing loops widen any deviation: compound interest, word of mouth, arms races. Balancing loops pull the system back to a target: body temperature, restocking, competition. Sustained growth always has a reinforcing loop behind it, and growth that stops always has a balancing loop that has begun to bite.
When a growth curve flattens, look for the balancing loop that just engaged rather than for the effort that supposedly stopped.
The system traps have names
Meadows catalogues the failure structures that keep recurring: policy resistance, the tragedy of the commons, drift to low performance, escalation, shifting the burden to the intervenor, rule beating, seeking the wrong goal. None of them requires a villain. Each is generated by the arrangement itself.
Drift to low performance is the quiet one: this quarter missed, so next quarter's target comes down a little. Every single step is reasonable.
Bounded rationality
Everyone in a system optimises inside the information they can actually see, and the sum of those local optima is often the worst global outcome. So exhortation fails and replacement fails. What works is changing what people can see when they decide.
Putting server cost per team into a weekly email beats ten meetings about reducing server cost.
How do I use it today?
Pick one problem you have solved repeatedly and that keeps coming back. Instead of asking whose fault it is, ask three things: what is accumulating here and how long is the delay, which loop is currently in charge, and what feedback is missing that would let the person deciding see the consequence immediately. Most chronic problems do not need more discipline. They need a meter moved into the hallway.
Deep read
Read alongside
Further
Lines to keep
A system's behaviour cannot be read off its parts.
People find the leverage point, then push it the wrong way.
We cannot control a system. We can learn to dance with it.