MRP Can Plan the Materials but Not the Shop Floor: What APS Actually Fixes

Plenty of factories have ERP and run MRP every night, and still find that the plan the system produces cannot be run. The material requirements are there. But which order to start first, on which machine, at what time? MRP has no answer. So planners go back to Excel, and the system becomes decoration.

The cause is a built-in assumption of MRP: infinite capacity and fixed lead times. It answers one question, "when should material be released?", and assumes every machine can always fit the job, changeovers take no time and everyone can do every task. Reality is full of constraints: machines are finite, changeovers stop the line, operator skills differ, and there are only so many sets of tooling. Those constraints are exactly what makes a plan "unschedulable". APS (Advanced Planning and Scheduling) exists to produce an executable plan under those real constraints.

Who does what: four layers of planning

To place APS, separate the four layers: ERP manages money and orders, MRP manages material, APS manages sequencing, MES manages execution. ERP handles orders and resources at the month and week level. MRP calculates material needs at the week and day level. APS refines the plan down to each operation, each machine and each minute, sends the schedule to MES, and takes real-time progress back to re-plan dynamically. Without the APS layer there is always a gap between planning and execution.

How APS schedules

APS is not first-come-first-served. It weighs rules and algorithms against finite capacity. Three basic approaches are usually combined:

• Forward scheduling starts from today and finishes as early as possible. It suits situations where delivery is the priority.
• Backward scheduling works back from the due date and starts as late as allowed. It minimizes inventory.
• Bottleneck scheduling (TOC) schedules the bottleneck resource first and lets everything else follow its rhythm, because the bottleneck sets the output of the whole line.

Add priority rules such as EDD (earliest due date), SPT (shortest processing time) and CR (critical ratio), plus batching of similar jobs to reduce changeovers, and APS can build a sequence that balances delivery and efficiency.

What it is really worth

APS brings into one calculation the constraints that make MRP fail: capacity, material availability, routing sequence, machine capability, operator skills, tooling and changeover time. The goals it must pursue at the same time (on-time delivery, equipment utilization, minimum inventory, fewest changeovers) often conflict. Pushing utilization up usually raises inventory; squeezing changeovers can cost delivery dates. A human scheduler always sacrifices one for another. The core of APS is an optimizer that finds a workable balance between conflicting goals.

Why many APS projects fail at go-live

Because people stare at the algorithm and ignore the real weak point, which is data. APS results are roughly algorithm × data quality. If the routing, standard times, capacity calendar or BOM is wrong in even one place, the output is a precise error. So implementation goes in three steps: fix the data first, then get a finite-capacity schedule that can be executed, and only then optimize. The three most common traps are not in the algorithm either: treating APS as a toy without fixing data, chasing the optimal at the expense of what is executable, and failing to change the organization so the planner's role never shifts.

Three questions before you start

APS is not an algorithm you buy. It is a systems project covering data, integration and organization. Before starting, answer three questions honestly:

1. Are your routings and standard times accurate?
2. Can ERP and MES exchange data in both directions?
3. Are your planners ready to move from dispatching jobs to managing rules?

If you cannot pass these three, even an expensive system ends up as decoration.

Translated from the original Chinese article on the WeChat account “犀牛鸟智造”.