Developed by Jac A. M. Vennix (Group Model Building, 1996) and by George P. Richardson and David F. Andersen (Teamwork in Group Model Building, 1995).
Builds a system dynamics model in facilitated workshops with the people who must act on it, producing a shared model and shared understanding together.
The model a hospital's workshop drew together, and the policy it tested
Surgeons, emergency staff, ward nurses, and social workers drew these loops in two sessions. Full wards stretch the staff and slow discharges, which fills the wards further (R1); full wards also cancel planned operations (B1). The run opens with no weekend team; add one and find the smallest that stops the cancellations. Illustrative.
Beds occupied
Operations canceled that week
Week 16 of 16
- Beds occupied
- 491
- Operations canceled that week
- 9
- R1, Full wards slow discharge
- More patients stretch the staff, discharges slow, and more patients stay. The nurses' loop, and the one nobody had drawn before. Two negative (−) links: an even count, so the loop reinforces.
- B1, Full wards cancel surgery
- When beds run out, planned operations are canceled, which caps admissions. It holds the wards just under full, at the surgeons' expense. One negative (−) link: an odd count, so the loop balances.
- Admissions
- About 280 a week, emergency and planned. The surgeons insisted planned operations be drawn as their own flow.
- Beds occupied
- The stock everyone argued about: 500 beds, and trouble once more than 485 are full.
- Staff stretched
- Named by the ward nurses: above about 450 patients, the discharge paperwork waits.
- Operations canceled
- The surgeons' number: planned operations called off when no bed is free.
- Weekend discharge team
- The policy the group chose to test: staff who complete discharges on Saturday and Sunday.
- Discharges
- About 58 percent of occupied beds a week when staff are not stretched. The social workers added the weekend gap.
An example
A regional hospital canceled planned operations most weeks every winter, and each group had its own culprit. The surgeons blamed the emergency department, the emergency department blamed the wards, and the wards blamed social care for slow placements. An analyst's model the year before had been politely ignored.
The chief operating officer brought fourteen people from all four groups into two half-day workshops with a facilitator and a modeler. The group named the variables first, then drew the links on a wall, arguing over each sign. The ward nurses added the loop nobody had drawn: once more than about 450 beds were full, staff were stretched, discharge paperwork waited, and the wards filled further. The group agreed on the pattern the model had to reproduce: beds creeping toward 500 each winter, and cancellations once they passed 485.
Then they tested policies together on the screen. A weekend discharge team adding fifteen discharges a week stopped the slide. Ten did not. The board funded the team for the winter.
Why it works
People act on a model they helped build and ignore one they were handed. When the surgeons watch the nurses draw the loop that links full wards to slow discharges, they see their cancellations as part of the same system, and the argument about culprits turns into an argument about the structure, which can be settled by testing.
The model also holds the knowledge no single group had. The nurses knew about the paperwork, the social workers knew about the weekend gap, and the surgeons knew the cancellation threshold. Agreeing first on the behavior the model must explain keeps the drawing honest, and simulating together shows effects that intuition misses, such as a reinforcing loop that makes a modest change at the right point matter far more than its size suggests.
The thinking behind it
Jac A. M. Vennix Jac A. M. Vennix, a system dynamics researcher, set out the method in his 1996 book Group Model Building, published by Wiley. A facilitated team builds a system dynamics model of its own problem, so the model captures what the members know and the process builds a shared understanding of the problem and a commitment to act on it. The shared understanding, in his account, is as much the product as the model.
George P. Richardson George P. Richardson, a system dynamics scholar, wrote Teamwork in Group Model Building (1995), published in System Dynamics Review, with David F. Andersen. The paper sets out the roles a modeling team needs when it builds a system dynamics model together with a client group, which became the basis of group model building. Its concern is practical: who leads the talk, who draws, and who keeps the record.
David F. Andersen David F. Andersen, a system dynamics scholar at the University at Albany, developed group model building with Richardson, in work where the client group helps formulate the simulation model itself. Their 1995 paper names the roles the work needs: a facilitator, a modeler and reflector, a process coach, a recorder, and a gatekeeper who connects the team to the client organization and its decision makers.
How to run it
- Scope the question before the first session. Agree with the sponsor on one problem, a time horizon, and the groups whose knowledge and cooperation it needs.
- Staff the roles. Bring a facilitator, a modeler who draws as the group talks, a recorder, and an insider who knows the organization.
- Draw out the variables from the group. Ask each person to name the quantities that matter, and put every one on the wall before linking any.
- Map the loops together. Link the variables one arrow at a time, with the group agreeing on each sign, and label each loop as reinforcing or balancing.
- Agree on the behavior the model must explain. Sketch the pattern over time that everyone recognizes, such as beds rising each winter, before simulating.
- Simulate and test policies as a group. Run the model on a shared screen, try the policies people favor, and look at why each one works or fails.
- Write it up for those who act on it. Record the model, the tests, and the decision in plain terms for the people who will carry it out.
Where it goes wrong
The modeler builds it alone between sessions. A model that comes back polished and unfamiliar belongs to the modeler again, and the group stops defending it. Keep the drawing visible and change it in front of the group.
The loudest group sets the structure. If the surgeons draw the diagram, it will show the emergency department as the cause. The facilitator asks each group in turn for the links it sees, and writes down disagreements as open links.
The workshop ends at the diagram. A wall of loops feels like insight, but until a policy is tested against the agreed behavior, nobody knows which change matters. Leave time for simulating and for deciding.
When to use something else
- Causal Loop Diagram. When one analyst or a small team needs to map the feedback in a problem, without the workshop.
- Stock-and-Flow Models. When the model needs real quantities and rates to forecast, more than shared understanding.
- World Café. When a large group needs to surface views across many tables, and no model is needed yet.
Where it sits in SOLVE eX
Go to the source
- Group Model Building: Facilitating Team Learning Using System Dynamics, Wiley, 1996
- Teamwork in Group Model Building, System Dynamics Review, 1995