This weekend, I was driving somewhere. I had been to the town before, but not to this address. So, I plugged in the address into the GPS and off I went. GPS reported some traffic along the way that I usually take and took me through some back roads. Unfortunately, I had car trouble along the way and had to call for help. When a local car repair guy showed up, he asked me where I was driving from and where was I going to. When I told him, he was very surprised that I even happened to be on that road. When I told him that I was just following the GPS, he nodded and went about his work. Once he left, I started thinking about the conversation.
Update on October 15, 2019: CNN News story
Last week, my blog post on the topic of Optimization and suggested that optimization can sometimes lead to counterintuitive results. However, this experience made me think that the use of optimization (or other) models can also simply show new ways of getting things done. It is just that with our limited knowledge, we had never seen these pathways.
A Real-Life Supply Chain Optimization Example
Let us look at an example from the supply chain world. Arkieva had a customer in the plastics (chemicals) business. They had 6 older manufacturing facilities (about 30 years old) and 1 new facility that was commissioned about 5 years prior to us getting involved with them. They wanted to optimize the production given the existing network. Over the last 5 years, they had fallen into a pattern of usage where the newer plant produced all the high-volume products, and the older plants produced the ‘dogs and cats’. This was done so because it was assumed that this would get the most output out of the fancy (and expensive no doubt) equipment that had been installed at the new plant.[Read more: Who’s Guilty? Dormouse, March Hare or Mad Hatter?] However, once the optimization model was put in, it did the usual tradeoffs. It could see that the older plants were prone to break down when they ran too many products. It also was able to use the information that overall, the transitions from grade to grade was faster on the newer plant. Because of these and other factors, it recommended a different way of doing things. Select 1 or 2 high volume products per older plant, and load the newer plant with the dogs and cats. This allowed the older plants to run without breakdown and provide consistent throughput. The newer plant was able to take care of the various small volume demand very well without breaking down. Staffing requirements in older plants went down as well. Overall it provided much better results to the business.
Was it counterintuitive? To that business, it was. And how about a new road? Absolutely.
Of course, the story does not end at the model. We had to go through a very robust exercise of convincing the users to accept the solution. But that is for another blog post.