It’s human nature to want to get moving on new things as quickly as possible. Take a new piece of technology, for example. We immediately want to dive in, experience it, and touch it. Many of us don’t even take time to read the user manual before starting to use a new purchase.
But when it comes to using prescriptive analytics and optimization modeling techniques, it’s important to think before you do. If you start with conceptual modeling before diving into technical modeling, you can make sure that you are solving the right problem at the right level and getting the outcomes that you need.
This applies to anyone interested in using optimization to create more value, whether you are just starting with Remsoft, restarting after a short break from modeling, or you want to leverage our modeling platform to build a new model.
When we start a new optimization project, our clients’ initial reaction is usually, “Let’s talk about data.” or “Tell me how the Woodstock technology works.” But we always explain why this is not the best way to begin.
Instead, we advocate the importance of assessing business drivers and business values before diving into the modeling application and discussing technology and data. We want to first offer suggestions for best practices and share what the Remsoft team has learned through 150+ years of combined modeling experience.
Defining the problem
While some analysis methods, like big data and data mining, start with the data. Remsoft’s technology is powered by optimization and we begin by analyzing the decisions that you need to make. This decision-centric approach requires you to look at the problems you are trying to solve and prioritize them in terms of business importance.
It’s important to understand the desired drivers (objectives), the different actions and treatments available (decision variables), and the limits or boundary conditions of decisions (constraints). It is also necessary to define who benefits from the optimized schedules, who is making decisions, how those decisions are made, what information is required, and what assumptions are being made.
Once you have clarified the scope of the decisions to analyze, you should also determine what insights could be produced to lead to better expected outcomes. This translates not only into defining what schedule of activities you need to output but also into what possible future conditions (scenarios) you could analyze.
Asking the right questions
When Remsoft conducts solution configuration workshops with clients, we combine our extensive modeling experience with the deep knowledge and subject matter expertise of our clients to frame the business problem with discussions. Here are seven key questions we ask our clients to consider.
- What assets are you looking at?
Consider each asset’s characteristics and how they change. (i.e., the asset inventory and its growth/deterioration).
- What is the planning horizon? Think about your main driver. Is it sustainability (long-term), capacity planning (mid-term), or resource scheduling (short-term)?
- What activities and treatments drive value? Also, think through under what conditions they can be scheduled, along with any predefined actions that must be represented (decision variables).
- What is the desired reporting period? Determine the frequency and duration of activities and if schedules should be created weekly, monthly, quarterly, or yearly (period width).
- Do actions generate products that are part of a supply chain? Also, consider whether the products need to be delivered to destinations (allocation problem).
- What constraints and objectives must be balanced? Determine if there are any business rules that limit the scope of activities and what quantities must be balanced (constraints).
- What is your main business objective? Think about whether you want to minimize costs or maximize value (objective) and which activities that drive value and cost you want the model to decide upon.
Another thing to note is that not all scheduling problems can or should be solved with exact mathematical optimization methods such as linear programming. However, both resource allocation and supply chain optimization are well suited to linear programming.
As you can see, there is a lot to consider before you start defining data needs. You can set your organization up for success by first confirming that your problem is amenable to linear programming, identifying the required scenarios and desired insights, and then formulating a conceptual version of your optimization model.
While no one can predict the future, Remsoft optimization solutions can help decision-makers choose the best scenarios possible given the data available at the time. You can think of a model as a digital abstraction of reality, and as a guide it can be a very powerful decision-making aide.