This particular case example involves leveraging a set of data points to optimize production planning based on demand, specific price points and capacity across multiple facilities. As further background, the project required cost elements to include product, transport, duty and other related areas. In addition, supplier lines had to be qualified for each product before any allocation could be made and the ultimate "expressive pricing" included regional variances, volume bands, and multiple discount structures. In other words, the ultimate solution required significant complexity, even compared to hairy sourcing-specific optimization events.
Garry told Spend Matters that the "problem presented and solved was to supply over 500 components to any one of over 30 manufacturing plants to meet a demand plan for finished goods in the USA FMCG arena. The planning horizon for the finished goods and the components they are comprised from was a 5-year rolling demand forecast ... the requirements of the client and the products themselves made this a problem beyond most planning applications on the market today."
Specifically, "the task was simply to produce the lowest-cost solution to meet the provided demand forecast." But as Garry suggests, "those of you familiar with such things will know problems of this nature are well known as being 'computationally hard'."
Stay tuned as we dig into the situation and outcome in Part 2 of this post.