It was the 1950s and General Electric was the largest appliance manufacturer in the US. In one of their factories in Kentucky, the managers were struggling with what seemed to be an intractable problem – there were times of the year when the factory would be working 3 or even 4 shifts, and then there would be other times when they had to lay off workers. It was easy enough to explain this away to demand fluctuations – except that Jay Forrester, then a professor at MIT, thought it was too simplistic an explanation. An Electrical Engineer by training (and one of the many brilliant minds who collectively laid the foundations of modern computing with their work in WW-II), he had studied feedback control systems – and that had taught him to take a systems view of problems, as opposed to focusing on a narrow sliver biased by a localized observation. In his own words, ‘After talking with GE managers about how they made hiring and inventory decisions, I started to do some simulation. This was simulation using pencil and paper on one notebook page. It started at the top with columns for inventories, employees, and orders … It became evident that here was potential for an oscillatory or unstable system that was entirely internally determined. Even with constant incoming orders, one could get employment instability as a consequence of commonly used decision-making policies.’ Over the years, this specific problem came to be known as the ‘bullwhip effect’ – perhaps more importantly, this project helped Jay Forrester crystallize the idea of ‘System Dynamics’1. Yet another brilliant example of inter-disciplinary thinking: the ability to apply ideas from one area (electrical systems) to a seemingly different one (inventory management) – the ‘long connections’ that often result in breakthroughs.
Systems Dynamics: what is different?
We need to start with our current mental models that usually frame problems as linear cause and effect. This event-oriented view of the world is helpful in that it helps us keeping moving the ball by operating in discrete problem-solution steps (after all, it is much easier to fund projects that promise a clear solution to a clear problem in a pre-specified time). And usually, reality doesn’t play out in such a simplistic, linear fashion:
- There is a time-delay between cause and effect: decisions take time to implement and during that time, the problem itself changes its shape and form
- There is a feedback loop: current decisions and actions shape the future situation, which in turn changes the impact of the actions
- There are other agents with their own motives and actions (within the organization and beyond) that influence the current problem space
All these put together, create a problem ecosystem that is fundamentally more complex than the simplistic linear view and forces us to take a Systems view of problems. It is this approach that makes Jay Forrester’s body of work so important and foundational to problem solving. Over the years, System Dynamics has grown to be a hugely important field with applications in inter-disciplinary sciences – biology, ecology and so on.
What does it mean to us?
It is obvious that this is a powerful way of looking at problems in the Enterprise – even more in our current world of ever increasing complexity. Yet, Systems Dynamics has found limited use – mostly because we continue to be comfortable with linear cause-and-effect mental models: in other words, we choose simplistic methods over the more comprehensive ones: which is not always such a bad thing, especially when the problem is localized enough. Here are a few of my suggestions:
- System Dynamics is useful when you are trying to answer ‘policy level’ questions which have far reaching implications. This is when you are looking for directional answers to help make long term decisions – e.g. should you write-off all delinquent vehicle loans that meet a certain set of specific criteria and if you do that, how does it impact downstream behavior for these customers who may have other products? This is obviously impossible model with a statistical model or even learn from a controlled field experiment
- System Dynamics is not a good option when you are trying to answer a specific question: e.g. what is the projected lift from an email campaign? Even though we all know that email campaigns have a decay effect and in many cases, external factors may strongly impact customer behavior. (especially at times like these where things change dramatically overnight) In such cases, System Dynamics could be an overkill – you might be better off with simpler solutions
- The time and effort it takes to define, setup and most importantly, gather the data to be able to effectively create a simulation model can be daunting: so it behooves the age-old trade-off between efficiency and effectiveness of building these complex models
- And finally, a good place to start is to get your teams to develop some of the underlying ideas behind System Dynamics – they are extremely useful in framing and solving problems in a more effective manner:
- The concept of endogenous vs. exogenous factors: System dynamics seeks endogenous explanations – in other words, don’t always take the easy way out of lumping explainability into ‘environmental factors’ or worse, error functions
- The concept of a time horizon: decisions take time to play out and any modeling exercise needs to factor the time dimension in a meaningful way.
- The concept of simulation: Develop a quantitative model that can be used to reproduce the problem and the future conditions that may arise when you implement the solution. This is a big area in itself – deserves a deeper discussion
One thing is certain: as we continue to work through this ‘new normal’, System Dynamics could be an important tool in your problem solving armor.