Systems thinking

The Language of Systems Thinking: Feedback and Delays

In short Everyday language describes the world as straight lines from cause to effect, but organizations behave in circles. Peter Senge offers a small vocabulary (reinforcing feedback, balancing feedback and delays) that lets you describe why a situation behaves as it does, and find where to act.

Based on: P. Senge, The Fifth Discipline: The Art and Practice of the Learning Organization, ch. 5.

On this page

Key ideas#

Detail complexity vs. dynamic complexity#

Senge separates two kinds of complexity:

  • Detail complexity: many interacting parts, as in assembling a complex machine or managing stock in a large store. Hard, but not necessarily hard in a systems sense.
  • Dynamic complexity: the short-term and long-term effects of an action differ, or local effects are the opposite of system-wide ones.

Systems thinking deals with the second kind, and standard forecasting and planning tools, however sophisticated, tend to miss it. A common mistake is to fight complexity with more complexity: adding variables to a model instead of seeing the pattern of change.

Circles, not lines#

Senge’s signature example is filling a glass of water. It feels like a one-way action, but it is a loop: you compare the desired level with the actual level, adjust the tap, the water flow changes the level, and the gap shrinks. Each element is both cause and effect. The structure drives the behavior.

This also changes how you think about responsibility. In a loop everyone contributes to the problem, though not everyone has the same power to change it. Hunting for a single culprit, or blaming only yourself, is a product of linear thinking.

Three building blocks#

  • Reinforcing feedback amplifies change in one direction: growth, or an accelerating decline such as a run on a bank. Where you see growth, a reinforcing loop is at work.
  • Balancing feedback pulls the system toward a goal, explicit or hidden: a thermostat, or a habit you cannot shake.
  • Delays mean effects show up gradually. They hide the system’s nature and make it look as if a fix worked, or failed, when the real effect has not arrived yet.

Senge argues that any system can be described with these three elements. They are the nouns and verbs of the language; system archetypes are its common sentences.

Escalation: two “defensive” lines make a loop#

Senge’s example is an arms race. Each side saw its build-up as a linear, defensive response to the other’s threat. Put the two lines together and they form a reinforcing loop: one side’s build-up raises the threat the other perceives, which triggers its own build-up, and so on. Each side acted rationally locally, and together they produced the opposite of what both wanted. Senge names this pattern escalation and finds the same structure in gang rivalries, bitter divorces and advertising wars.

How to apply it#

Draw a causal loop diagram:

  1. Describe the problem as behavior over time (“waiting times rise every spring and fall every autumn”), not as a single event.
  2. Name the variables as quantities that can rise or fall: “backlog”, not “the backlog problem”.
  3. Link them with arrows. Mark each link as same direction (more A, more B) or opposite direction (more A, less B).
  4. Close the loops. A loop with an even number of opposite links is reinforcing; an odd number makes it balancing.
  5. Mark the links where effects arrive slowly.
  6. Check the diagram explains the behavior from step 1. If not, a loop or delay is missing.
  7. Look for where to act: weaken a harmful reinforcing loop, remove the limit in a balancing one, or shorten a delay.

Example#

The Workshop’s diagram for spring backlogs has three parts:

  • Reinforcing loop: satisfied customers lead to more recommendations, more bookings, more revenue, more mechanics hired, shorter waits, and more satisfied customers.
  • Balancing loop: more bookings mean a longer backlog, longer waits, and in time fewer bookings.
  • Delay: hiring and training a mechanic takes months.

The diagram explains a pattern the owner had blamed on “lazy managers”. In spring the backlog grows and the owner hires. By the time new mechanics are productive the season is over, so there is overcapacity in autumn. The lever is not stricter managers but a shorter delay: train seasonal mechanics in winter, or pre-book spring services.

Pitfalls#

  • Adding more variables to a model → look for the few loops that explain the pattern.
  • Describing events instead of patterns → plot the behavior over time first.
  • Ignoring delays → mark them explicitly; they explain most “surprises”.
  • Searching for someone to blame → ask which structure produces the behavior.

Sources and further reading#

  • P. Senge. The Fifth Discipline: The Art and Practice of the Learning Organization. Doubleday, 1990, ch. 5.