Working with a national packaged food manufacturer, Data Driven Supply Chain used AI and decision science to identify additional manufacturing capacity without significant capital investment, reducing dependency on contract manufacturers.
After analyzing current-state production scheduling techniques, Data Driven Supply Chain (DDSC) custom-built an optimization model that recommended a finite production schedule for a snack food plant. Compared to the plant’s original scheduling approach, this optimization model identified approximately 5% capacity improvement potential in the plant, reducing co-packer volumes without adding or upgrading production lines.
Background
Due to COVID-19 stay-at-home orders, a national packaged foods manufacturer saw its snack food demand skyrocket. For several of its snack brands, produced at a single, owned plant, demand exceeded internal manufacturing capacity. As a result, the manufacturer engaged contract manufacturers (co-packers) to close the capacity gap.
With contract manufacturing cost-per-unit higher than the cost of internal manufacturing, the packaged foods manufacturer engaged DDSC to identify bottlenecks in their owned plant and recommend ways to increase manufacturing capacity without significant capital investment.
Considerations
The manufacturer’s owned plant produced several snack foods brands, resulting in dozens of SKUs across multiple flavors and package sizes. The manufacturing process was a complex sequence of cooking the base product, adding custom ingredients for each flavor, then bagging the product in multiple sizes. The bagged snacks were then distributed for sale in grocery stores, convenience stores, cafeterias, and more.
Plant leadership used a production wheel to create the monthly finite production schedule. This production wheel used common-sense heuristics, such as scheduling very long production runs of top-selling SKUs to limit the number of changeovers into those top SKUs.
Downtime occurred with changeovers between SKUs. Changeover activities included replacing ingredient hoppers and cleaning the line of allergen residue. Because some changeovers would take longer than others (e.g., changing from an allergen SKU to a non-allergen SKU, required more cleaning time than the reverse), increasing capacity in the plant hinged on optimizing the finite production schedule and being deliberate about when changeovers occurred.
With this knowledge, DDSC focused on understanding how to improve the finite production schedule and intra-plant work-in-progress (WIP) flow using AI and optimization modeling.
The Data-Driven Supply Chain Approach
The Data-Driven Supply Chain approach involved first understanding the complex nature of the plant’s operations. Changeover times varied by SKU/SKU pair, and a production run required a close orchestration of several sequential activities.
DDSC then began developing an optimization model for finite production scheduling. DDSC was careful to respect and reference the existing production wheel approach to scheduling. Using the Chesterton’s Fence rule as a guide (“don’t remove a fence until you know why it was put there”), this operator-empathetic approach assured the manufacturer team that, while the company already had an excellent way of finite scheduling, the DDSC approach would focus on finding ways to make the process even more efficient and productive.
DDSC’s optimization model output recommended, over the course of a fiscal month, the minute-by-minute finite production plan for the plant, including changeover time specific to each leading SKU / trailing SKU pair.
Model results generally matched the heuristics of the production wheel approach, such as scheduling very long runs for top-selling SKUs; but the model squeezed out about 5% more capacity (measured in pounds of product) than the production wheel during the highest-demand months. This translated into millions of dollars of identified savings by reducing dependence upon contract manufacturing.
Scenario analysis in this model also analyzed how much capacity could be freed up by rationalizing the lowest-selling SKUs from the product portfolio. (Some SKUs required almost as much changeover time as production time).
Outcome
Data Driven Supply Chain built a customized optimization model to recommend a finite production schedule for the snack food plant. This optimization model, which ran on a desktop computer using open-source modeling software and a commercial optimization solver, was transitioned to the manufacturer’s supply chain analytics team for integration into regular business processes. This approach could also be adopted for the manufacturer’s other capacity-constrained plants.
Data Driven Supply Chain provided the packaged foods manufacturer with an AI and decision science approach to achieve greater production capacity out of its existing capital investments, reducing dependence on contract manufacturing and increasing efficiency.