Systems Thinking: Stocks, Flows, Feedback and Delays, and Why We Misjudge Them

Systems thinking is the habit of explaining behaviour by structure: what accumulates (stocks), what fills and drains it (flows), which effects loop back on their causes (feedback), and how long all of that takes (delays). It sounds obvious, and people are bad at it. When MIT Sloan students drew how water in a bathtub changes over time, given only the inflow and outflow, they averaged 48% on the task. This post explains the four core ideas, the research on how educated adults misjudge them, and what helps you practise.

Systems thinking in one paragraph

Systems thinking replaces "A causes B" with "this structure produces this pattern over time". A backlog, a cash balance, a team's trust and the CO2 in the air are all stocks. They change only through their flows, they respond to feedback, and they respond late.

Donella Meadows, one of the field's clearest writers, put the two loop types in a sentence in her essay on places to intervene in a system: "A negative feedback loop is self-correcting; a positive feedback loop is self-reinforcing." A thermostat is the first kind. Compound interest and word of mouth are the second.

The practical promise is better decisions in situations where the obvious fix makes things worse, or helps now and hurts later. The research below says the promise is real but the starting skill level is low.

The four building blocks of systems thinking

Each block is simple on its own. The trouble starts when they combine.

Stocks are quantities that accumulate: inventory, savings, the number of people in a room, unread email. A stock is a memory of everything that has flowed in and out. That is why it changes slowly even when its flows change fast.

Flows are rates: hires per month, orders per week, litres per minute. The rule linking the two is plain bookkeeping. The stock rises whenever inflow exceeds outflow, whatever the inflow is doing. A falling inflow can still mean a rising stock.

Feedback loops connect a stock back to its own flows. A low inventory triggers more production (balancing). More users attract more users (reinforcing). Most real behaviour comes from several loops pulling against each other.

Delays sit between action and effect. Meadows calls delays in feedback loops "critical determinants of system behavior". A delay is what turns a sensible correction into an overshoot: you keep pushing because the result has not shown up yet, and by the time it shows up you have pushed too far.

The bathtub studies: how educated adults misjudge stocks and flows

The most cited evidence comes from John Sterman's group at MIT. In Booth Sweeney and Sterman's inventory of systems thinking tasks, students in an introductory system dynamics course at MIT Sloan got about 10 minutes per task, unpaid and ungraded. More than half had studied engineering, computer science, maths or science.

The bathtub task is close to the simplest dynamic problem there is. A graph shows water flowing in and draining out; you sketch the amount of water in the tub. There is no feedback and the numbers are round. A cash-flow version swaps the tub for a bank account.

With square-wave flows, average performance was 77%. With a sawtooth inflow it fell to 48% across 150 students. Only 40% put the peaks and troughs of the stock in the right places. Only 28% drew the stock's slope in line with the net flow, which is the core rule of accumulation. Many drew a stock that simply copied the pattern of the net flow.

A manufacturing task added one delay and one balancing loop. Customer orders step up; you sketch production. Average performance was 41%. Only 44% showed production overshooting orders, which it must do to rebuild the inventory lost in the meantime. And 89% drew production paths that broke conservation of material: the units they had production make up later did not match the units lost earlier.

Why stock-flow failure is so stubborn

The obvious explanations do not hold up. Cronin, Gonzalez and Sterman (2009) ran a series of experiments with highly educated adults and found that poor performance was not explained by trouble reading graphs, lack of context, low motivation or limited cognitive capacity. They called the pattern stock-flow failure and found it hard to shift.

Their account of the cause is a shortcut they call the correlation heuristic: people expect the stock to look like the flow. If more people enter a shop at minute 8, the shop must be fullest at minute 8. It is not; it is fullest when entries stop exceeding exits.

A small think-aloud study shows the gap clearly. Aşık and Dogança Küçük (2021) gave 40 undergraduates the classic department store task. On the flow questions (when did most people enter and leave), 88% and 78% answered correctly. On the stock questions (when were the most and fewest people inside), only 28% and 25% did. Their protocols showed weak planning, monitoring and checking, which links stock-flow failure to metacognition: people did not stop to test the answer that felt right.

Systems thinking failures outside the lab

The same mistakes show up where the stakes are high. In the beer distribution game, players run a simulated supply chain with orders, shipments and delays. Sterman's 1989 study in Management Science found that the players' combined decisions produced swings that systematically departed from optimal behaviour. His analysis traced the poor results to "misperceptions of feedback", above all players' insensitivity to how their own orders fed back into the system.

Climate policy is the bathtub at planetary scale. Sterman and Booth Sweeney (2007) asked MIT graduate students what it takes to stabilise atmospheric greenhouse gas concentrations. Most believed concentrations could stabilise while emissions kept exceeding removal. As MIT News summarised, emissions run at about twice the rate of removal, so stabilising concentrations means cutting emissions by at least half. A tub filling faster than it drains keeps rising.

At work the shapes are familiar. A hiring freeze that does not shrink the backlog for months. A price cut whose demand spike hides the slower loss of margin. A team that adds reviewers to speed up releases and slows them down. Each is a stock, a flow, a loop and a delay that a single-cause explanation misses.

How to practise systems thinking: what the evidence supports

The honest summary: quick fixes help a little, and real teaching helps more. A review in Baghaei Lakeh and Ghaffarzadegan's paper notes that changing the graph, the context or the motivation has failed to improve stock-flow reasoning, while education in stocks and flows is one of the few interventions that reliably works.

Their own experiment tested cheap nudges with 184 online participants. Answering four unrelated analytical questions before the department store task raised correct answers on one stock question from 6.12% to 22.09%. A simple warning that the task is harder than it looks raised it to 16.51%. The authors are candid about the limits: more than 78% of participants still got at least one stock question wrong.

Two practical lessons follow. First, slowing down pays. The warning seems to work by pushing people out of the fast, intuitive answer. Anything that crowds out careful reasoning, such as heavy cognitive load, pulls the other way. Second, a second look helps. In the think-aloud study, being told their first answer was incorrect let another 20% of students solve the department store task on their next attempt.

Meadows offers a third lesson for where to aim. In her ranking of leverage points, "constants, parameters, numbers" sit at the bottom. Delays, feedback loops, information flows and goals sit higher. Most meetings argue about the numbers.

How you reason about these structures out loud also says a lot about how you think. If you want to see your own reasoning in numbers, NOOVA runs a 20-minute voice dialogue instead of hours of tests.

A checklist for practising systems thinking this week

1. Name one stock you are responsible for and write down its inflows and outflows.
2. Sketch, before checking, how that stock will move over the next month if the flows stay as they are.
3. Ask "is inflow above or below outflow right now?" instead of "is inflow rising?".
4. Draw one feedback loop around a recurring problem and mark it balancing or reinforcing.
5. Estimate the delay between your last decision and its visible effect, then wait that long before correcting.
6. Check every plan for conservation: what you take out somewhere must be put back somewhere.
7. Move one discussion from adjusting a number to changing a rule, a goal or an information flow.

Sources

• Bathtub Dynamics: Preliminary Results of a Systems Thinking Inventory, Booth Sweeney and Sterman, International System Dynamics Conference, 2000
• Why Don't Well-Educated Adults Understand Accumulation?, Cronin, Gonzalez and Sterman, Organizational Behavior and Human Decision Processes, 2009
• Modeling Managerial Behavior: Misperceptions of Feedback in a Dynamic Decision Making Experiment, Sterman, Management Science, 1989
• Understanding Public Complacency About Climate Change, Sterman and Booth Sweeney, Climatic Change, 2007
• Understanding climate change complacency, MIT News, 2008
• Metacognition in action as a possible explanation for stock-flow failure, Aşık and Dogança Küçük, System Dynamics Review, 2021
• The Dual-Process Theory and Understanding of Stocks and Flows, Baghaei Lakeh and Ghaffarzadegan, System Dynamics Review, UCL Discovery copy
• Leverage Points: Places to Intervene in a System, Donella Meadows, The Donella Meadows Project