What a system is

A system is a set of factors or actors that interact with one another. The factors can be behaviours, people, teams, organisations, policies or resource levels. The important word is interact. A list of problems is not a system.

The guidance draws a line between three kinds of system.

• Simple. Few parts, clear cause and effect, and a predictable result. The guidance uses a bicycle: it is predictable, and changing one part has a known effect.
• Complicated. Many more parts, needing specialist expertise to understand and coordinate, but still predictable when it is running correctly. The guidance uses a space rocket.
• Complex. Many interacting parts, where the behaviour cannot be read off from the parts. Health and care pathways sit here.

A referral route between a general practice, a community team and a hospital is rarely a rocket. It is complex, and the difficulty is not the number of steps. The steps respond to each other, to the people inside them, and to everything happening around them.

Five properties that make a service complex

The guidance names five attributes of complex systems. All five show up in an ordinary service pathway.

• Change over time. The system is dynamic. It exchanges people, information and money with its environment, so its behaviour shifts without anyone deciding to change it.
• Emergence. The whole behaves in ways you cannot predict from the parts. The guidance's example is the unequal distribution of health across a population, which emerges from the interaction of employment, housing, food and other systems.
• Feedback. A change can reinforce or suppress further change. If fewer people commute by car, cycling may become more appealing, which cuts car journeys further.
• Non-linearity. The size of the effect is not proportional to the size of the input. A smoking reduction programme may move slowly, then fall quickly, then slow again.
• Adaptation. People and organisations learn and change their behaviour. After the Soft Drinks Industry Levy, some costs were passed to consumers, but not always on the targeted drinks.

Two consequences follow. An intervention is an event inside a moving system, not a lever on a still one, so the system will respond to it. And a local problem is usually both cause and effect of the wider pattern.

The parts of a map: elements, links and polarity

Most causal system maps share three components, and the group should use the same words for all three.

• Elements, also called nodes, factors or entities. These are the constituent parts, drawn as boxes or bubbles.
• Connections, also called links, edges or arcs. These show the direction, and sometimes the strength, of a relation between two elements.
• Networks. The whole set of elements and connections, which shows the structure of the system or part of it.

In a causal loop diagram the elements are written as factors that can rise or fall. That is a working rule, not a style choice. 'Waiting times' is a factor, because it can go up or down. 'Poor communication' is a verdict on top of a factor and gives the group nothing to link to. Write 'consistency of information passed between teams' instead, which can be higher or lower.

Each link carries a polarity. A positive link means an increase in one element raises the other, and a decrease lowers it. A negative link means an increase in one lowers the other. Maps may also carry delay marks, where the influence of one factor on another takes time to appear. In practice a delay mark records that a delay exists, not how long it is.

Feedback, and why pathways surprise you

A feedback loop appears when a chain of causes returns to influence the factor it started from. Reinforcing loops reinvest a change and amplify it. Balancing loops resist a change and hold the system where it was. The guidance gives an example from maternal care: better quality of care raises attendance, higher attendance raises workload, higher workload lowers staff motivation, and lower motivation lowers quality of care again. That loop is balancing, so the first improvement is partly cancelled. A second loop runs from quality of care to maternal mortality to staff motivation and back, and it reinforces.

This is the most useful thing a group can learn to see. Services rarely fail because one step is missing. They hold their current performance because a balancing loop keeps pulling it back, which is why a single process fix tends to drift back to its old level. You can spot the type without software: count the negative links around the loop. An odd number makes it balancing, an even number (including none) makes it reinforcing.

Boundaries, and why the map is not the service

A boundary sits at the junction between a system and its environment. It is not a fixed wall: boundaries are continually created, maintained and degraded. Boundary setting is therefore a decision, not a discovery. A boundary statement can name a time window as well as a scope: the route from first contact to discharge, over these months. Someone has to decide whether the map runs from first contact to discharge, or from housing to employment, and that is normally you plus whoever is paying.

Three limits follow, and they are why a pathway map should never be presented as a description of the pathway.

• The map is partial. It is bounded by the knowledge and conditions of the people who made it, at one point in time.
• The map is not reproducible. A different group would have drawn a different map, and this is not a design fault.
• The map flattens difference. Relations often work differently for different groups, so inequalities are hard to show. Mapping relies on diverse participation to add that lived-experience context.

Mental models and leverage points

People carry assumptions about how a system works: what is inside it, what is outside, what drives what. Systems thinking calls these mental models and asks for them to be made explicit, so that differences can be seen rather than argued about in the abstract. It also lets you name a disagreement precisely, because it usually turns out to be about one link rather than the whole picture.

A leverage point is a place where a small shift produces a large change across the system. The guidance is clear that there is no gold standard way of finding them; inspecting feedback loops, network analysis and stakeholder knowledge are all used. Two terms are worth knowing, because they are easy to spot in a finished map: sinks, which influence nothing else, and drivers, which influence a problem without being influenced by it.

The vocabulary to settle before you map

Four decisions, made before the workshop, will save the analysis later.

• Name the system and its boundary. One sentence, written down and agreed with the group in the first ten minutes.
• Write every element as a factor that can rise or fall. Reject labels and verdicts at the point they are written.
• Agree what counts as a link. A link means one factor directly changes another. If it needs a story in the middle, it is two links.
• Decide now that the map is provisional. That keeps validation in the plan, rather than in a methods section written afterwards.