Shop Floor Control (SFC): APICS Definition & Dispatch Rules

What shop floor control really is

Shop floor control (SFC) is the discipline of turning a planned production schedule into actual motion on the floor — which order starts next, on which machine, in which sequence, with which priority, and with what feedback flowing back upstream to confirm that reality matches the plan. The canonical definition, from the ASCM/APICS Dictionary (14th edition): "a system for utilizing data from the shop floor as well as data-processing files to maintain and communicate status information on shop orders and work centers." That definition is five decades old in substance — it pre-dates the term MES by about fifteen years — and the reason it has survived unchanged is that the underlying problem has not changed either. The plan produced in ERP on Friday afternoon meets the physical reality of Monday morning, and something has to decide, cycle by cycle, what gets dispatched next.

I have been doing this job, in different chairs, since 1998. Instandhaltung at Johnson Controls in Rastatt, Six Sigma Black Belt on headliner lines, PLC engineer in the JIT Center of Excellence, expatriate running a plant in Changchun, then team lead for global MES and traceability across Johnson Controls Electronics — 900+ connected machines, 750+ users, 30+ processes across China, Mexico, the US, Tunisia, Macedonia, France and Russia. After that Visteon, iTAC, Dürr, now SYMESTIC. The constant across every one of those chairs was shop floor control — the gap between what the ERP claimed was happening and what the plant was actually doing. The gap is always real, always larger than management believes, and always the root cause of the problems that get blamed on other things: missed deliveries, excess WIP, expedited freight, quality containment. Closing that gap is what SFC is for. Everything else in the manufacturing execution stack — OEE measurement, predictive quality, schedule adherence, on-time delivery — depends on whether SFC works. When it does not, those downstream metrics are fiction with decimal places.

The five-state order lifecycle — where plans become motion

The most common SFC failure mode I see is not dispatching-rule choice or terminal-technology choice — it is that the plant's systems treat "open order" as a single state when it is actually five. Most reporting problems, most schedule-adherence debates, and most of the "I thought that order was already running" surprises trace back to collapsing these five states into one or two.

The practical test for whether a plant's SFC actually works: at any moment during production, can you answer per order which of the five states it is in, without a human interpreting paperwork or guessing? Plants that can answer yes typically run at schedule-adherence rates in the 85–95% range. Plants that answer yes for states 1, 2 and 5 but collapse 3 and 4 into "somewhere on the floor" typically run 60–75%. Plants that have to send a supervisor to walk the floor to tell you where an order is have no SFC to speak of, regardless of what their ERP sales rep has told them they bought. I have audited all three kinds; the gap between them is measured in weeks of lead time and percentage points of on-time delivery, not in software licence fees.

Dispatching rules — the five that matter and when to use which

The core mechanic of SFC at the machine level is the dispatching rule: when a machine finishes one order and looks at the queue of released orders waiting for it, which one does it pick up next? Five rules dominate industrial practice, each with a specific operating regime where it outperforms the others. Plants that apply a single rule universally are almost always applying the wrong rule for at least some of their work centers.

The operational reality I observe on the floor: most plants are running informally on what I call the supervisor-shout rule — whatever the shift leader asks the operators to run next, based on the specific customer pressure of that hour. The supervisor-shout rule is not wrong in principle — a good supervisor incorporates information the formal rule cannot see, like a customer call from fifteen minutes ago or a container truck leaving at 16:30. The problem is that the supervisor-shout rule is invisible to the ERP, invisible to the schedule, and invisible to the reporting layer, which means the plant's formal OEE and schedule-adherence numbers are describing a dispatching regime that nobody actually follows. Fixing SFC starts with making the real dispatching rule explicit and visible, not with picking a theoretically optimal one. SPT looks great on paper and wrecks delivery performance for the long orders it starves. EDD looks great until you realise three orders have the same due date and the rule has nothing to break the tie. There is no universally correct rule; there is only the rule that fits your product mix, routing complexity, and cost-of-lateness asymmetry.

Input/output control — the SFC mechanism almost nobody uses

If the dispatching rule decides which order runs next, input/output control decides whether the queue at a work center is growing, stable, or shrinking. The concept was formalised by Oliver Wight in 1970 ("Input/Output Control: A Real Handle on Lead Time," Production and Inventory Management Quarterly, 3rd quarter 1970) and reinforced by George Plossl through the 1970s, and it is the single most undervalued mechanism in modern SFC practice. Most plants I audit have not heard of it, or have heard of it and never implemented it, and the consequence is a persistent failure mode that defeats every amount of dispatching-rule tuning: the queue at the bottleneck keeps growing because input is released faster than output can clear it, and no dispatching rule can fix a queue that is structurally increasing in size.

The mechanics are simple enough to fit on one whiteboard. For every critical work center, measure:

The insight that makes input/output control so powerful: if the queue is growing, no amount of dispatching tuning will shorten lead times, because Little's Law guarantees that average flow time through the work centre equals average WIP divided by average throughput. You cannot outrun the arithmetic. If you want lead time to come down, either input has to decrease (release fewer orders) or output has to increase (more capacity, fewer stoppages). Changing the dispatching rule only changes which orders are late, not how late the average order is.

This is the concept that transformed my thinking when I was running the Johnson Controls global MES program in the early 2010s, and it is the concept I now ask customers about first when they tell me their plant has a "lead time problem." The answer is almost always that nobody is measuring input and output at the bottleneck, that the queue has been growing for six to eighteen months, and that dispatching-rule experiments are not going to reverse the trend because the trend is structural. Input/output control is not glamorous, not new, not AI-driven; it is Wight-and-Plossl 1970s manufacturing management, and plants that do it reliably run lead times 30–50 % shorter than plants that do not, with the same machines, the same people, the same product mix.

Order release policies — the upstream choice that shapes everything

The release decision — moving an order from state 1 (Scheduled) to state 2 (Released) — is where the plant's SFC philosophy is encoded, whether the plant knows it or not. Four release policies dominate industrial practice, and they produce radically different shopfloor behaviours from the same underlying demand.

The practical point: the plant whose ERP is configured for backward infinite-capacity release (the default) and whose SFC layer runs EDD dispatching at the machine is not running a deliberate SFC strategy — it is running the ERP's unconscious defaults modulated by the supervisor's daily judgement. Making the release policy explicit is usually a higher-leverage intervention than any dispatching-rule change, because release is upstream of everything else that happens in the five-state lifecycle.

The Released-Not-Running pile — SFC's chronic blind spot

The single most diagnostic metric for SFC health is one that most plants do not track: the Released-Not-Running pile — the count of orders that the ERP/MES says are in state 2 (Released) but that no machine is currently in state 4 (Reported) on. These are orders that have been formally dispatched to the floor, that are consuming allocated materials, that the ERP thinks are being worked on, but that in reality are sitting somewhere — on a pallet outside the bottleneck, in a tote at the line entrance, on paperwork on a supervisor's desk waiting for material, on a machine that is down.

In a healthy SFC environment, the Released-Not-Running pile is small and stable — typically 5–15% of total released orders, representing normal queuing buffer. In a dysfunctional SFC environment, the pile grows month over month, reaching 40–60% of released orders, which is a diagnostic-grade signal that the plant's release policy is pushing more work onto the floor than the floor can absorb. The pile is also where most of the "lost" orders in a plant live — orders that everyone assumes are being worked on, that the customer service rep quoted a lead time against, and that will materialise as late surprises in three weeks. Tracking the Released-Not-Running count per work center, and alarming when it crosses a threshold, is a higher-value SFC feature than almost any dispatching-rule sophistication, because it surfaces the problem that dispatching cannot solve.

The Priority-Rule Fallacy — when the fix is upstream

The pattern I have watched play out in dozens of plants across four continents: management notices that on-time delivery is slipping, convenes a working group to optimise the dispatching rule, debates SPT versus EDD versus CR for six to twelve weeks, implements the winner, and sees on-time delivery improve by 2–4 percentage points for six months before gradually reverting to the baseline. The Priority-Rule Fallacy is the belief that dispatching-rule optimisation is the answer to lateness, when the actual cause is input/output imbalance at the bottleneck that no dispatching rule can fix. The rule determines which orders are late, not how many. Solving dispatching when the real problem is release gets a temporary improvement that masks the structural issue, which is the worst possible outcome because it defers the real intervention by years while consuming the organisational energy that would have gone into diagnosing the actual cause.

The diagnostic: if reducing the release rate at the upstream work center reduces lateness at the downstream one, the problem was release, not dispatching. If it does not, then dispatching is worth optimising. Doing this test before investing in dispatching-rule work is thirty minutes of analysis that saves twelve weeks of argument.

From a Johnson Controls JIT plant serving a premium German OEM, 2012: the plant produced sequenced seat-cover assemblies that had to arrive at the OEM's assembly line in the exact order the vehicles came down the paint shop, with a four-hour call-off window and a five-figure per-minute stop penalty if a sequence arrived late or out of order. On-time delivery had been stuck at 92.3 % for eight months against a contractual target of 99.5 %, and the operations team was in its fourteenth week of arguing about whether the dispatching rule at the sewing cells should be switched from a hybrid EDD/S-O configuration to pure S/O with a tighter tie-breaking rule. The mathematical modelling showed S/O would improve on-time by approximately 1.8 percentage points under steady state, which would still leave the plant 5+ points short of target but would be directional progress. I was brought in as the MES team lead to implement the sequencing changes, and I made the unpopular decision to delay the software change by a week and measure input/output at the three work centers that fed the sewing cells. What the measurements showed, within four shifts of data, was that the cutting cell upstream of sewing had been running at 108 % of its planned input against 96 % of its planned output for roughly fourteen weeks, which meant the queue of cut panels waiting to enter sewing had been growing at a rate of approximately 12 panels per shift for ninety-eight shifts. The queue had reached a size where the average panel waited 5.4 hours to enter sewing, against a plant assumption of 1.8 hours that had been baked into the sequencing logic six years earlier by someone who was no longer at the plant. No dispatching rule at the sewing cell was ever going to fix an upstream queue that the sequencing logic did not know existed. The actual intervention was a release-rate cap at the cutting cell — formal input/output control, Wight 1970, implemented as a CONWIP policy between cutting and sewing with a fixed cap of 35 panels in the intermediate buffer. Within three weeks of enforcement, the buffer stabilised at 34.1 panels, average wait time dropped from 5.4 to 1.9 hours, and on-time delivery reached 98.7 % and then 99.2 % and then 99.6 % by the end of the second month. The dispatching-rule debate resumed afterward and the team eventually did switch from hybrid EDD/S-O to pure S/O with tighter tie-breaking, which delivered an additional 0.3 percentage points on top of the input/output fix. The instructive number, that I quote whenever a new customer is convinced their lateness problem is a dispatching problem: 99.2 of the 7.3 percentage-point improvement came from input/output control, 0.3 from the dispatching rule the team had been arguing about for fourteen weeks. The Priority-Rule Fallacy in one plant, one chart, one contract-compliant recovery. Not because the dispatching rule did not matter — it did, at the margin — but because the margin was a rounding error against the structural issue nobody had looked for. I now ask the input/output question first on every SFC engagement, before I let any conversation about dispatching rules start. The plants that understand this reach world-class SFC performance with surprising speed; the plants that do not spend years in dispatching-rule debates that never quite solve the underlying problem.

SFC vs MES vs MOM vs APS — four disciplines glossaries routinely confuse

The four acronyms are not interchangeable. Confusing them is not pedantic; it leads to procurement decisions where plants buy capabilities they already have and fail to buy capabilities they need. The four-way disambiguation:

The practical implication: when a plant says "we need an MES," they usually mean "we need SFC that is better than what ERP gives us." When they say "we need APS," they mean "we need a scheduling engine that respects finite capacity." When they say "we need MOM," they mean "we need a reference architecture before we pick software." All three are legitimate needs; treating them as the same need is the root of 60 % of the failed manufacturing-software projects I have investigated in my career.

The 80/15/5 cause distribution of late orders

After enough SFC audits across enough industries, the same causal distribution keeps showing up for late orders, in roughly consistent proportions — consistent enough that I now use it as a predictive framework when a customer describes their lateness problem before I have seen any data. Approximately 80 % of late orders have their root cause upstream of the dispatching layer: insufficient capacity, input/output imbalance, missing materials, excess release. Approximately 15 % have their cause in dispatching itself: wrong rule, missing tie-breaking logic, supervisor-override that contradicts the stated rule. Approximately 5 % are genuinely unpredictable disruptions — machine failure, quality containment, urgent customer change — that no SFC discipline could have prevented.

The investment implication, exactly analogous to the predictive-quality 80/15/5 pattern: if you have not yet solved the 80 % of lateness that lives upstream of dispatching, solving the 15 % that lives in dispatching is solving the wrong problem. Plants that sequence their SFC maturity correctly — release policy first, input/output control second, dispatching rule third — capture 95 % of the achievable improvement at roughly 30 % of the investment. Plants that start with dispatching and never look upstream typically deliver the temporary 2–4 percentage-point improvement I described earlier and then wonder why their gains do not compound.

How this fits into the SYMESTIC platform

SYMESTIC implements SFC on the five-state order lifecycle as a first-class data model — every order in the system has an explicit state in (Scheduled, Released, Dispatched, Reported, Closed), every transition is timestamped and auditable, and the Released-Not-Running count per work center is surfaced as a default dashboard metric because it is the single most diagnostic SFC signal a plant has. The dispatching layer supports all five standard rules (FCFS, SPT, EDD, CR, S/O) configurable per work center, with priority-override escalation paths that make the supervisor-shout rule explicit when it is applied rather than invisible. The release layer supports all four standard policies (backward-infinite from ERP default, forward-finite via APS integration, CONWIP, kanban/POLCA for pull environments) with input/output monitoring as a standard instrumentation, because input/output control is the Wight-1970 mechanism that most plants have never implemented and that delivers more SFC benefit than any other single intervention. ERP integration is bidirectional with the major platforms (SAP R/3 via ABAP IDoc, Microsoft Dynamics/Navision, Infor/InforCOM, proAlpha) — orders flow from ERP in state 1, transition through states 2–4 in SFC, and return to ERP in state 5 with full quantity/time/quality reporting for costing and closure. The SFC data feeds directly into the OEE calculation (reported orders drive performance; dispatched-but-not-running orders drive availability analysis), schedule adherence (planned-vs-actual on dispatched sequence), on-time delivery (closed-order timestamp against customer promise), and rolled throughput yield (per-operation pass rate across the five-state transitions), because the whole point of treating the order lifecycle as a first-class model is that every downstream metric dependent on it is computed from the same auditable source of truth rather than reconstructed from conflicting fragments. See also alarm management for the event layer that surfaces SFC disruptions in real time, predictive quality for the analogous 80/15/5 framing on the quality side, and process documentation for the upstream definition model that SFC references when executing.

FAQ

What is shop floor control (SFC)?
Shop floor control is the discipline of turning a planned production schedule into actual motion on the floor — deciding which order starts next, on which machine, in which priority, and with what feedback flowing back upstream. The canonical ASCM/APICS definition: "a system for utilizing data from the shop floor as well as data-processing files to maintain and communicate status information on shop orders and work centers." SFC predates the term MES by about fifteen years and remains the core discipline that every manufacturing execution system implements, whatever the vendor calls it on the packaging.

What is the difference between SFC and MES?
SFC is the discipline; MES is the software category that implements it. An MES contains SFC as its core function plus additional capabilities — quality management, maintenance, traceability, performance tracking, documentation. A plant can do SFC with pencil and paper and travellers (and many still do, with surprising effectiveness); it cannot do MES that way because MES definitionally implies the software integration. The practical sales conversation that conflates the two is a root cause of procurement confusion: plants shopping for MES often need SFC discipline first, because the best software in the world cannot fix undisciplined order-lifecycle management.

What are the standard dispatching rules and when should I use each?
Five rules dominate practice: FCFS (First Come First Served) for fairness-critical or regulated environments; SPT (Shortest Processing Time) for clearing overloaded queues fast but at the cost of starving long orders; EDD (Earliest Due Date) for minimising maximum lateness — the standard in automotive and aerospace; CR (Critical Ratio, time-remaining ÷ work-remaining) for multi-operation routings where both time and work matter; S/O (Slack per Operation) for complex discrete assembly with variable operation counts. Universal rule choice is almost always wrong — different work centers in the same plant benefit from different rules, and plants that apply one rule globally are running the wrong rule for at least some of their production.

What is input/output control and why does it matter?
Input/output control, formalised by Oliver Wight in 1970, is the discipline of measuring and bounding the rate of order arrival at a work center against the rate of order completion. If arrival exceeds completion over time, the queue grows, lead time grows, and no dispatching rule can reverse the trend because Little's Law guarantees flow time equals WIP divided by throughput. The mechanism is not glamorous and not new, but plants that implement it reliably run lead times 30–50 % shorter than plants that do not, with the same machines and people. It is the single most undervalued mechanism in modern SFC practice and the one I ask about first on every customer engagement.

What is the Released-Not-Running pile?
The count of orders that the ERP/MES lists as released to the floor but that no machine is currently actively reporting on. Healthy SFC environments run 5–15 % of released orders in this state as normal queuing buffer; dysfunctional environments run 40–60 %. A growing Released-Not-Running pile is a diagnostic-grade signal that release policy is pushing more work onto the floor than the floor can absorb, and it is where most of a plant's "lost" orders live — orders the ERP thinks are being worked on, that customer service has quoted a lead time against, and that will materialise as late surprises three weeks from now. Tracking this count per work center, and alarming when it crosses a threshold, is a higher-value SFC feature than most dispatching-rule sophistication.

What is the difference between SFC, MES, MOM, and APS?
Four distinct concepts. SFC is the discipline of operational control; it is the parent concept, coined by APICS in the 1970s. MES is the software category that implements SFC plus additional manufacturing functions; defined by MESA International in 1997 and by VDI 5600. MOM (Manufacturing Operations Management) is the ISA-95 architectural umbrella framing manufacturing as four operational domains (production, quality, maintenance, inventory); it is a reference model, not a software product. APS (Advanced Planning & Scheduling) is the finite-capacity planning engine upstream of SFC that produces the schedule SFC executes. Confusing these four is the root of roughly 60 % of the failed manufacturing-software projects I have seen — plants buy what they already have and fail to buy what they need.

What is the Priority-Rule Fallacy?
The belief that optimising dispatching rules will fix lateness, when the actual cause is input/output imbalance or release-policy error upstream of dispatching. The pattern repeats reliably: management notices lateness, spends 6–12 weeks optimising the dispatching rule, gets a 2–4 percentage-point improvement that reverts to baseline within six months because the structural cause was never addressed. The diagnostic test is thirty minutes of analysis — if reducing release rate at the upstream work center reduces lateness downstream, the problem was release, not dispatching. Running this test before investing in dispatching work saves months of effort solving the wrong problem.

What is the realistic impact of implementing disciplined SFC?
For a plant starting from ERP-default backward-infinite release, no input/output control, informal supervisor-shout dispatching, and state-collapsed reporting, a disciplined first-year SFC programme typically delivers: on-time delivery improvement of 10–20 percentage points, WIP reduction of 25–40 %, lead-time reduction of 30–50 %, and schedule-adherence improvement of 15–30 percentage points. The dispatching-rule optimisation that plants typically focus on contributes 2–5 percentage points of the total; input/output control and release-policy discipline deliver the bulk. Compounded over three years with quality, maintenance, and OEE programmes running in parallel, the plant's operational performance is typically unrecognisable against baseline — not because any single intervention was transformative, but because SFC discipline makes every downstream improvement actually stick.

Related: MES: definition, functions & benefits · OEE: definition, calculation & practice · MES software compared · OEE software · Schedule adherence · On-Time Delivery (OTD) · Predictive quality · Alarm management · Process documentation · Rolled Throughput Yield · Scrap rate vs. rework rate · Change control · Recipe management · MDE (machine data acquisition) · BDE (production data acquisition) · Production control module · Production planning module · Production metrics module · Automotive · Metal processing · For production managers · For operational excellence · For COOs & plant managers. External references: ASCM/APICS Dictionary (canonical reference for Shop Floor Control, dispatching rules, and order-lifecycle terminology) · APICS body of knowledge (Wight's input/output control, Plossl's manufacturing-management tradition).

About the author

Christian Fieg

Head of Sales at SYMESTIC. More than 25 years in manufacturing industry on four continents. Six Sigma Black Belt. Career: Maintenance engineer at Johnson Controls Rastatt (1998), Six Sigma Black Belt in headliner manufacturing (three years), PLC engineer in JIT Center of Excellence (2003–2006), expatriate plant lead at FAWER Johnson Controls Changchun, China (2006), Team Leader Business Analyst Global Electronics at Johnson Controls (2006–2013, 900+ connected machines, 750+ users, 30+ processes across China, Mexico, USA, Tunisia, Macedonia, France, Russia), Manager Center of Excellence at Visteon Corporation (2013–2015, global MES program end-to-end), Sales Manager MES DACH at iTAC Software (2015–2018), Senior Sales Manager at Dürr AG (2018–2021), Head of Sales at SYMESTIC since 2021. Author of "OEE: Eine Zahl, viele Lügen" (2025). Expertise: Manufacturing Execution Systems, shop floor control and dispatching, input/output control, APICS/ASCM body of knowledge, global MES rollouts, OEE, Six Sigma (Black Belt), traceability, Smart Factory, JIT/JIS processes, automotive production, Cloud MES. · LinkedIn · Book: "OEE: Eine Zahl, viele Lügen"