A behavior-over-time graph (BOTG) is a simple plot of an important variable against time. It is often the best first tool in a systems inquiry because it forces a team to specify which behavior needs explanation.
How to create one
- Name a variable that can rise or fall.
- Select a time horizon long enough to reveal the pattern.
- Mark known data when available.
- Sketch uncertain sections and label them as estimates.
- Add a desired trajectory or competing stakeholder views.
- Note events that may have changed structure.
What the shape can reveal
Exponential growth suggests reinforcing feedback; a plateau suggests a balancing constraint; repeated overshoot may indicate delayed balancing action; S-shaped growth suggests early reinforcement followed by a limit. These are hypotheses, not diagnoses.
Use multiple variables
Plot workload, quality, backlog, and staff capability separately rather than compressing them into one score. Compare their timing. If output rises before defects and turnover, the delay may explain why leaders misread the policy.
Common mistakes
- Using vague variables such as “the problem.”
- Choosing a time window that begins after the pattern started.
- Presenting a memory-based sketch as measured data.
- Inferring causality from similar curves.
- Ignoring different stakeholder experiences.
The graph prepares a stronger question for a causal loop diagram: what structure could generate this shape?
References
- Sterman, J. D. (2000). Business Dynamics. Irwin/McGraw-Hill.
- Richmond, B. (1994). Systems Thinking/System Dynamics: Let’s Just Get On With It. isee systems.
- JMU System Dynamics Learning Guide.
Facilitating a group exercise
Give participants the same variable and ask them to sketch its past and expected future independently. Differences reveal assumptions about the starting date, measurement, population, and desired outcome. Discuss those differences before averaging them away.
Add a “business as usual” line and one or more desired trajectories. The gap clarifies the policy challenge. Mark confidence ranges when data are sparse and distinguish an observed line from a remembered or anticipated one.
From graph to evidence
Locate available time-series data, but check whether definitions changed. A sudden improvement may reflect a revised reporting rule rather than system behavior. Disaggregate by location, group, or case type because averages can hide divergent patterns.
From evidence to structure
Use the graph to evaluate loop hypotheses. If a proposed reinforcing loop cannot explain the observed plateau, look for a balancing constraint. If oscillation is present, examine delays, decision rules, and pipeline effects. The graph keeps causal mapping accountable to behavior.
Discussion