System Dynamics Modelling is an opportunity for entrepreneurs to learn from the complexity of startups’ systems

System dynamics is its ability to help entrepreneurs gain a better understanding of the operating structures of their business as well as learn from the outcome of their past and future decisions. This is important because it empowers them to choose high-leverage policies through controlled experiments (Sterman, 2000). This means that entrepreneurs are able to have their own management laboratory and test their hypotheses before implementing any strategy.

The benefits of SD in the practice is highlighted in the case study of a biotechnology startup firm. John D. W. Morecroft, David C. Lane, and Paul S. Viita have reported their SD modelling project in their paper Modeling Growth Strategy in a Biotechnology Startup Firm published in 1991. This case study shows how SD can be implement in managers workflow and emphasis on the additional value brought by modeling. The article describes how the SD team have involved the business executives of a biotechnology startup in the formulation and simulation of their growth strategy (Morecroft et al., 1991). The company, Bio Industrial Products (BIP), is specialised in selling cleaning biological products.

Creating a shared understanding of the business

Figure 1: Stock and flow diagram of marketing and distribution (left hand side) and customer base (right hand side)

• SD helps managers unify their understanding of the organisation’s structure as well as maintain a consistent thinking. As every mental model is unique, actors are pushing the system in different directions, which leads to implementation failures (Sterman, 1994). It is therefore critical that all actors have a shared agreement on the goal and functioning of the system. In the context of the case study, the decision-makers of the company have been engaged in a discussion in order to create a map of the structure of their business based on their views, assumptions and policies. Figure 1shows the relationship between sales and distributions. Building the map has enabled the management team to identify the “leverage points” (Donella Meadows, 1999) of the system by following the dependencies of its different components. These points are important because they indicate where one should intervene in a system (Donella Meadows, 1999). In this case, ‘delivery time’ appears to be the constraint and act as the link between the company and the market. If the firm accumulates delays in product delivery, the number of distributors will decrease, product sold will decrease over time and eventually the growth of the company will stagnate. When evaluating the impact of the model building workshop, managers expressed how the process gave them more awareness of the business and especially on the critical relationships between distributors maintenance and delivery time (to maintain distributors it was important to understand their sensitivity to time delivery).
• When building the model it is reasonable to ask whether SD can really overcome biases and not introducing them during the formulation process. Although the model is supported by mathematical equations, it was also designed to represent the behaviour of the decisions-makers, it is therefore hard to conceptualise it objectively. The quality of the model still heavily relies on the accurate elicitation of the real behaviour of the system that must be acknowledged by managers. It was nevertheless mentioned how the model enabled to have shared understanding of the business. The process however required a facilitator and several hours invested to fine tune the model. Overall, SD can bring value to managers by stimulating discussions, engaging stakeholders and create alignment on the working of the business, yet, it is a questionable whether managers would take the initiative to engage in modelling in their day-to-day activities.

Running experiments to evaluate the impact of a decision

Figure 2: Comparison of two simulation: The optimistic scenario with the realistic scenario

• SDM provides an edge to managers as it allows them to test their assumptions. It offers an open environment for experimenting and visualising the impact of their decisions without engaging in irreversible risks or outstanding costs.
• Many decisions made keep failing over and over again because agents are not learning from past experiences (Sterman, 1994). In the case of the biotechnology firm, the executive management team was able to explore the relationship between sales growth and distributors sensitivity to time delivery under different settings. Through computer simulation, they tested hypotheses by changing the parameters of the system such as marketing time allocation. SD is neither constrained by the time nor the context in which it is deployed. It is therefore able to display the true state of a system. Thus, managers are less susceptible to their own biases, even if sometimes leaders rather ignore the brutal facts (Collins et al., 2001).
• Learning from the system requires managers to be willing to proactively interact with the model, SD alone can’t actively ‘teach’ the decision-makers about the system. Nevertheless, simulation allowed the company to understand key strategic elements of the business including the importance of the time allocated in maintaining their relationships with its distributors. This was brought to light when the team compared two scenarios (Figure 2): an optimistic scenario where there is no loss of distributors, with a more realistic scenario representing distributors sensitivity to the company’s attention. The sharp decrease between the two figures lies in the change of one parameter representing the “attitude” of distributors toward the company. One can argue that qualitative models can still produce insights that are timely and as much relevant with fewer resources than a full simulation, yet, the strength here lies in the ability of the model to show the instant impact of change as it can be seen in figure 2. In addition, the steady decrease of the growth rate after 24 months cannot be inferred from diagrams only. Although the management team was aware of the importance of the distributors in the whole system, they had additional understanding of how these variables are mutually influential.
• Furthermore, being able to visualise and compare alerts more than a simple diagram. Visualisation has a considerable positive impact on the quality of decision-making as it conveys information in a way that is better processed by the human brain (Zabukovec and Jaklic, 2015). One can argue that today there are several software that provides data visualisation with a greater accuracy than SD. However, the goal here is not to make estimations or predictions but rather understanding the behaviour of the interacting elements of the system and this can be done faster with SD simulation. Because it requires fewer data than other simulation models (e.g. Discrete Event Simulation), the parameters can be changed without focusing on the precision of the numbers.
• In addition, the managing director of the company stated that by testing his assumptions he gained insights on business issues that were not possible through dialogue only. These insights had in turn driven new action to improve the bottlenecks found in business processes. All things considered, managers can learn from controlled experiments by testing policies before adopting them, and therefore reducing risks and uncertainties.
• Their willingness to engage with the system is however necessary but the rewards are worthwhile as they are receiving immediate feedback that would otherwise take months or years in the real world (Sterman, 1994).