Automating the Model Building Process
The holy grail of supply chain design is to have a high-fidelity ‘digital twin’ at your fingertips. From this living model of your physical supply chain, you can quickly engage in scenario planning by pulling the supply chain levers under consideration and assessing the impact to key metrics. While extremely valuable insights are unlocked by applying scenarios to the digital supply chain twin, this metaphoric pot of gold is far-too-often at the end of an onerous trudge through data management and baseline modeling.
With such extensive effort expended on data management and baseline modeling, the ultimate question becomes, “How can you accelerate the time to value?” As with any defined process, data management and baseline modeling can be largely automated. Moreover, the upfront effort of such automation may be less than you would think.
Whether supply chain design is a core competency or a support function within your organization, there are likely some common themes across all of your modeling initiatives. Perhaps they consistently revolve around decisions of capital expenditure, how to best serve a new market, or the impact of rationalizing assets. While the investments, geographies, property, and equipment in question may be different each time, the structure of the automation pertains less to the data and more to the business question itself. To this end, automating the model building process creates value in the following ways.
Repeatability Over Time
Repeatability over time is perhaps the most obvious reason to automate. If you know you will be refreshing a model for supply chain planning with any regularity, why reinvent the wheel again later? Broadly, there are two use cases for refreshing a model — routine-driven and event-driven — and while the former is generally known from the start to require a future refresh, the latter is often realized only after the fact. Below are some examples of each:
- Routine-driven (e.g. annual capital budgeting; weekly sales and operations planning)
- Event-driven (e.g. a global pandemic; fuel price spike; unexpected disaster such as port strike or facility fire; political uncertainties; availability of new commercial real estate)
Repeatability Across Markets
The quality of the insight you provide through your models will undoubtedly stimulate a thirst for more within your organization. Once you’ve proven the value of your supply chain analytics through a specific project, it will be prevalent in leadership’s collective mind to call upon you for similar value in the future. Perhaps the examples below sound familiar:
- Geographies (e.g. “Can we perform a site selection analysis for Australia just like what was delivered for Brazil?”)
- Business units (e.g. “We’d like us to rationalize sourcing assignments for glass, just like we did for aluminum.”)
Remember, while the input data itself may change, the structure of the model and vast majority of the corresponding automation pertains more to the type of business question itself. Furthermore, your leadership will be able to more quickly digest standard outputs in a standard manner to how they were constructed for prior projects.
Repeatability Across Team Members
Help someone else the way you wish someone had for you. Help set yourself up for success. Help your team grow and evolve.
- Facilitate specialization and division of labor (e.g. allow more senior team members to focus on modeling approach and scenario analysis by transitioning execution of a data workflow to more junior team members)
- Accelerate recall of an older project via workflow self-documentation
- Enable succession and continuity for planned or unexpected organization and personnel changes
Through this lens, automation ‘enforces’ standardization and the accompanying benefits. Particularly useful is documented workflow visualized through a lexicon of intuitive icons representing specific data transformations linked to the respective inputs and outputs of each process step.
Repeatability Within a Project
Even in circumstances where design projects are legitimately ‘one-off’ requests, the low barrier to automation coupled with the likely need for iterative re-querying of data, adjusting filters, and modifying aggregation structure may still prove a net benefit within a single project.
- Even if the model approach and structure is not-fully-defined, automation makes experimenting and testing a highly modular and configurable activity.
- The self-documenting nature of automated workflows provides visibility to key data decisions and assumptions used in the model-building process.
Automating the model building process accelerates the time to unlock the valuable insights desired by you and your stakeholders. With this, the question truly becomes, what are your reasons not to automate?