Creating Useful Models

Models attempt to capture the essential nature of a small part of reality to promote understanding. The industrial statistician, George Box observed in 1979 that “all models are wrong, but some are useful”. This begs the question how do you create a model which is useful and fit for purpose? We believe that successful modelling activity depends primarily on the ability of the modeller to extract useful information from the heads of the people affected by the problem. Translation of that knowledge into the language of the toolset is an important but secondary factor.

Adopting the following guidelines, which are adapted from John Sterman’s book 'Business Dynamics', will help to ensure the value and utility of your models:
  • Develop the model to solve a problem do not create a general model of the organisation.
  • Integrate modelling into the change initiative from the beginning
  • Challenge the need for a modelling approach to make sure that the techniques that will be applied to solving the client’s problem are the right ones
  • Do not rule out the use of other techniques such as benchmarking where they can add value.
  • Keep implementation in mind throughout the project
  • Modelling works best when it is treated as a iterative and as a joint initiative between client and consultant
  • Keep the model visible when it is being developed and make sure that clients are involved early and remain deeply engaged with its development
  • Validation means testing often and regularly
  • Get a preliminary model working as early as possible and then add detail as needed (but no more than that).
  • Capturing the feedback loops in play is more important than creating a highly detailed model
  • the tools are easy to learn to use, but don’t use novices to construct your model. Use expert modellers who are experienced in business, facilitation and who understand mathematics.

The other key success factor is the following of a disciplined process. An outline of the steps that we go through follows. These map neatly onto our end to end process. It should be kept in mind that this is not a linear, step by step formula. Once you have gone through Start Up, there will be many iterations around the stages as new information emerges that must be taken into account.

Start Up

As with any initiative, an effective startup increases the chance of a successful project. Typically, startup activities will include bringing stakeholders to the same level of awareness about the problem and how it is going to be addressed. Familiarity with the tools and what they can do will also be on the agenda. Playing the Beer Game is a popular ice breaker which exposes people to the concepts.

Describe the Problem

Motivated by undesirable performance, the problem is formalised in terms of issues, concepts, key variables, structure and relevant policies. A hypothesis about the problem’s cause is created alongside an initial model. The boundaries of the model and its time horizon are established. This is a team endeavour supported by the use of large group techniques and mess mapping.

Map The Structure

Based on the initial hypothesis, the model is developed to show stocks and flows and feedback loops. All variables are defined, and relationships between levels and flows are described with equations ensuring consistency of units as this is done.

Simulate

Early simulations may demonstrate unexpected behaviour which is the trigger for returning to the hypothesis, structure and equations with a view to refining them in the light of improved understanding. This iteration may happen several times until the model adequately reproduces the problem behaviour and reflects it’s historical roots.

Challenge Management Thinking

Most misbehaviour of corporate, social and governmental systems arises from erroneous intuitive thinking about complex non-linear behaviour. Sufficient understanding should have been developed by this stage which should allow the team to present their findings to the management team. Invariably this will lead to preconceptions and assumptions being challenged as reality is hauled blinking into the sunlight.

Having presented the evidence, attention is turned to solutions often in the form of new policies, changes to existing policies, changes to the structure or changes to the way decisions are made. The most promising scenarios can be tested through simulation, ensuring that policies interact optimally, de-risking the selection of a way forward. Proposals for policy change should be encouraged from all corners of the organisation but most especially from those affected by them.

Communicate and Educate

This step works towards a consensus for implementation. Events based on the use of learning environments are used to disseminate the insights gained by the modelling team as far and wide as necessary. The model will demonstrate how current problems are being caused and also demonstrate how the chosen solution deals with them. Sometimes the model is presented as a flight simulator and the audience is given the opportunity to learn by adjusting parameters for themselves. This can be a very powerful learning experience.

Implementation can mean reversing policies which are deeply embedded within the organisation and changing strongly held beliefs. This requires a period of intense education and debate and sometimes revisiting earlier stages as new information comes to light.

Implement

At this point, the agreed changes are implemented. Success will largely be determined by the thoroughness and attention to detail of preceding stages. All of the tools in the change management toolbox have applicability here and implementation cannot be rushed if changes are to be embedded in a lasting way.

Learn

As implementation completes, evaluation and learning follow. It is important that actual results are consistently compared to those produced by simulation and any differences explained with adjustments to the model being made if necessary.

Once a model has been used to solve a particular organisational problem, It is not unusual for there to be a move to expand it to accommodate other areas of the business. While this is not necessarily wrong, caution is required as expanding the scope of the model may well increase its complexity to the point where it becomes difficult/impossible to maintain.


Legal | Privacy