The Machinery of Systems: Stocks, Flows, Feedback Loops, and Delays

By Pritesh Yadav June 22, 2026 19 min read —

In the first chapter you learned the big idea of systems thinking: structure drives behavior. The way a system is wired — not the people inside it — is what produces most of what we see. That is the headline. This chapter is the engine room. Here you learn the actual parts the machine is built from, and how they fit together to produce growth, stability, surprise, and collapse.

There are only four core pieces. Once you can see them, you can read almost any system — a bank account, a forest, a software team, your own habits. The four pieces are:

Stocks — the things that pile up.
Flows — the rates that fill or drain those piles.
Feedback loops — the way a stock loops back to change its own flows.
Delays — the lags between doing something and seeing the result.

That is the whole alphabet. The rest of systems thinking is just grammar built on these four letters. Let's learn each one slowly, with everyday examples, and then watch them combine to produce the behavior that fools almost everyone.

Key takeaway: Almost every system you will ever care about is built from stocks, flows, feedback loops, and delays. Learn to spot these four, and "complicated" situations start to make sense.

2.1 Stocks: the things that pile up

Let's define the first word carefully, because everything else stands on it.

Stock: Anything that accumulates — that builds up or drains down — and that you could measure at a single moment in time. A stock is a "how much is there right now" quantity.

The classic picture is the water in a bathtub. At any instant you can stop and measure the water level. That level is a stock.

Analogy: Think of a bathtub. The amount of water sitting in it right now is the stock. You could pause time, take a ruler, and read off exactly how much is there.

Stocks are everywhere once you look:

Money in your bank account.
The number of unread emails in your inbox.
Your body weight.
Inventory sitting in a warehouse.
The amount of trust in a friendship.
The carbon dioxide (CO₂) in the atmosphere.
The amount of unfinished, messy code in a software project (often called "technical debt").

Donella Meadows, one of the clearest voices in this field, called a stock "the present memory of the history of changing flows." That phrase is worth unpacking. Your bank balance today is not a fresh fact — it is the accumulated record of every deposit and withdrawal you have ever made. The stock remembers. Your body weight is the memory of years of eating and exercising. Trust between two people is the memory of every kept and broken promise. A stock is the system's storage of its own past.

Key takeaway: A stock is an accumulation you can measure at one moment. It is the memory of everything that has flowed in and out of it until now.

2.2 Flows: the rates that change a stock

A stock just sits there unless something changes it. Those changes are flows.

Flow: A rate at which a stock fills up or drains down. Flows are measured per unit of time — per second, per day, per month. An inflow raises the stock; an outflow lowers it.

Back to the bathtub. The faucet is the inflow — water pouring in. The drain is the outflow — water leaving. The water level (the stock) goes up or down depending on those two flows.

      faucet (inflow)
          |
          v
     +-----------+
     |  ~~~~~~~  |   <- water level = STOCK
     |  ~~~~~~~  |
     +-----------+
          |
          v
       drain (outflow)

Every stock has at least one inflow and usually an outflow. Here are some pairs:

Stock (the pile)	Inflow (fills it)	Outflow (drains it)
Bank balance	Deposits / income	Withdrawals / spending
Body weight	Calories eaten	Calories burned
Company headcount	People hired	People who quit or are fired
Inventory	Goods produced / received	Goods sold / shipped
Atmospheric CO₂	Emissions	Absorption by oceans/plants
Population	Births	Deaths

Here is a teaching point that sounds small but changes how you act in the world:

Key takeaway: You control flows; you experience stocks. You cannot directly grab the water level — you can only turn the faucet or open the drain. You cannot directly set your bank balance — you can only change income and spending. To move a stock, you must change a flow.

This is why "just decide to weigh less" doesn't work, and "just decide to have more savings" doesn't work. Stocks are not directly adjustable. They only move when you change the rates that feed or drain them — and then you have to wait.

2.3 The bathtub trap: why people misread stocks and flows

Now we hit the single most common error in all of systems thinking. It is so reliable that MIT professor John Sterman ran experiments showing that even top graduate students get it wrong. It is called the bathtub dynamics mistake.

The error is believing that reducing the inflow automatically reduces the stock. It does not. A stock keeps rising as long as inflow is bigger than outflow — even if the inflow is shrinking.

Analogy: Imagine the bathtub is filling fast and is nearly full. You turn the faucet down to a trickle. Has the water level dropped? No. It is still rising — just more slowly — because even a trickle is more water coming in than the slow drain can let out. The tub will overflow. To make the level fall, the inflow must drop below the outflow, not merely get smaller.

Example: This is exactly the climate confusion. People hear "we reduced emissions this year" and assume the amount of CO₂ in the atmosphere is now going down. It is not. CO₂ is a stock. Emissions are the inflow; natural absorption is the outflow. Emissions are still far higher than absorption, so the stock keeps climbing. Reducing the inflow only slows the rise. The stock will keep growing until emissions fall below what the planet can absorb. Same logic as the tub.

Common mistake: Confusing a flow with the stock it feeds. "We cut the inflow, so the pile must be shrinking." Wrong. The pile keeps growing until inflow drops below outflow. This single confusion explains bad reasoning about debt, weight, climate, inventory backlogs, and project bug counts.

One more idea while we are here. When inflow exactly equals outflow, the stock holds steady even though water is still moving. This is called dynamic equilibrium — a balanced, but living, state.

Dynamic equilibrium: When inflow and outflow are equal, so the stock stays level even though both flows keep running. (A bathtub with the faucet and drain perfectly matched: water moves, level holds.)

2.4 Feedback loops: the engine of behavior

Stocks and flows alone are just plumbing. The magic — and the trouble — starts when a stock loops back to change its own flows. That is a feedback loop.

Feedback loop: A closed chain of cause and effect in which a change in a stock travels around and comes back to affect the very flows that change that stock. Output becomes input. The loop, not any single part, governs the behavior.

The everyday example is a thermostat. The room temperature (a stock) controls the heater. The heater controls the room temperature. Round and round. Notice the circle: temperature → heater → temperature. That circle is the feedback loop.

Once you start looking for loops, you stop seeing straight lines of cause and effect and start seeing circles. This shift — from lines to circles — is the heart of systems thinking.

There are exactly two kinds of feedback loop. Learn both and you have the engine.

Balancing (negative) feedback loops — the stabilizers

Balancing loop: A goal-seeking loop that resists change. It pushes a stock toward some target and holds it there. It damps things down and seeks stability. Diagrammers label it B.

The word "negative" here does not mean "bad." It means the loop opposes (negates) whatever change is happening. It is a stabilizer.

Analogy: The thermostat again. If the room gets too cold, the heater turns on until it hits the target temperature, then shuts off. If it gets too hot, it stops heating. The loop is always pushing back toward the goal of, say, 21°C. Filling a glass of water is the same — you slow the pour as it nears full. Your body sweating to cool down is a balancing loop holding your temperature steady.

Balancing loops are the reason most things in your life don't fly off to infinity. They are nature's and society's thermostats: they hold blood sugar steady, hold prices around a sensible level, hold a thermostat at room temperature, and bring a crash diet's weight loss to a halt when the body fights to restore its old set-point.

Reinforcing (positive) feedback loops — the amplifiers

Reinforcing loop: A self-amplifying loop where more leads to more (or less leads to less). It does not seek a goal — it runs away, producing explosive growth or accelerating collapse. Diagrammers label it R.

Again, "positive" does not mean "good." It means the loop reinforces and amplifies the change in the same direction.

Analogy: Compound interest. Money in a savings account earns interest, which is added to the money, which then earns even more interest. More money makes more money makes more money. The pile snowballs. The same shape drives a rumor spreading (each person who hears it tells more people), a panic bank run (fear makes people withdraw, which spreads more fear), and the saying "the rich get richer."

Example: Habits run on reinforcing loops. You go for a run, feel good, so you run again, get a bit fitter, which makes running more pleasant, so you run more. More leads to more — a "virtuous cycle." Debt runs the same loop in reverse: you borrow, the interest grows your debt, which makes the payments harder, so you borrow more. More leads to more — a "vicious cycle." Same machinery, opposite feeling.

Here is a crucial truth about reinforcing loops:

Key takeaway: A reinforcing loop never runs forever. Something always stops it. Eventually it runs into a balancing loop — a limit. Money compounding hits the limit of the economy; a population growing hits the limit of food; a rumor runs out of people who haven't heard it. Whenever you see runaway growth, ask: "Where is the limit that will eventually catch it?"

	Balancing (B)	Reinforcing (R)
What it does	Resists change, seeks a goal	Amplifies change, no goal
Behavior over time	Settles toward a target; stabilizes	Explodes (growth) or collapses
Everyday picture	Thermostat, sweating, filling a glass	Compound interest, rumor, bank run
Nickname	Stabilizer / goal-seeker	Snowball / vicious-or-virtuous cycle

2.5 Delays: the source of nearly every surprise

If feedback loops are the engine, delays are why the engine constantly catches us off guard.

Delay (lag): The time gap between an action and its visible effect — between cause and consequence. Delays are normal and unavoidable, but they make systems hard to steer.

The perfect example sits in your bathroom.

Analogy: the shower with slow plumbing. You step in; the water is cold. You turn the dial toward hot. Nothing happens — the hot water hasn't traveled through the pipe yet. Impatient, you crank it further. Suddenly scalding water hits you. You yank it back toward cold, overcorrect, and freeze. Then you do it again. You end up oscillating between too hot and too cold, never settling. Nothing was wrong with the plumbing or with you — the problem is the delay between turning the dial and feeling the result.

That oscillating dance is not bad luck. It is the predictable output of a balancing loop (you trying to reach a comfortable temperature) plus a delay. Delays turn smooth steering into wild swinging.

Example: A company sees sales rising, so it hires more staff. But hiring, onboarding, and training take months — a delay. By the time the new people are productive, the sales spike has passed, and now the company is overstaffed and cutting jobs. Then sales rise again and they under-hire. The workforce oscillates, just like the shower, because of the lag between deciding to hire and the hire actually helping.

Example: the bullwhip effect. In a supply chain, a small bump in customer demand at the store gets passed up to the distributor, then the factory, then the raw-material supplier. Because each link reacts to delayed, secondhand information, the orders get wildly exaggerated as they travel up the chain — a tiny wiggle at the bottom becomes a giant whip-crack at the top. Peter Senge built a famous teaching game (the "Beer Game") entirely around this delay-driven chaos.

Best practice: When you see a system oscillating — swinging back and forth, overshooting, or when people complain "our fix isn't working yet" — look for a delay first. The cure is usually patience and smaller, slower adjustments, not bigger, faster ones. In the shower, the fix is to nudge the dial a little and wait.

2.6 What delays do inside loops: overshoot and collapse

Put delays together with the two kinds of loop and you get the dramatic behaviors that systems are famous for.

Overshoot: Going past a limit or target before the delayed signal that you've gone too far finally arrives. (You scald yourself in the shower because the "too hot!" signal came late.)
Oscillation: The back-and-forth swinging that a delay creates inside a balancing loop. (Hot, cold, hot, cold.)
Overshoot-and-collapse: When reinforcing growth charges through a limit so fast — and the limit itself gets damaged in the process — that the system doesn't just stop, it crashes and can't recover.

Example: A deer population in a valley grows year after year (a reinforcing loop). With a delay, it blows past the amount of food the land can supply. By the time starvation kicks in, there are far too many deer, and they have stripped and damaged the vegetation. The population doesn't gently level off — it crashes, and because the land was overgrazed, it can't bounce back to where it was. This overshoot-and-collapse pattern is the central warning of Meadows' famous study Limits to Growth.

 stock
  ^
  |            .-''-.  <- overshoot (past the limit)
  |          .'      '.
  |        .'          \    <- collapse
  |......./.............\.......  limit line
  |     .'               \____
  |  __.'
  +--------------------------------> time

Key takeaway: Smooth, gentle behavior comes from loops without much delay. Wild behavior — swinging, overshooting, crashing — comes from delays inside loops. If a system keeps surprising you, the delay is usually hiding in plain sight.

2.7 Nonlinearity and tipping points: when small causes have huge effects

So far we have mostly imagined that bigger pushes give bigger results. Real systems often don't work that way. They are nonlinear.

Nonlinear: Cause and effect are not proportional. A small change can produce a giant effect, or a huge effort can produce almost nothing. The relationship bends instead of running in a straight line.

Analogy: The straw that broke the camel's back. The camel carried ten thousand straws with no problem. One more — identical to the rest — and it collapses. The last straw was not special; the system had reached a point where its response stopped being proportional.

Example: Adding fertilizer to a crop. A little helps a lot. More helps a bit. Even more does nothing. Too much actually poisons the soil and kills the crop. Doubling the fertilizer does not double the harvest — the relationship curves and then turns around. Anyone who assumes "more input = more output, always" will badly misjudge what to do.

Closely related is the idea of a tipping point (also called a threshold).

Tipping point / threshold: A critical level beyond which the whole system suddenly flips into a new and different state — often a state that is hard or impossible to reverse.
Path dependence: The idea that where a system can go next is limited by where it has already been. History matters; some doors, once shut, stay shut.

Analogy: Heating a pot of water. 97°C — still water. 98°C — still water. 99°C — still water. 100°C — suddenly, violently, it boils and turns to steam. Nothing dramatic happened for almost the whole journey, and then everything changed at once. That sudden flip is a phase change at a threshold.

Example: A lake slowly takes in fertilizer runoff for years with no visible problem — the water stays clear. Then one season it tips: algae explode, oxygen vanishes, fish die, and the lake turns green and lifeless seemingly overnight. The slow build-up was invisible; the flip was sudden; and getting the lake back to clear is far harder than it was to push it over the edge.

Common mistake: Assuming proportionality — "twice the effort gives twice the result." Nonlinearity and tipping points break this everywhere. You can push a system for ages with little to show, then a tiny extra push flips it into a whole new state. This is why both slow progress and sudden disasters surprise people who think in straight lines.

2.8 Putting the pieces together: reading a system

You now have the full machinery. Let's watch the parts combine in one familiar story so the connections lock in: technical debt in a software team — but the same shape fits a messy house, a neglected garden, or unpaid bills.

The stock: messy, fragile code (technical debt) piled up in the project.
Inflow: every rushed, hacky shortcut adds to the pile.
Outflow: time spent cleaning and fixing the code drains the pile.
A reinforcing loop (R): the messier the code, the slower and more error-prone new work becomes, so the team rushes more, which adds more mess. More leads to more — the pile snowballs.
A delay: the pain of today's shortcut doesn't show up for weeks or months, so it feels free at the time.
A tipping point: eventually the codebase becomes so tangled that a small new feature triggers a cascade of bugs — the system flips from "manageable" to "everything is on fire."

Notice that nobody on the team is lazy or stupid. Each shortcut was a reasonable choice under pressure. The trouble lives in the structure — a reinforcing loop with a delay and a hidden threshold — exactly the lesson from Chapter 1: structure drives behavior.

  rushed shortcuts ---(+)---> messy code (STOCK)
        ^                          |
        |                         (+)
        |                          v
        +----(+)---- slower, buggier work
              (the more mess, the more rushing)
                = REINFORCING LOOP (R)
       [delay: pain shows up weeks later //]

Best practice: To read any system, ask four questions in order. (1) What is the stock — what's piling up or draining? (2) What are the flows feeding and draining it? (3) Is there a loop — does the stock circle back to change its own flows, and is that loop a stabilizer (B) or an amplifier (R)? (4) Where are the delays? Four questions, and a confusing situation becomes a readable machine.

2.9 A first taste of diagramming: the + and − labels

You will learn full causal loop diagrams in a later chapter, but the basic notation is simple enough to introduce now, because it makes the loops above precise. A causal loop diagram is just variables connected by arrows, and each arrow gets a label.

A + arrow means the two variables move in the same direction: more of A causes more of B (or less of A causes less of B).
A − arrow means they move in opposite directions: more of A causes less of B.

Common mistake: Reading + as "good" and − as "bad." The signs have nothing to do with desirability. They describe only the direction the two variables move together. "More crime → more police" is a + link even though crime is bad. Keep direction and desirability completely separate.

There is one elegant trick worth remembering: to tell whether a whole loop is reinforcing or balancing, count the minus signs around it. An even number (including zero) of minus links makes the loop reinforcing (R); an odd number makes it balancing (B). Delays are usually marked with a small double-slash // on the arrow, because — as you now know — they are the key to predicting oscillation.

2.10 Chapter recap

You have learned the four building blocks that nearly every system is made of, and the behaviors they produce when combined.

Stock: What piles up; the memory of past flows. You measure it at a moment.
Flow: The rate that fills (inflow) or drains (outflow) a stock. You control flows; you experience stocks.
Bathtub trap: Cutting the inflow does not shrink the stock; the stock keeps rising until inflow drops below outflow.
Balancing loop (B): Goal-seeking, stabilizing — the thermostat. Resists change.
Reinforcing loop (R): Self-amplifying — compound interest. Runs away until it hits a limit.
Delay: The lag between cause and effect; the source of oscillation, overshoot, and most surprises.
Nonlinearity: Cause and effect are not proportional; small pushes can have huge effects.
Tipping point: A threshold beyond which the system flips into a new, often irreversible, state.

Key takeaway: Stocks pile up, flows change them, feedback loops make stocks steer their own flows, and delays make the steering go wild. Master these four, and the strange behaviors of real systems — runaway growth, stubborn stability, oscillation, and sudden collapse — stop being mysteries and start being predictable machinery.

In the next chapters we will use this machinery to recognize the handful of recurring traps these structures produce (the "system archetypes"), and to find the high-leverage spots where a small, well-placed push can change everything.

Continue reading

🧭 Ten Disciplines

Seeing the Whole: What a System Is and Why It Behaves the Way It Does

🧭 Ten Disciplines

Leverage Points, System Traps, and Thinking in Systems (Advanced & Real-World)

🧭 Ten Disciplines

Scarcity, Choice, and How Prices Carry Information (Micro Foundations)