Contents
Key points
- In production planning, an APS mainly improves on-time delivery, lead time, work in progress, setup share, OEE and planning effort.
- On-time delivery rises because commitments are made against real capacity and confirmed material dates – not because anything is produced faster.
- Lead time and work in progress fall because queue times, flat buffers and premature production disappear.
- Setup share, OEE and planning effort improve through better sequences and less manual replanning; almost half of the planning effort occurs outside the planning team.
- Not everything improves at once: utilisation often falls, and the logistics objectives conflict with one another. An APS makes this conflict manageable but does not resolve it.
All figures in this article are model calculations on a fictional example plant, not project results.
Why can the effect of an APS not be stated as a flat percentage?
Production planning metrics do not improve through an APS by fixed standard values. The effect depends on which constraints exist in the plant, how good the master data is and which processes are changed at the same time. Percentages without this information are therefore hardly reliable.
On top of this comes the attribution problem: a planning system is never introduced on its own. Master data is cleaned up and feedback processes and responsibilities change in parallel. Which share of the improvement is due to which change cannot be separated cleanly afterwards – not least because the three factors act multiplicatively:
- Good data without a planning system gives a better starting point that nobody uses – no person calculates a thousand operations against capacity, material and qualification every day.
- A planning system without good data produces plausible-looking, wrong plans. That is worse than no system at all, because somebody believes them.
- Both without changed processes have no effect if the plan is overwritten by hand after it has been issued.
More meaningful than a percentage is therefore the question of which mechanism a metric can respond through in a specific plant at all. If the mechanism does not apply, even the best software achieves nothing.
The model plant used in this article
All calculations refer to the same fictional plant: twelve lines, one of them a bottleneck running two shifts with 80 available hours per week and 46 production weeks a year. 1,200 production orders per year with an average of six operations. Setup and cleaning share at the bottleneck 18 per cent, on-time delivery 82 per cent, lead time 18 working days of which 2.2 days are actual processing. Work in progress 1.4 million euros, finished goods inventory 2.1 million euros, inventory holding cost rate 18 per cent, planning team 1.5 full-time equivalents.
How does an APS affect production planning metrics?
Production planning metrics do not change by chance. Every improvement results from a specific planning decision that acts through a mechanism you can check. Four chains of effect are the most important:
- Optimise the sequence → fewer setups and cleans → setup share falls, OEE rises
- Schedule against real capacity → achievable commitments → on-time delivery rises
- Move predecessors along → no premature production → work in progress falls
- Replace buffers with scheduling → less queue and waiting time → lead time falls
Why does on-time delivery rise – and what do OTD, OTIF and delivery reliability measure?
On-time delivery is one of the most important metrics in production planning. An APS improves it because delivery dates are planned against real capacity, confirmed material dates and the current order situation – not because anything is produced faster. The date determined this way is often later, and that is exactly why it is kept.
Without detailed scheduling, a delivery date is built from assumptions: standard lead times per production stage, flat replenishment times from the material master and safety buffers that were added at some point. None of these knows the current order situation – the date is an estimate that turns out right or wrong by chance.
OTD, OTIF, delivery reliability and schedule adherence: what do they measure?
The terms are often used interchangeably but measure against different reference points: schedule adherence against the internal plan, OTD and OTIF against the confirmed date, and requested-date adherence against the customer’s original request. For assessing an APS project the distinction is crucial, because the metrics develop differently.
| Term | What is measured | Reference point |
|---|---|---|
| Schedule adherence (schedule adherence) | Was the operation produced as planned? | Internal plan |
| Delivery date adherence, OTD (on time delivery) | Share of orders delivered on time | Confirmed date |
| OTIF (on time in full), delivery reliability | On time and complete in quantity – stricter than OTD | Confirmed date |
| Requested date adherence | Share of orders delivered on the originally requested date | Customer requested date |
| Service level, Service level | Umbrella term for logistics performance, depending on the definition including availability and flexibility | Varies |
| Service rate, Fill rate | Share of requirements that can be met immediately from stock | Inventory |
When people talk about the benefit of detailed scheduling, they almost always mean OTD or OTIF against the confirmed date – and it is precisely this metric that responds most clearly.
Does on-time delivery get worse at first after go-live?
Only in the planning view. Because the system calculates honestly for the first time, unachievable legacy commitments become visible and the reported delay rises in the short term. Actual delivery reliability towards the customer does not get worse as a result: the old date was not being kept anyway; the new one is later but achievable. In the medium term even requested-date adherence rises, because less equipment time is lost to avoidable setups and the plan becomes more stable – more orders can be produced within the requested period at all.
Model calculation
With 1,200 orders a year and on-time delivery of 82 per cent, 216 orders are late. If on-time delivery rises to 93 per cent, 84 remain – 132 late deliveries avoided per year. At an average of two hours per escalation, spread across sales, planning and production control, that corresponds to around 264 hours of pure communication rework.
The prerequisite is data quality: standard lead times and planned times from costing lead to wrong dates in an APS too. Which data objects need to be right for this is covered in the article on data objects for production planning.
Why does the lead time fall without anyone working faster?
Lead time is the time from the release of a production order to its completion and a central production planning metric. An APS usually reduces it not through faster processing but through shorter queue, waiting and buffer times.
In most plants, processing and setup account for less than a fifth of the lead time; the rest is queuing, waiting and safety buffers in routings and lead times. Detailed scheduling replaces these flat buffers with calculation: an operation starts when material, resource and staff are available, not after a fixed transition time has elapsed.
Model calculation
In the model plant the lead time falls from 18 to 9.5 working days, with processing unchanged at 2.2 days. According to Little’s Law, at constant output work in progress falls in the same proportion – by almost half, from 1.4 to around 0.74 million euros. At an 18 per cent holding cost rate that means around 119,000 euros lower inventory cost per year.
Why does work in progress fall?
Work in progress (WIP) is the value of all released but not yet completed orders in production. An APS reduces it because it takes delays into account not only forwards but also backwards through the order network.
Work in progress builds up wherever something is finished before it is needed. A typical trigger: a purchased part arrives later and the assembly date is postponed – but the upstream operations carry on unchanged and produce into a buffer. An APS moves these operations along. Nothing is produced prematurely, and the freed capacity is available to other orders. What this looks like in practice is shown in the article on production planning in machinery and plant engineering.
The second lever is lot size: campaigns oriented to demand and shelf life rather than a fixed target create less advance production. And because, according to Little’s Law, work in progress equals throughput times lead time, both metrics fall together.
How does an APS improve setup share and OEE?
Setup share and OEE are among the most important production metrics wherever sequence effects matter. The setup share is the proportion of equipment time spent on setups and cleaning; OEE (overall equipment effectiveness) is the product of availability, performance and quality. An APS improves both mainly through availability.
Where setup or cleaning effort depends on the sequence, the sequence directly determines available capacity. Detailed scheduling that knows transition matrices and forms campaigns reduces the number of costly changeovers with the equipment unchanged. On top of this come fewer material-related stoppages, because availability is checked before the start, and less start-up scrap thanks to calmer sequences.
Model calculation
The setup and cleaning share at the bottleneck falls from 18 to 12 per cent of equipment time. At 80 hours of weekly capacity, that is 4.8 additional productive hours per week, or around 221 hours a year – almost three extra production weeks at the bottleneck, without investing in equipment.
The mechanisms behind this are covered in detail under campaign formation and cleaning optimisation and, for electronics manufacturing, under SMD line planning and setup optimisation.
How far does planning effort fall – and where exactly?
Planning effort is a frequently underestimated production planning metric, because only part of it is visible in the planning team. It also arises in coordination meetings, in workforce scheduling and in daily replanning on the shop floor – and that part is often the largest.
Where planning effort actually occurs
| Pot of effort | Why it arises | What reduces it |
|---|---|---|
| Planning team | The plan is built by hand rather than assessed: setting sequences, hunting for conflicts, recalculating dates, maintaining lists | The plan is generated; the planner checks, decides and handles exceptions |
| Coordination and status meetings | Every department brings its own version. A large part of the meeting time goes into agreeing on the facts before anything is even decided | A shared data basis. The meeting starts with the decision rather than the reconciliation |
| Workforce scheduling | Shift roster and production plan are created separately and merged manually | Skills and attendance data become a constraint in the planning run instead of rework |
| Shadow planning on the shop floor | Supervisors and team leaders re-plan because the instruction is not workable, not current or not complete. This effort appears in no statistic and is rarely called planning | A plan that knows the actual constraints and is therefore workable without rework |
The fourth pot is the least visible and often the largest. Where production control spends an hour every morning translating the targets into a workable sequence, that is planning – it is just not called that and is measured nowhere.
Model calculation
64 hours saved per week correspond, over 46 production weeks, to around 2,944 hours a year – roughly 1.8 full-time equivalents. Of these, 34 hours fall to the planning team and the remaining 30 to coordination meetings, workforce scheduling and shadow planning. A supervisor who spends an hour less a day replanning is an hour more on the shop floor.
What the freed time is used for
The greatest effect often lies not in fewer planning hours but in freed capacity for master data maintenance, leadership and improvement work – in places where there has chronically been no time.
| Where time is freed up | What it is typically used for |
|---|---|
| Planning team | Computing scenarios instead of maintaining lists, bottleneck analyses, preparing capacity decisions, looking after the planning model |
| Production engineering | Master data maintenance – checking operation times against feedback, adding alternative resources, revising routings |
| Supervisors and production control | Leadership on the shop floor, disruption analysis, training new staff, implementing improvement measures |
| Process and lean managers | Setup time reduction using SMED, value stream analyses, layout improvements, standardising procedures |
| Production management | Medium-term capacity and skills planning instead of daily firefighting |
Two of these uses reinforce themselves. Better master data makes the plan better, less replanning is needed and time is freed again – the control loop is described in the article on data objects. And lean measures such as SMED shorten the individual setup time while detailed scheduling reduces the number of setups; the two levers multiply.
In fairness
these hours can rarely be translated into headcount reduction, and they should not be. Anyone who calculates the benefit purely as saved headcount will not realise it – and will lose acceptance along the way.
Calculated with 46 production weeks. The euro amount values the time; it is not a saving in the sense of headcount reduction. Experience shows that the pot for re-planning on the shop floor is estimated far too low on the first attempt – keeping a record for a week is worthwhile.
Which further production metrics respond?
Alongside the core production planning metrics, further production metrics often respond as well – such as finished goods inventory, backlog, response time to disruptions or the overtime rate. They complement the core metrics and follow the same mechanisms.
| Metric | Mechanism | Model calculation |
|---|---|---|
| Finished goods inventory | Lot sizes closer to demand and shorter advance production, because capacity and dates are known | From 2.1 to 1.7 million euros – around 72,000 euros lower inventory costs per year |
| Backlog | Overload becomes visible before it arises; orders are not released with impossible dates in the first place | Backlog hours at the bottleneck from 340 to under 100 per quarter |
| Response time to disruptions | Rescheduling in minutes rather than hours allows scenarios to be compared instead of a gut decision | From half a day to under 30 minutes to a reliable answer |
| Overtime and agency staff rate | Capacity bottlenecks become visible weeks ahead rather than on a Friday afternoon | Overtime rate from 7.5 to 5.0 per cent |
| Energy cost per unit | Better loading of chamber equipment, fewer heating cycles, peak load avoidance | Strongly dependent on the equipment – see energy-aware production planning |
| Skills coverage | Single-person dependencies become visible in the model and therefore manageable | Number of critical activities with only one available person from 9 to 4 |
Which metrics do not improve automatically – and why?
Not all production planning metrics improve automatically with an APS. This is not a limitation of the software but a consequence of central objectives such as utilisation, inventory, lead time and on-time delivery being in structural conflict. Knowing this before the project sets expectations correctly.
| Metric | What happens | Why that is right |
|---|---|---|
| Utilisation | Often falls, especially away from the bottleneck | High utilisation is a cost variable, not a performance variable. It can always be raised by releasing more orders – at the expense of inventory and lead time |
| Quoted lead time | Initially gets longer | The new date is realistic rather than optimistic; the old, shorter commitment was never achievable |
| Reported delay | Rises in the first weeks | Unachievable legacy commitments become visible before they are worked off – before, they were simply not measured honestly |
| Inventory ahead of the bottleneck | Can rise deliberately | A bottleneck waiting for material costs more than the buffer in front of it |
| Processing cost per unit | Stays unchanged | An APS does not change the work content, only its arrangement; at most the allocated setup and energy costs fall |
The conflict between the logistics objectives
High on-time delivery, short lead times, low inventory and high utilisation cannot all be maximised at once – described in the literature as the scheduling dilemma. An APS does not resolve this conflict, but it makes it transparent and consistently manageable.
The question “How much better will metric X get?” therefore cannot be answered without the objective weighting: a plant that puts on-time delivery above everything will see a different inventory development from one that prioritises utilisation. How the weighting can be set and applied consistently is described in the MCP weighting rule.
The most common conflict after go-live
the system shows later dates and lower utilisation figures than before and therefore initially looks like a deterioration. In fact it is showing the truth for the first time. Anyone who does not explain this effect in advance loses the argument in the first week.
How do you measure production planning metrics before and after APS implementation?
Production planning metrics can only be assessed reliably if the baseline, the comparison period and unaffected control variables are documented properly. A simple before-and-after comparison is usually not enough, because master data, processes and staff change at the same time.
- Record the baseline from dataOver at least six, preferably twelve months, so that order mix and seasonal effects become visible.
- Calculate against a historical reference weekApply the model to a week actually run and compare it with what really happened – only then do all other variables stay constant.
- Track control variablesSuch as scrap rate or sickness absence. If they move too, it was not the planning.
- Separate data effect from software effectImprovements that come from clarifying master data should not be attributed to the software. Both are valuable, but they are two different things.
- Observe for at least twelve monthsInventories and lead times only respond once the existing work in progress has been reduced.
Calculated with 46 production weeks. The inventory effect follows the relationship known as Little's law: at constant output, work in progress falls in proportion to the lead time. Not included are the value of the bottleneck time gained, lower planning and escalation effort, and the effects of higher on-time delivery.
Conclusion: what does this mean for an APS project?
An APS does not improve production planning metrics across the board but through mechanisms you can check: realistic scheduling, better sequences, predecessors moved along and calculation instead of flat buffers. Before the project it is therefore worth clarifying for each target metric whether its mechanism applies in your own plant, what the baseline looks like and how the objectives are weighted. Explaining in advance why utilisation and reported dates may initially look worse wins the argument after go-live instead of losing it.
For manufacturers
Where is the biggest lever in your planning?
Every production operation plans differently. Let’s look at your process together: where do bottlenecks occur today, which use case offers the greatest leverage, and what results can realistically be achieved?
Book a potential analysisFrequently asked questions on production planning metrics
Before the project: the deviation between planned and reported operation times, and the baseline of the target metrics such as OTD and lead time. After go-live, add the share of operations that are still moved by hand after the planning run. Time deviations and manual corrections say more about the quality of model and data than any result metric, and they can be derived from existing data without extra effort.
Data quality, constraint model and planning process have to fit together: realistic operation times and master data, a model that knows the constraints that actually limit the plant, and a process in which the plan is run rather than overwritten by hand. If one of these is missing, new plans are produced but not automatically better results. The objective weighting must also be set – otherwise it is unclear which metric is supposed to improve at all.
Planning effort falls immediately with go-live. Setup share and on-time delivery respond within a few weeks, once the sequences are actually run. Inventories and lead time take several months, because existing work in progress has to be reduced first. OEE follows last because it accumulates several effects; a reliable overall picture emerges after about twelve months.
No. Each effect depends on a mechanism that does not apply everywhere: setup share only falls with sequence-dependent setup or cleaning times, work in progress mainly in multi-stage production with order networks, and on-time delivery where dates have so far been derived from flat lead times. In single-stage production with few variants and stable lots the effects are considerably smaller. It is therefore worth checking, metric by metric, whether its mechanism exists in your own plant.
Because they make mechanism and order of magnitude transparent: every figure can be recalculated and transferred to your own values. Project results from other plants are hardly transferable, because their baseline, type of manufacturing and parallel process changes are unknown. What is achievable in a plant can only be determined from its own data anyway – the calculators on this page are a first step.
Sources and notes
All figures in this article are model calculations on the fictional plant described at the outset. They show the direction and order of magnitude of the mechanism concerned, not the result of a specific project. The theoretical foundations on the logistics objectives come from the literature below.
- H. Lödding: Verfahren der Fertigungssteuerung – Grundlagen, Beschreibung, Konfiguration (Handbook of Manufacturing Control). Springer Vieweg, VDI series. Chapters on the logistics objectives and their interactions.
- H.-P. Wiendahl: Fertigungsregelung – Logistische Beherrschung von Fertigungsabläufen auf Basis des Trichtermodells (Load-Oriented Manufacturing Control). Carl Hanser Verlag, 1997. The basis for the throughput diagram and the distinction between logistics performance and logistics costs.
- P. Nyhuis, H.-P. Wiendahl: Logistische Kennlinien (Fundamentals of Production Logistics). Springer. Includes the derivation of the relationship between inventory, output and lead time.
- ISO 22400 – key performance indicators for manufacturing operations management, including the definitions of availability, performance, quality and OEE.
An overview of selection and implementation is given in the guide Production planning software, and an overview of all technical articles under End-to-end planning.
Daniel Walkiewicz
Daniel Walkiewicz ist Partner bei MCP Algorithm Factory, verantwortet dort die Produktionsfeinplanung und APS-Prozessberatung und setzt seit 2010 Planungsprojekte bei MCP um. Das Unternehmen hat als Ganzes über 300 Planungsprojekte in Betrieb genommen und betreibt eines der größten Opcenter-APS-Teams in Europa.