Production Planning vs Reality: Why APS Plans Fail
DBR77 Team7 min read
On this page
APS plans fail because they are built on assumptions that stop being true soon after release: standard cycle times instead of real ones, machines assumed available, material assumed on hand and a full crew on every shift. A plan holds when it is fed by live machine states, actual cycle times, material status and staffing, and when planners have clear rules for when to replan and when to let the floor absorb a deviation.
The problem usually sits in the data the planning software works with.
Why the plan drifts after release
A plan forecasts the shift from what is known at release time.
The OPC Foundation's summary of ISA-95 lists detailed scheduling, work dispatching and work or product tracking among the same Level 3 activities. In many plants, though, they live in different tools. The APS sits next to ERP and reads routings and capacities that are updated once a year. The MES knows that line 2 stopped at 07:14. When those two do not talk, the plan and the floor go separate ways from the first hour, and the shift lead fills the gap with phone calls.
The plan uses the routing's standard cycle time, but the older press runs slower.
Five assumptions that break plans
Each broken assumption has a data source that can fix it, and a person who should own it.
| Assumption in the plan | What the floor sees | Data that fixes it | Owner |
|---|---|---|---|
| Standard cycle time from the routing | The older press runs slower than the routing says | Actual cycle times per machine and part, from the MES | Process engineer |
| Machine available all shift | Unplanned stops, and maintenance that was never in the plan | Live machine states, maintenance windows from the CMMS | Maintenance planner |
| Material on hand | Lot on quality hold, in the wrong place or delivered late | Warehouse stock by status, inbound deliveries | Warehouse lead |
| Full, qualified crew | Absence, or no one trained for the second machine | Shift roster and skills matrix | Shift lead |
| Fixed changeover time | Changeover depends on which product came before | Changeover times by sequence, from MES history | Planner |
Review the first and last rows monthly. Routings and changeover times drift slowly, so nobody notices until the plan is wrong every day.
When to replan and when to absorb
Replanning after every deviation is as bad as never replanning, because the floor stops trusting a plan that changes every hour. Write down the rules:
- Frozen window. The current shift, or the next four hours, changes only for safety, quality, a stop on the constraint or missing material.
- Replan triggers. The constraint machine is expected to stop for more than 60 minutes. Material for an order in the frozen window is not available. A rush order is approved by sales and planning together.
- Absorb on the floor. Short delays on machines that are not the constraint. The shift lead may swap orders on one line within agreed limits.
- Every change gets a version and a reason, so the next review can see why the plan moved.
A level sequence is easier to keep and easier to repair. The Lean Enterprise Institute describes heijunka as leveling the type and quantity of production over a fixed period of time, which avoids batching. A small, repeating mix recovers from a stop faster than one large batch that blocks a machine for two days.
Plan around the constraint first
The Theory of Constraints gives a simple order. Its five focusing steps start with identifying the system's constraint, exploiting it and subordinating everything else to it.
In planning terms: schedule the bottleneck first. Protect it with a small buffer of material and work ahead of it. Put maintenance windows and changeovers where they cost it least. Then schedule the other machines to feed it. A disruption away from the constraint often needs no replan at all.
Illustrative example: a plant runs 14 CNC machines on two shifts. The 5-axis cell is the constraint.
At 08:30, a coolant pump fails on a 3-axis machine. It is not the constraint, so the rule says absorb. The shift lead moves two short orders to the twin machine, and the plan stays as it is.
At 11:10, the 5-axis cell stops, and the technician expects three hours of repair. The rule says replan. The planner builds two scenarios: run the delayed orders on the next shift with overtime, or move one order to an older 5-axis machine with a longer cycle. The second keeps both customer dates. The planner publishes version 3 with the reason, the warehouse gets new picks in the new sequence, and customer service sees one order move by half a day.
How to measure plan quality
Measure the plan as well as the output:
- Schedule adherence per shift: orders started and finished within their window.
- Plan versions per day, with the reason for each.
- Deviations by cause: machine, material, people, cycle time, changeover.
- Cycle time gap between the routing and actual runs, per part.
- Frozen window breaks per week.
Material is often the largest cause, and Warehouse and Intralogistics Optimization With Real-Time Data shows how to connect the warehouse to the line. If the APS, MES and WMS come from different vendors, see How to Unify MES, WMS, QMS and CMMS Without Replacing Everything.
How this works in DBR77 IRIS
IRIS APS in DBR77 IRIS plans with finite capacity for machines and labor, including changeover times by product sequence. Planners define setup times, capacities and shift patterns, and material shortages from MRP are shown next to the plan. What-if scenarios run side by side, so lead times, costs and utilization can be compared before a plan is committed. Every plan is versioned, and approval workflows make sure the right people sign off before release.
Order progress from IRIS MES feeds back into the schedule, and deviations from plan are flagged for the planner. IRIS MRP shows which orders lack material, so the planner can move them before they reach the line. Machine status from DBR77 IIoT shows what is actually running, and scenarios can be tested on DBR77 DT before they reach the floor.
IRIS also tracks schedule adherence and flags bottleneck resources on the schedule. The planner decides which scenario becomes the plan, and that decision is logged with the plan version.
FAQ
Should the APS reschedule automatically?
Inside clear limits, a proposal can be prepared automatically. Releasing a new plan to the floor should stay with the planner, especially inside the frozen window.
How often should routings be updated?
Compare actual and standard cycle times monthly, and update the parts with the largest gap. A routing that is never checked drifts far from the floor within a year.
Do we need an APS if we have ERP planning?
ERP planning is often built on MRP logic that does not check machine and crew limits in detail. If your plan regularly fails on machine or crew limits, you need finite scheduling, from an APS or an equivalent module.
Conclusion
Take last week's plan and find which assumption broke first: cycle time, machine availability, material, crew or changeover. Feed that one from live data, agree a frozen window and replan triggers with the floor, and plan the constraint first. Then track schedule adherence and deviation causes every week to see which assumption to fix next.
Sources
- OPC Foundation, OPC 10030 ISA-95 Common Object Model: ISA-95 Summary
- Lean Enterprise Institute, Heijunka
- Theory of Constraints Institute, The Five Focusing Steps (POOGI)