01 — Orientation
You already run on systems
You already operate inside systems, whether or not you have ever called them that. A morning routine, a team, a budget, an inbox: each is a set of parts connected in a way that produces a behaviour. Systems thinking is just the habit of looking at the connections and the behaviour, not only the parts.
Parts are the easy bit
Behaviour over blame
02 — Anatomy
Parts, links, purpose
A system is three things at once: a set of parts, the connections between them, and a purpose. Miss any one and you are not looking at the system, only a piece of it. The purpose is the easiest to overlook and the most revealing, because a system’s real purpose is shown by what it actually does, not by what it says it is for.
2.1
Parts
The visible components. Swappable, and usually where attention lands first. Necessary, but rarely where the behaviour lives.
2.2
Connections
How the parts affect each other: flows of material, money, information, influence. This wiring is where most behaviour is decided.
2.3
Purpose
What the system is actually aiming at, read from its behaviour. A system that reliably produces delay is, in effect, aiming at delay.
2.4
The whole
Emergent behaviour that no single part has on its own. You cannot find it by studying the parts in isolation.
03 — Stocks & flows
What accumulates, what moves
The most basic unit of any system is a stock: something that accumulates. Money in an account, water in a bath, tasks in a backlog, trust in a relationship. Stocks change only through flows, the rates at which things move in and out. Get this one distinction and much of the rest follows.
Stocks are the memory
Flows are the levers
Figure 01
A stock and its flows
Level changes only when a flow runs
Act on the valves, not the box. A stock sits between an inflow and an outflow. The level rises or falls only when the flows differ; the box only ever reflects what the valves have done.
04 — Feedback
Feedback loops
Systems act on themselves through feedback: an output loops back to influence the input that produced it. There are only two kinds, and almost every system behaviour you will ever meet is some combination of them. Learn to spot which loop is running and you can predict what a system will do next.
4.1
Balancing
A loop that seeks a target and corrects toward it. A thermostat, a stock replenished when it runs low. Balancing loops give stability and resist change, including change you want.
4.2
Reinforcing
A loop that amplifies itself. Interest on savings, word of mouth, a rumour. Reinforcing loops produce growth or collapse, and they compound, quietly at first.
4.3
Which is running
Ask whether the system is settling toward something or running away from it. Settling is balancing; running away is reinforcing.
4.4
Loops in pairs
Real systems run several loops at once, often pulling against each other. Behaviour over time is the net of whichever loop is currently winning.
Figure 02
Balancing and reinforcing
Two loops, every behaviour. Reinforcing (R) compounds and produces growth or collapse; balancing (B) seeks a target and self-corrects. Every system behaviour is some mix of these two, running together.
05 — Delays
Why systems mislead
Between a cause and its effect there is almost always a delay, and delays are where systems fool people. When feedback arrives late, the natural response is to overcorrect, which sends the system swinging the other way. Much of what looks like instability is really a delay that nobody accounted for.
Delays hide cause
Overshoot and oscillate
Figure 03
Overshoot
A delayed response overshoots before it settles
Late corrections swing. A delayed system chasing a target corrects late, so the response overshoots and swings before it settles. Smaller, slower moves damp the swing; harder pushes amplify it.
06 — Constraints
Bottlenecks and constraints
In any chain of steps, one stage sets the pace for the whole. It is the bottleneck, the constraint, and it is almost never where the noise is loudest. Improving any stage other than the constraint does not speed up the system; it just piles work in front of the real limit. This is where the fundamentals start to earn their keep on real processes.
6.1
Find the limit
Look for where work piles up and everything downstream waits. That queue marks the constraint, wherever the complaints happen to be pointing.
6.2
Feed it first
The constraint should never be idle or starved. An hour lost at the constraint is an hour lost for the whole system; an hour lost elsewhere is usually free.
6.3
Improve there
Effort spent easing the constraint moves the whole system. Effort spent anywhere else improves a number that was never the problem.
6.4
Watch it move
Relieve one constraint and a new one appears somewhere else. That is success, not failure; find the new limit and repeat.
Figure 04
The constraint
90/hr
90/hr
40/hr
Constraint
90/hr
Throughput is set by the narrowest stage, not the sum
The narrowest stage sets the pace. Four stages, one narrow. Throughput is set by the slowest stage, not the average, so widening any other stage changes nothing. Find the narrow point before you optimise anything.
07 — Leverage
Leverage points
Not all interventions are equal. Some points in a system, when nudged, barely move it; others shift the whole thing. These are leverage points, and the counter-intuitive lesson, drawn from Donella Meadows, is that the places we most often push, the numbers, are the weakest, while the places that would move everything, goals and mindset, are the ones we rarely touch.
Weak: the numbers
Middle: the structure
Strong: the goal
Figure 05
The leverage ladder
Aim higher up the ladder. Leverage rises from numbers at the bottom, through buffers, flows, loops, information and rules, to goals and paradigm at the top. We push low and wonder why little moves.
08 — Resilience
Designing a process that holds
Once you can see stocks, flows, loops and constraints, you can design a process that holds up rather than one that merely works on a good day. A robust process is not the fastest one; it is the one that degrades gracefully when something goes wrong, because something always does.
8.1
Design for the bad day
Plan around the failure, the absence, the spike, not the calm average. A process that only works when everything is fine is not finished.
8.2
Keep some slack
A system run at one hundred percent has no room to absorb a shock, so every shock becomes a crisis. Slack looks like waste and behaves like resilience.
8.3
Name the failure modes
List the few ways this will realistically break, and decide the response in advance. A known failure with a plan is an inconvenience, not an emergency.
8.4
Prefer buffers to heroics
A small buffer of stock, time, or capacity absorbs variation quietly. Relying on someone to save the day is a design flaw wearing a cape.
09 — Boundaries
Seeing the system you are in
Before you can act on a system, you have to draw its edges, and where you draw them decides what you can see. Draw the boundary too tight and the real cause sits just outside it; draw it too wide and you drown. The boundary is a choice, not a fact, so choose it on purpose.
Map before you touch
Mind the boundary
Figure 06
Drawing the boundary
Draw the line, then decide what stays outside it
The line is a decision. A system boundary, drawn deliberately. Parts and their connections sit inside; inputs and outputs cross the edge. What you place outside the line is as consequential as what you keep in.
10 — Traps
Common traps
A handful of mistakes recur across almost every system, and naming them is the fastest way to stop repeating them. Nearly all are the same error in different clothes: acting on a part while ignoring the structure it sits inside.
Optimising the part
Missing the loop
Blaming the person
Ignoring the delay
11 — First actions
The A–Z
A checklist to start from. Run it against any system you are trying to understand or improve, from a workflow to a whole organisation.
A
Name the system
Its parts, connections, and real purpose.
B
Read the purpose from behaviour
Not from the mission statement.
C
Find the stocks
What is accumulating here?
D
Find the flows
What changes each stock, and how fast?
E
Act on flows
Not on the stock by hand.
F
Spot the loops
Is this settling or running away?
G
Name each loop
Balancing or reinforcing?
H
Look for the delay
Between an action and its result.
I
Make smaller moves
Where the response is slow.
J
Find the constraint
Where does work pile up?
K
Feed the constraint first
Keep it from going idle.
L
Improve at the constraint
Not everywhere else.
M
Expect a new constraint
Once you relieve the old one.
N
Climb the leverage ladder
Past the numbers.
O
Ask what produces the number
What loop, rule, or goal produces the number you dislike.
P
Design for the bad day
Not the average.
Q
Keep deliberate slack
Do not run at full.
R
List the failure modes
And pre-decide the response.
S
Draw the boundary on purpose
Then question it.
T
When a fix keeps failing
Work on the structure, not the part.