Thinking in Systems

A book about why the same production problems keep coming back, and what to do about it.

What it's about

A short, clear primer on systems thinking. Everything is described in terms of stocks (things that accumulate), flows (what fills and drains them), and feedback loops that either balance a system or reinforce whatever it is already doing. From those few pieces, Meadows builds up why complex systems surprise us, the traps they fall into, and where it is worth pushing to change them.

What stuck

Behaviour comes from structure. If a system keeps producing the same bad outcome, blaming the latest event is usually wrong. The structure, how the loops are wired, is producing that outcome, and it will keep doing so until the wiring changes.

Delays cause oscillation. When the feedback arrives late, people overcorrect. Anyone who has watched an autoscaler flap or a team swing between "ship faster" and "freeze everything" has seen this.

The traps have names. Policy resistance, drift to low performance, escalation, shifting the burden to an intervenor: once you can name the pattern, you stop being surprised by it. Drift to low performance in particular, where the standard quietly lowers itself to match results, describes alert fatigue better than any incident review I have read.

Leverage points are rarely where the effort goes. Tuning parameters is the easiest change and usually the weakest. Changing the rules, the information flows, or the goal of a system does far more.

What I'd apply

• In incident reviews, ask which loop produced the failure, not only which change triggered it.
• Before adding another alert, look for the balancing loop that is missing. More signal into a loop nobody closes is just noise.
• Watch for goals that quietly drift. If a target has been "temporarily" lowered twice, the system has already chosen.