Stock-and-Flow Models

Thinking Tools3.3 Mapping the System Around the ProblemAlso known as Stock and Flow Diagrams, System Dynamics Simulation

Originated by Jay W. Forrester (Industrial Dynamics, 1961), best known through Donella H. Meadows, Dennis L. Meadows, Jørgen Randers, and William W. Behrens III (The Limits to Growth, 1972).

Represents a system as quantities that accumulate and the rates that fill or drain them, then simulates the model forward to show behavior over time.

An engineering team is a stock: it fills only as fast as starts beat departures

Engineers are a stock filled by starts, which arrive two months after each offer is signed, and drained by departures, a share of everyone on the team. Set the hires and the quit rate, then Play: the team grows only while starts exceed departures. Set hires to zero for a freeze. Illustrative.

Select a loop to trace it, or a box to read it.B1, The drain. More engineers means more departures at the same rate, which pulls the stock back toward the level where starts and departures match. One negative (−) link: an odd count, so the loop balances.Offers signed (flow). The inflow decision: offers accepted each month, set on the slider.Waiting to start (stock). A small stock of its own: people serving two months' notice at their old jobs. Half of it starts each month.Engineers (stock). The stock: 80 today. It rises only while starts are larger than departures, however large either one is.Departures (flow). The outflow: the quit rate times the number of engineers, so a bigger team loses more people each month at the same rate.Quit rate. The share of engineers who leave in a month, set on the slider.
3
2.0%

Engineers

0: 801: 802: 813: 824: 845: 856: 867: 878: 899: 9010: 9111: 9212: 9313: 9514: 9615: 9716: 9817: 9918: 10019: 10120: 10221: 10322: 10423: 10524: 10605010015001234567891011121314151617181920212223240: 80: 802: 81: 824: 84: 856: 86: 878: 89: 9010: 91: 9212: 93: 9514: 96: 9716: 98: 9918: 100: 10120: 102: 10322: 104: 10524: 106050100150024681012141618202224

Starting a month

0: 2.01: 2.52: 2.83: 2.94: 2.95: 3.06: 3.07: 3.08: 3.09: 3.010: 3.011: 3.012: 3.013: 3.014: 3.015: 3.016: 3.017: 3.018: 3.019: 3.020: 3.021: 3.022: 3.023: 3.024: 3.00.02.04.001234567891011121314151617181920212223240: 2.0: 2.52: 2.8: 2.94: 2.9: 3.06: 3.0: 3.08: 3.0: 3.010: 3.0: 3.012: 3.0: 3.014: 3.0: 3.016: 3.0: 3.018: 3.0: 3.020: 3.0: 3.022: 3.0: 3.024: 3.00.02.04.0024681012141618202224

Leaving a month

0: 1.61: 1.62: 1.63: 1.64: 1.75: 1.76: 1.77: 1.78: 1.89: 1.810: 1.811: 1.812: 1.913: 1.914: 1.915: 1.916: 2.017: 2.018: 2.019: 2.020: 2.021: 2.122: 2.123: 2.124: 2.10.01.02.03.04.001234567891011121314151617181920212223240: 1.6: 1.62: 1.6: 1.64: 1.7: 1.76: 1.7: 1.78: 1.8: 1.810: 1.8: 1.812: 1.9: 1.914: 1.9: 1.916: 2.0: 2.018: 2.0: 2.020: 2.0: 2.122: 2.1: 2.124: 2.10.01.02.03.04.0024681012141618202224

Month 24 of 24

Engineers
106
Starting a month
3.0
Leaving a month
2.1

B1, The drain
More engineers means more departures at the same rate, which pulls the stock back toward the level where starts and departures match. One negative (−) link: an odd count, so the loop balances.
Offers signed (flow)
The inflow decision: offers accepted each month, set on the slider.
Waiting to start (stock)
A small stock of its own: people serving two months' notice at their old jobs. Half of it starts each month.
Engineers (stock)
The stock: 80 today. It rises only while starts are larger than departures, however large either one is.
Departures (flow)
The outflow: the quit rate times the number of engineers, so a bigger team loses more people each month at the same rate.
Quit rate
The share of engineers who leave in a month, set on the slider.

An example

The CTO of a logistics software firm had 80 engineers and a board target of 100 within a year. Recruiting promised three signed offers a month, and the plan simply added 36 people to 80. The head of people operations drew it as a stock with an inflow and an outflow instead. New hires serve two months of notice before they start, and about 2 percent of engineers leave each month, 1.6 people at today's size and more as the team grows.

Run forward, the model put the team at 92 after twelve months and at 100 only in month nineteen. Four signed offers a month would reach 100 in about eleven. The board then asked how fast the team would shrink if a budget freeze stopped all hiring today. The same model answered in a minute: starts would continue for two months as signed offers arrived, then the team would lose a little under one and a half engineers a month, reaching about 66 a year after the freeze began.

Why it works

People judge accumulations badly. Asked how a team, an inventory, or a cash balance will move, most look at the inflow alone, or assume the stock follows the inflow's shape. A stock changes only by the difference between what flows in and what flows out, and it keeps each change until a flow alters it. So a stock moves slowly, lags its flows, and can keep rising while its inflow falls, as long as the inflow stays above the outflow.

Writing the model down forces every flow to be named and given a rate, including the ones a plan leaves out, such as departures that grow with the team. Running it forward turns a target into a date and a policy into a path, and it shows which flow moves the stock fastest. Here, keeping people mattered as much as hiring them.

The thinking behind it

Jay W. Forrester Jay W. Forrester, an MIT engineer and management theorist, founded system dynamics with Industrial Dynamics (MIT Press, 1961), which models an organization as stocks, flows, and feedback loops with delays and runs the model forward to see how it behaves. The book drew on a study of Sprague Electric and gave the first account of what is now called the bullwhip effect. Stocks that accumulate and the rates that fill and drain them are the method's basic vocabulary.

Donella H. Meadows Donella H. Meadows was lead author of The Limits to Growth (Universe Books, 1972), written with Dennis L. Meadows, Jørgen Randers, and William W. Behrens III, the system dynamics study that made stock-and-flow modeling widely known by running the growth of population and industry against finite resources. In Thinking in Systems (2008), published after her death, she explained stocks, flows, feedback loops, and delays for general readers, and in Leverage Points (1999) she ranked the places to intervene in a system.

How to run it

  1. Set the boundary. Decide what the model includes and what it takes as given, such as the job market.
  2. Name the stocks. List the quantities that accumulate: engineers, people serving notice, open requisitions.
  3. Name every flow in and out. Give each stock its inflows and outflows, including easily forgotten ones such as departures and retirements.
  4. Write each flow as a rule. A fixed number a month, a share of a stock, or a delay such as two months of notice.
  5. Start from today's values. Enter the current levels and rates, and mark which ones are guesses.
  6. Check it against history. Replay the last year or two; if the model cannot reproduce them, fix it before trusting it.
  7. Run the policies. Try each option, such as more hires, fewer departures, or a freeze, and compare the dates and levels.

Where it goes wrong

The first failure is a model without its outflow. Plans that add hires to headcount, orders to inventory, or sign-ups to users while ignoring departures miss every date. Spot it when the plan's line is perfectly straight, and recover by adding the drain as a share of the stock.

The second is false precision. A model with fifty parameters looks rigorous and rests on fifty guesses; keep it small and say which numbers are estimates. The third is skipping the check against history. A model that cannot replay last year's numbers will not forecast next year's, however clean its diagram looks.

When to use something else

  • Causal Loop Diagram. When the team is still agreeing on what drives what, before any rates are worth estimating.
  • Queueing Theory. When the stock is a queue of work and the question is how the wait depends on arrivals and capacity.
  • Monte Carlo Simulation. When the flows themselves are uncertain and the decision needs a range of outcomes with odds attached.

Where it sits in SOLVE eX

Go to the source

  • Industrial Dynamics, 1961
  • The Limits to Growth, 1972

Back to the catalog