Beware of using optimization without examining what happens if your assumptions change! Unintended consequences could arise. This article is based on a nice paper by some masters’ students at MIT.
The optimization program relies on the assumptions, which are input as data (for example, anticipated demand levels here). The optimization predicts what we call a discrete solution (for example, open this warehouse, close that one). But these changes, in practice, cannot be made easily. It’s costly to keep opening and shutting warehouses; how frequently do we want to do that? So the ‘jumpiness’ of the solutions as assumptions change plays a large role in our decision making, more, actually, than the predicted savings from the optimal solution.
Where the models excel, though is in allowing us to create many scenarios (sets of assumptions), and play them through the model, recording the results we get. Then the decision makers and analysts can examine the range of answers they get to extract a rationale for making their decisions. It’s an art as much as a science. And it’s a lot better than experimenting in real life by closing warehouses or opening them, with far less cost, anxiety, and risk.
That’s why simulations are useful. Out here in Santa Rosa we wouldn’t build a bridge without using a computer to simulate what would happen in various types of earthquakes. Why wouldn’t you do the same with your supply chain?