scroll
Predict, Not Just Respond

Algorithmic Supply Chain Planning – Food for Your Supply Chain

An algorithm is simply a procedure (sequence of steps) that ends at some point in time. To this end, your cooking recipe is an algorithm. Much like cooking recipes, some algorithms end in burnt or horrible tasting food. Therefore, how an algorithm works is what matters and what the end result defines the quality of the algorithm. Gartner refers to Algorithmic supply chain planning as those which are complex and have the ability to improve themselves through machine learning techniques. We refer to the latter as Self-Improving supply chains©.

The most popular tool for algorithmic supply chain planning is a spreadsheet, which happens to be the least desired because of the quality of plans produced. It is inaccurate, does not take all the variables into account and cannot account for the mix of products or realistic lead-times. It is also incapable of handling both material and capacity constraints at the same time! Unfortunately, many of the popular Sales and Operations Planning S&OP plans out there follow the same principle as spreadsheets resulting in plans which are far from accurate. And if you were to use them you would need to add a lot of ketchup and seasoning to make it possible to swallow. However, these packages look nice and have much better user-friendly interfaces than spreadsheets. But that does not change the end result, inaccurate plans that need a lot of work to make them executable! The latter is done through trial and error as well as what-if scenario planning.

Attributes of a good algorithm are flexibility, speed and scalability, the ability to model the real world, and of course accuracy of the end result. It is not unusual to fall into the pretty user interface trap and feel that this will make my job easier and faster. Quite the contrary, what makes a planner’s job easier is when accurate plans are produced fast and there is not that much need for making changes to the final plan or so-called what-if scenarios. Think about it, if the plans are risk avert and accurate, what is the need for what-if type of activity? The system can take into account millions of variables and produce plans which are optimized and has already taken into account potential risks. The latter means if the plan does not materialize because of a potential issue of a supplier or resource or weather, then the system has already a backup plan!

Back to algorithms! As you can see, good algorithms make a difference in your life and can help you to go home a lot earlier to spend time with your families instead of “massaging” the plans that are supposed to be optimal! Secondly, learning algorithms go one step further and keep getting better and adapting to their environment. So, as the supply chain changes then the system (or algorithm) changes with the underlying trends and therefore remains relevant. To use our earlier analogy, good learning algorithms keep providing you with good food and changing itself and adapting to what you feel like eating as opposed to feeding you the same menu day in day out. We conclude that Good algorithms give you consistently good tasting food no matter what the ingredients are. On the other hand, good learning algorithms give you better-tasting food that changes as your environment (or supply chain) changes.

For more information on self-improving supply chains and algorithms, click here Intelligent Planning

Share:

Share:

Leave a comment