Catalog
concept#Architecture#Governance#Product#Reliability

System Dynamics

Methodology for modelling complex dynamic systems using stocks, flows and feedback loops to analyse behavior over time.

System dynamics is a methodological approach for modeling complex, dynamic systems using stocks, flows and feedback loops.
Established
High

Classification

  • High
  • Organizational
  • Organizational
  • Advanced

Technical context

Vensim / Stella (modelling tools)PySD / Python ecosystem for analysis and automationBI and visualisation tools for presenting results

Principles & goals

Focus on stocks, flows and feedback loops to explain behaviour.Model over time with explicit consideration of delays.Iterative calibration, validation and inclusion of stakeholder knowledge.
Discovery
Enterprise, Domain

Use cases & scenarios

Compromises

  • Incorrect or unvalidated assumptions lead to misleading outcomes.
  • Overfitting to historical data reduces predictive power.
  • Organisational resistance prevents adoption of model results.
  • Start with simple models and gradually increase complexity.
  • Document assumptions and validation steps explicitly.
  • Closely involve domain experts and conduct regular reviews.

I/O & resources

  • Time series of relevant system variables (e.g. demand, inventory)
  • Expert knowledge on causalities and feedbacks
  • Organisational and process descriptions
  • Scenario simulations and time‑series forecasts
  • Policy recommendations and leverage points
  • Visualised model structures and reporting artefacts

Description

System dynamics is a methodological approach for modeling complex, dynamic systems using stocks, flows and feedback loops. It supports understanding the causes of behavior over time and analysing policy or design options. Typical applications include strategic planning, policy analysis and organisational simulation. Models are iteratively calibrated and validated.

✔

  • Enables deeper understanding of causal relations and long‑term effects.
  • Supports robust scenario and policy evaluation.
  • Visualises system behaviour and facilitates stakeholder communication.

✖

  • Result quality strongly depends on data and model assumptions.
  • Complex models can become opaque and overloaded.
  • Requires specialised expertise and time for calibration.

  • Forecast deviation

    Difference between modelled projection and real measurements over defined periods.

  • Sensitivity index

    Measures how strongly outcomes react to changes in key parameters.

  • Communication acceptance

    Degree of acceptance and understandability of model results among stakeholders.

Urban Dynamics (Forrester)

Classic example simulating urban growth and decay processes with feedbacks.

Sterman: Business Dynamics teaching examples

Teaching examples and models from John Sterman's courses on business dynamics.

Supply chain simulation model

Use case analysing inventories, lead times and feedbacks in logistics.

1

Define problem frame and identify stakeholders.

2

Design structural model with stocks, flows and feedbacks.

3

Estimate parameters, fit to historical data and calibrate.

4

Run scenarios, sensitivity analyses and validation.

⚠️ Technical debt & bottlenecks

  • Unstructured model library without version control.
  • Insufficient testing and validation practices for models.
  • Hardcoded parameters in models instead of configurable inputs.
Data availability and qualityUnclear or inconsistent model assumptionsLack of stakeholder commitment
  • Using unvalidated parameters as basis for policy decisions.
  • Replacing qualitative analysis entirely with a model.
  • Tuning models solely to justify preconceived decisions.
  • Confusing correlation with causal mechanism.
  • Underestimating the importance of time delays.
  • Neglecting nonlinear effects when scaling.
Systems thinking and causal modellingQuantitative data analysis and time series knowledgeCommunication and stakeholder facilitation
Feedbacks and their amplifying/dampening effectsTime delays in information and material flowsNonlinearities and saturation effects
  • • Limited historical data length for some variables
  • • Budget and timeline constraints for model development
  • • License restrictions of commercial modelling tools