Behaviour Over Time Charts

Before you look at the data, sketch what you think you will see. A Behaviour Over Time chart is a simple freehand plot of how a key variable changes over time. It takes two minutes to draw and does something no statistical tool can: it makes your mental model explicit, commits you to a prediction, and prevents you from unconsciously fitting your interpretation to the data after the fact. It is the starting point of the improvement method — Step 1, before the numbers.

What a BOT chart is

A Behaviour Over Time (BOT) chart is a concept from system dynamics, developed in the tradition of Jay Forrester at MIT and popularised by Peter Senge in The Fifth Discipline (1990). It is simply a sketch — typically freehand, on paper or a whiteboard — of how one or more key variables change over time. The horizontal axis is time. The vertical axis is the variable of interest: revenue, patient wait times, error rate, support cases, market share. No numbers are required. The shape is the point.

The BOT chart is the first step in understanding a system problem. Before collecting data, before running analysis, before building a model, you draw what you think is happening. The shape you sketch encodes your assumptions about the system: is this variable growing, declining, cycling, stuck? Has it changed recently or is it long-established? Is the pattern getting worse or stabilising?

Why draw it before looking at the data

The most common analytical error in improvement work is not statistical — it is cognitive. When you look at a time series and then decide what it means, your interpretation is inevitably shaped by what you see. A flat line looks like stability. A recent uptick looks like improvement. A gradual decline looks manageable. The pattern you perceive depends heavily on where you start the series, what scale you use, and what you were hoping to find.

Drawing the BOT before you look at the data does three things:

• It surfaces your mental model. The shape you sketch is a direct representation of what you believe is happening. Comparing it to the actual Bootstrap CUSUM result tells you whether your mental model is accurate — and if it is not, precisely where it is wrong.
• It prevents reverse-engineering. Once you have committed to a prediction in writing, you cannot unconsciously adjust your interpretation to fit the result. The prediction is the Plan step of PDSA. The Bootstrap CUSUM result is the Study step. The gap between them is the learning.
• It aligns teams before analysis. When a team draws BOT charts independently and then compares them, the disagreements are immediately visible. Two people looking at the same organisation see different patterns — and those differences reveal assumptions that would otherwise never be surfaced. The BOT exercise is a team diagnostic tool as much as an analytical one.

The six archetypal patterns

System dynamics identifies six recurring patterns that appear across very different organisations and sectors. Each has a characteristic shape and a specific loop structure that produces it. Recognising the pattern is the first step toward understanding the loop — and breaking it.

Patterns and loops — the connection

Every pattern is produced by a loop structure. This is the central insight of system dynamics: the behaviour of a system over time is determined by its feedback structure, not by the intentions or efforts of the people operating within it. Understanding the pattern is the first step toward identifying the loop. Identifying the loop is the first step toward changing it.

The relationship is precise:

• Reinforcing loop (R) dominant → Exponential growth or exponential decay
• Balancing loop (B) dominant → Goal-seeking, approach to a stable level
• B loop with delay → Oscillation — the system overshoots and corrects repeatedly
• R then B → S-curve — fast growth until a limit activates
• Multiple competing loops → Complex patterns combining the above
• Locked system / unaddressed constraint → Stagnation

Once you can name the pattern, you can hypothesise the loop. Once you can draw the loop, you can identify what needs to change to alter the pattern. This is the path from BOT chart to Causal Loop Diagram — the BOT shows what is happening over time, the CLD explains why.

Where BOT fits in the 7-step improvement method

BOT and Bootstrap CUSUM — prediction and test

The BOT chart and Bootstrap CUSUM are the prediction-and-test pair at the heart of the improvement method. The BOT is what you expect to see. Bootstrap CUSUM is what the data actually shows. Together they form a complete hypothesis-testing cycle that is more rigorous than either tool alone.

The BOT prevents post-hoc rationalisation. Without a pre-committed sketch, any Bootstrap CUSUM result can be interpreted as confirming what you already believed. A change point appears: “yes, that is what we expected.” No change point appears: “yes, the system is stable as we expected.” The BOT sketch, committed before the analysis, removes this escape route. Either the pattern you predicted appeared or it did not. The answer is binary and honest.

Bootstrap CUSUM makes the BOT prediction precise. A BOT sketch says “I expect an upward shift.” Bootstrap CUSUM says “an upward change point appeared on 14 March 2024 at 97.3% confidence.” The precision transforms an intuition into a statistical statement — one that can be reported, compared to predictions, and used to make the case for scaling or stopping.

BOT charts from the site articles

Every article on this site has an implicit BOT chart. Here is how each maps to the archetypes:

• NHS A&E — Stagnation then decay. Flat near target for several years, then a structural shift downward. Four Bootstrap CUSUM stages. No upward change point. The decay pattern is driven by a reinforcing loop: performance pressure → short-term fixes → pressure relief → structural fix never made → performance deteriorates further.
• Dementia diagnosis — S-curve then decay. Growth to target (reinforcing loop: awareness → referrals → diagnoses → more awareness), then plateau (balancing: target reached → pressure reduces), then collapse (COVID special cause) and partial recovery. The BOT is not simple — it requires three loop phases to explain.
• UK electricity emissions — Delayed decay then rapid fall. Stagnation until 2013 (constraint: economics of coal generation unchanged), then a sharp structural decline (reinforcing loop: carbon price → coal uneconomic → gas/renewables → lower emissions → investment in renewables → further cost reduction). One Bootstrap CUSUM change point at 99.8% confidence.
• Hydrogen plant — Hidden oscillation. The efficiency gap looks like noise in the raw data. Bootstrap CUSUM on residuals reveals the underlying pattern. The BOT sketch would have shown: we expect the efficiency gap to be consistent and persistent, not random — the pattern of a structural constraint, not random variation.

Related concepts