Strengthening your supply chain one link at a time.
The Makings of a Good Network Analysis
When thinking about what makes a good network analysis (also referred to as a network optimization or network study), I remember the famous line from “The Karate Kid” (the original version), when Mr. Miyagi says that your life must have “balance.” The same can be said for your network analysis. Having a balanced network model is very important when trying to optimize your supply chain. You can achieve proper balance during the all-important baseline modeling process, which is a critical step in making sure your model can replicate history before you turn it loose on future-state scenario modeling. Overstating or understating baseline modeling costs, as they compare to historical spends, can lead to trouble. Having a balanced baseline model, as it compares to historical flows and costs, will make sure you have not built an unfair bias into the network model. Some examples of this balance include:
Transportation costs versus fixed facility costs (building and indirect labor). If your model overstates fixed facility costs, compared to transportation costs, the process may result in fewer than the optimal number of locations.
Inventory carrying costs versus all other costs. If your model overstates inventory carrying costs, which usually occurs by having too high of a carrying cost percentage, it may be difficult to justify additional facilities, when in reality, this might be the optimal solution. Outside of the food industry, this percentage is typically around 8 to 10 percent.
Inbound transportation versus outbound transportation. If your model overstates outbound transportation costs compared to inbound costs, the model will tend to locate your facilities closer to the customer. If you understate outbound transportation costs, the model will tend to locate your facilities closer to the product source (manufacturing/vendors).
Aggregation strategy. This relates to both customer and product aggregation. If you aggregate too much, you can lose accuracy, and you may lose trust in the validity of the results. If you do not aggregate enough, you may encounter long run times that make it difficult to debug the model and run the number of scenarios required, leading to an increased project schedule.
While balance is important, it is not the only thing that makes for a good network analysis. Having over 25 years of modeling experience, and having completed well over 100 network modeling projects, I have seen mistakes that others have made in this process.
Data. The saying, “garbage in, garbage out,” is also true when performing a network analysis. This goes for both data availability and data quality. Lack of either, or both, can lead to making too many assumptions. Good and valid assumptions are certainly necessary and an important part of any network analysis, but making too many can lead to a worthless exercise.
Executive buy-in. The results of these studies, whether they are distribution-focused, manufacturing-focused, or both, will shape the future direction for the next five-plus years of your organization. Having executive buy-in to the process and growth strategy is critical to making sure this project does not end up in the circular file (trash can).
Candidate locations. This is one of the most common mistakes I see from inexperienced modelers. I have heard more than a few executives comment, “doesn’t the network model just pick the locations.” If you are running a center-of-gravity analysis, that would be true, but if you are performing a full supply chain network analysis, you need to provide legitimate choices for the model to pick from. For example, I once performed an audit of a network analysis performed by another company. In their analysis for the U.S., they only had one candidate west of Memphis, TN. Not shockingly, it picked that candidate every time to service the west. That is like asking someone to pick a number from one to one.
There are so many other factors that differentiate a good network study from a bad one, such as network constraints (service and facility), inco terms, profiling, etc. (we’ll save those for another blog post). As Mr. Miyagi once said, “Person who catch fly with chopstick accomplish anything.” Well, I think I’d prefer a good network model to a set of chopsticks to optimize your supply chain network, but to each their own.