Advances in Supply Chain Analytics
With the advent of edge devices and the streams of real time data that they provide, modern Supply Chain and Inventory Planning has been revolutionised. Huge advances have been made with just-in-time production, cross-border manufacturing, and predictive analytics ensuring that businesses can look forwards and plan adequately during periods of business-as-usual. The analytics involved in areas such as demand forecasting and growth trends have advanced significantly with the advent of Big Data and Machine Learning. But as good as these near time analytics are, they are still fundamentally based on backwards looking data, and when decisions need to be made to prepare for fundamental change, the best tool remains human intuition. But the world is fast paced, and even the best business leaders can struggle to cope with the rate of change, or to make the right decision when there is a tectonic shift of which they have no prior experience. For example, what happens to inventory levels when there are structural changes to logistics, or when trade deals change the rules for export/import? The complex nature of supply chains creates significant challenges for decision makers when the business is hit by these “Black Swans”.
Understanding Complex Systems: Lessons from F1
It might not at first appear to be obvious, but there are clear parallels with the challenges of fluid dynamics. Small changes to the shape of even a single component can have outsized effects on the aerodynamics of a car. Consider Formula1 cars that have been optimised to function within a specific set of rules. How do these teams set about re-defining their aerodynamics packages when there has been a fundamental shift in how their cars can function? The solution is that much of the heavy lifting is done using simulators that create a sandbox for testing ideas. Underpinning fluid dynamics are a set of equations that are used to model the flows of gases or liquids, that when combined produce complex and non-intuitive results. But the huge benefit of creating a mathematical model to replicate these systems is that computers can then be asked to optimise for a specific outcome. In effect, the computer model can determine what the right aerodynamic shape would be.
Understanding Complex Supply Chains
Systems Dynamics is a methodology for studying the relationships between objects in a system and has been used to model everything from ecological systems to national economies. System Dynamics lets us define complex systems of mathematical equations using a visual language. With this approach we can model complex supply chains and simulate into the future while capturing all the time-sensitive complex interactions between the components of the system. We can then use those simulations to run what-if scenarios.
The remainder of this article demonstrates how to build a relatively simple systems dynamics model of inventory management and to show some of the real-world insights that Silico’s technology enables.
Step 1 - Setting up my Inventory
The first step in creating a model is to define the individual elements as Stocks (things that accumulate and decumulate) and Flows (things that modify the stocks).
We can create a very simple dynamic where goods accumulate according to our orders, there is a delay until delivery, and the goods then accumulate as Inventory.
As you can see the graphical interface that Silico provides makes it simple to create this relationship and define the objects. All of the complex maths is hidden away from the end-user so you can simply focus on describing the business.
Step 2 - Define the Dynamics
From here we can start to add some of the dynamics that we are trying to model. In this example we can add some customer dynamics and our desired inventory cover.
With simple drag and drop functionality we can join the different model elements. We have joined the number of customer orders with the number of weeks cover that we require. This calculation then feeds into both our Items on Order and our capacity to deliver.
Step 3 - Adding Adjustment Delays
The next step is to add in some adjustment delays, which is to say that there is a delay to my ability to make Inventory adjustments, due to delays which are inherent in any supply chain, as well as in the demand forecasting process. Modelling time delays is straightforward in this framework which confers significant advantage over non-dynamic modelling tools such as spreadsheets.
As you can see, these time delays will then flow through the model, and depending on the structure of the relationships, might generate reinforcing feedback loops which manifest in complex phenomena such as the bullwhip effect.
Step 4 - Adding Commercial Structure
The final step for this model is to add some commercial aspects. If we wish to understand how inventory changes might change the P&L then we need to identify this in the model.
We are able to add some assumptions around our cost of holding inventory and we can also add any costs of unfulfilled orders that are represented here by the ‘cost of non-delivery’. This lets us start to model the trade-offs of holding inventory.
Step 5 - Adding Values and Formulas
For this model we have created a number of downstream calculations that are triggered by my number of customer orders. For the purposes of illustration we have defined this as 200 orders with a step up by 50 orders after 5 months.
We can add any formula or values to make this as realistic as possible, including importing actual order data straight into Silico. This ensures that we can define the behaviour of our model as it steps through time.
Step 6 - Dynamic Dashboards
In 5 easy to follow steps we have created a dynamic model of inventory management.
As we run our model through time the Silico dashboard lets us play with any of the key model parameters and get instant feedback on the impact on any of the metrics or KPIs modelled.
If needed, we can also export the data that we have generated in Silico into other data science workbooks or visualisation tools for further analysis.
Exploring the Results
Even in this relatively simple and easy-to-build model we are already able to produce some powerful insights. In the dashboard that we have created we can add a slider to change the dynamics behind our supply and demand and examine the result this has on the model. This is represented by the ‘Inventory Cover’ and ‘Delivery Time’ sliders at the top of the Dashboard.
As we move the slider from left to right, we change the inventory cover and delivery times. As we can see this does not have a straightforward effect on our items that we receive or those that we are unable to deliver. This is because the relationships between order and delivery are non-linear due to the time delays involved in several areas. Under certain conditions these dynamics come together to create unusual results. This type of insight, one that takes into account second and third order effects, is incredibly powerful and is difficult to achieve so easily with other tools.
The obvious next step now that we have generated these insights is to try to optimise our system in order to produce the best possible combined outcome. Silico also provides optimisation tools to ensure that this can be achieved. In running an optimisation analysis you can define the outcomes that you want your business to achieve, and allow the Silico solution to search for the right mix of the various levers at your disposal to produce that result.
Inventory Management is just one area that Silico can help with. This same approach can be used to explore other business challenges, from Workforce Planning to Marketing Analytics. Even more powerfully, simulations from across business units can be linked together and, for the first time, give you a truly holistic framework on which to base important decisions.
For further information on Silico please visit www.silicoai.com