Five in 5: Advanced planning systems in manufacturing

With Brian Barish, Energy, Resources & Industrials principal, and Michelle Davis, Supply Chain & Network Operations senior manager

1. How can advanced planning and scheduling systems enable competitiveness in the manufacturing market?

Brian: All production systems can be described as a trade-off among capacity, inventory and time. If that sounds familiar, it's because it all comes down to Little's law. The trade-offs among these three manifest when variability is introduced in the production environment. Imagine I had a perfectly frozen schedule with material that showed up at the perfect moment and operations that executed perfect cycle times. I could plan for the least amount of capacity and inventory as possible. It would be the leanest operation you'd ever seen. But as soon as variability is injected into the system, these trade-off decisions begin to appear. If the objective is very low lead times given variability, the manufacturing system needs a lot of capacity and inventory to support it.

Maintaining market competitiveness means having a planning strategy that aligns your inventory and capacity with your production times. These strategies are tied together with scheduling. Symptoms of these being out of alignment are long lead times, low service levels and massive amounts of work in process inventory. The most important lever for any operations leader to consider for maximizing the competitiveness of their manufacturing operations is scheduling.

2. What are the most important considerations when evaluating if an advanced planning system is right for my organization?

Brian: If you rewind the clock back to the 1980s and 1990s, creating a daily schedule was computationally complex. Daily schedules required an understanding of multistep operations, routings, task lists, capacity limits, machine requirements, etc. It was nearly impossible to sequence and time jobs with any degree of accuracy. Therefore, computers would provide schedules at a much rougher cut and lower level of fidelity—with humans necessary to intervene and apply intuitive intelligence to work around edge cases. As computation capacity exploded, computers became more equipped at detailed, higher-fidelity operational scheduling. But just because the computer can do more detailed scheduling, should it?

Today, we see a lot of organizations struggle to provide the right quantity and quality of data to support advanced planning. For clients who have fully integrated autonomous lines where everything is controlled on a conveyor and tasks can be executed down to the minute, a daily operations schedule may be right for you. But in my experience, for many US manufacturers, there will always be a human in the loop, and that planned intervention needs to be appropriately accounted for. It’s important to rightsize the schedule to the organization’s needs. Asking a planning system to do more than its master data can support can drive just as much churn as not having a schedule to begin with. This means knowing strategically when to schedule at a lower fidelity and accept variability in the system.

3. In this Deloitte Dbrief, you discussed system nervousness as the change in plans that occur due to the continual rescheduling of orders. How can planning systems be better designed to manage system nervousness?

Michelle: There really is no such thing as a perfect schedule. Clients can spend a lot of resources designing and implementing a system to get to the "right" answer, but perfect plans go out the window the second they hit the factory floor. One of the most important decisions an organization can make is determining what degrees of freedom they will accept for planning horizons.

In the Dbrief, we discussed liquid, slush and frozen time periods. In the frozen time period, no alterations should be made to the demand forecast as the planning period is within lead time. Building on Brian’s points, when planning systems (computerized or human-driven) become too involved in daily execution, organizations may experience increased variability or system nervousness.

4. How can an organization begin to dial in what the right liquid, slush and frozen time horizons are for them?

Michelle: Once an organization defines its inventory, capacity and cycle time strategies, there are granular decisions to be made to maximize their internal production performance and external supply base. For example, for an organization with internal production systems that have very high fixed costs for setups or changeovers, the frozen horizon should be on the longer side to minimize the frequency of these big investments. These investments could be in terms of scale, complexity, skilled labor requirements, etc. The other thing to consider is the length and agility of the supply base. Organizations that have highly specialized, inflexible or long-lead-time suppliers have a greater need to balance work in progress inventory with schedule flexibility. It’s not uncommon to have more than six months of sourcing and more than nine months of supplier development ahead of their most complicated products. Thus, there’s an impetus to plan and lock in their demand and production schedules well ahead of time to provide stability to both the suppliers and the plant.

5. Are we seeing clients deploy AI solutions to address production and operations planning today? What tools are they using, and what’s the impact they’re seeing as a result?

Michelle: We’re seeing a lot of clients dip their toes into the AI water. Advanced planning and scheduling are difficult because there are many constantly changing variables. Clients can be very successful using Generative AI (GenAI) to identify their data and information gaps, provide simple directions to close the gaps, and then perform rapid simulations on this foundational planning data.

Brian: An AI model will rapidly search for any inconsistencies in your data. For example, I would use AI to make real-time system adjustments to manufacturing durations based on observed production data.

Michelle: We recently built an AI tool for rapid data auditing that searched through datasets looking for missing or inconsistent data, such as vendor spelling differences. The information was then presented to the planners, introducing a human in the loop. Once approved, the AI made the system changes. It was a quick and easy way to clean and maintain planning parameters, heavily decreasing manual intervention. One great use case for an AI tool that overlays many disparate datasets is to quickly disseminate information that was historically siloed, kept on hard drives, or never formally captured in a database structure. An AI tool can quickly query this information and present it to a human, removing the complexity of data engineering and redundancy of data entry.