I was going to start with a joke I heard about sodium, but then thought, “Na”.
I’ve never tried to write a book but I imagine the author has a good idea about how the plot will end. Then they put all their creativity and skill to build the story, develop their characters, and of course create strands of unexpected subplots to craft an inevitable twist.
After you’ve understood the problem and decided that a simulation model will help you develop and assess solutions, the questions to ask are: “What information do I need the model to produce?” and “What’s the best way of presenting that information?” I can safely say that choosing to present the solution simply as a table of numbers is unlikely to be the most useful end product. As performance is described by multiple indicators, finding a way to communicate the combined impact is one of the most important decisions you need to make.
The standard way of using a model is to run a number of scenarios, each with different process data, business rules or assumptions, and examining the results. As it’s not unusual to run 100s or even 1000s of different scenarios, it’s easy to get “the paralysis of analysis” and begin to get lost in the quantity of data being generated. It’s useful to think about:
- Comparisons – how the indicators compare across a number of scenarios
- Contra-indicators – how indicators respond differently; for example throughput might increase but quality might suffer
- Responsiveness – indicators might vary at differing rates; if throughput increases at a faster rate than quality falls, there may be a “sweet spot” where the total value of quality product is a maximum
“I wouldn’t start from here” ― a stranger giving me directions
A more unusual way to start at the end is to do this in the model itself. Similar to backcasting, this is a planning technique that starts at the end – specifying a desired future, and working backwards to identify the processes that are needed to make that happen. Whilst that sounds a bit like “Back to the Future”, it’s really a very constructive way to think; I’ve even built models that actually worked backwards!
In one example in the steel industry, a multi-step process was driven by a production schedule. Product changes resulted in significant changeover losses, so getting the right mix, volume and sequence of product to maximise production, minimise loss, and meet customer demand was essential. The production steps included multiple furnaces and continuous casting separated by buffers, with the system constantly “looking forward” to try and determine the best quantity and timing of the next product to launch. Instead of following the process from the front, we designed a simulation model that started with the customer demand and used this to define the end product quantities and timetable. This was then pushed back through the process and the model generated the raw materials and timetable at different stages. The company could then implement this timetable and be assured that the only losses were intrinsic to the process and not due to poor planning. And of course the output exactly matched the customer demand.
An even more stunning example of backcasting was a model developed for a nuclear waste management company. As you can imagine, decommissioning is an extremely complex multi-stage process, and takes place over many, many years. The focus of the business is to improve overall safety by remediating the waste and decommissioning plant as early as possible. The simulation model I designed presented the decommissioning of multiple waste streams as a Gantt planning chart, which was the “report” that managers used and understood. The smart decision was to make the Gantt interactive. The manager could click on any start or end date, and drag it forwards or backwards. The model instantly responded by adjusting all affected process’s capacities, and identifying which business rules needed to be changed to achieve the revised plan. Rather smartly, it also showed that improving the safe condition of one waste stream might have the undesired effect of delaying the processing of another waste stream.
“…if you are not absolutely sure of a thing, it is so difficult to commit yourself to a definite course of action.” ― Agatha Christie, Murder Is Easy
Modelling is about making decisions. Get the model right and the results will be valuable. Presenting the results in a way that builds confidence to make the decision – well that’s priceless. You too might burst into song.