Dillygence
Production flow management: what should be simulated or not?
Production flow management: intuition is no longer enough. Discover which decisions the COO can make without simulation and those where it is vital.
Piloting industrial production: from intuition to simulation-based testing
According to Gartner, simulation is becoming a standard for industrial governance in complex organizations, especially for high-impact CAPEX and OPEX decisions. The signal is simple: experience is no longer enough when a decision involves millions, customer lead times, and a fragile production ramp-up (production rate increase). For a COO (Chief Operating Officer, Director of Operations) or an Industrial Director, production steering changes in nature: it no longer exists only to “hit the plan,” but to prove that a plan is physically feasible—site by site, product mix by product mix. Takeaway: intuition guides, simulation decides when financial risk becomes real for production steering.
The shopfloor tension: when the plant must hit the plan, cash, and safety… with constant disruption
A COO does not manage an average; they manage days with breakdowns, scrap, absences, material delays, and changeovers. From an Operations leadership standpoint, the challenge is different: preventing these local disruptions from turning into multi-site drift, excess stock, and daily trade-offs between production, sales, and finance. Each disruption forces a choice: protect throughput, protect quality, protect delivery, or protect cash. Without rules, the shopfloor becomes a system of implicit priorities—therefore unstable. With rules, the same reality becomes manageable, and simulation helps test those rules before enforcing them in production steering.
1) Production steering, clearly: a decision loop “from plan to shopfloor”
Industrial production steering can be understood as a short loop: decide, execute, observe, correct, learn. For a COO (Chief Operating Officer, Director of Operations), this loop only has value if it scales: multiple plants, multiple product families, and constraints that move. Much of the best market content explains tools or planning, but rarely describes this full loop with explicit disruption management. When the loop breaks, firefighting replaces management, and production steering deteriorates.
Actionable definitions: planning and scheduling, execute, monitor, improve
Production planning sets a direction: volumes, capacity, assumptions, and horizon. Scheduling turns that direction into an executable sequence, station by station, with real constraints and priority rules. Execution produces reality, with its variability and deviations. Monitoring and improvement turn those deviations into decisions that reduce the dominant variability in production steering.
The useful difference fits in one sentence: planning answers “what and when,” scheduling answers “in what order and with which resources,” execution answers “what actually happened,” and monitoring answers “what must change today” for production steering.
Flow steering: throughput, WIP, and lead time in the same equation
Throughput describes what the system actually outputs. WIP stands for Work In Progress (work-in-process inventory), meaning cash and space tied up. Lead time (throughput time) measures how fast work crosses the system—therefore the customer promise and commercial flexibility. If you optimize only local throughput, WIP grows and lead time slips, even with rigorous production steering.
Serious steering links these three variables and accepts an explicit trade-off. For an Industrial Director, this trade-off must also be comparable from one plant to another: same definitions, same rules, same trade-offs. Simulation makes the link visible because it reproduces the non-linear effects of queues. You reduce lead time without producing faster when you reduce waiting, not only when you increase pace—and that is a direct lever of production steering.
2) A COO's three decision zones: intuition, data, simulation
A COO does not decide the same way depending on risk, time horizon, and reversibility. A reversible, low-cost, local decision can be handled quickly. A costly, irreversible decision requires proof. Between the two, an intermediate zone exists—and it explains why many groups burn out on trade-offs between plants, programs, and functions in their production steering.
The intuition territory: culture, talent, conflicts, supplier negotiation
Algorithms calculate, but they do not carry a culture of safety, rigor, and respect. An inter-team conflict, a drift in commitment, or a supplier negotiation rely on trust and weak signals that are not quantified. In these cases, the COO's intuition remains decisive. The digital twin does not replace diplomacy, including in production steering.
The intermediate zone: fast trade-offs supported by BI
The “Grey Zone” corresponds to moderate-risk decisions that require a fast response. BI (Business Intelligence, decision-support analytics) helps here because it consolidates data and displays usable trends. Choosing a non-critical carrier or moving a batch forward on an underloaded line often fall into this category. The risk comes from gradual drift: a sequence of “moderate” decisions can create a major deviation, and the grey zone becomes a wall for production steering.
The proof wall: when the model becomes the referee between production, sales, and finance
Three decisions fall into the “no-mistake” zone: committing CAPEX (capital expenditure), validating a ramp-up (production rate increase), and testing logistics robustness under crises. On these topics, one mistake costs cash, credibility, and sometimes market share. The model becomes the referee because it arbitrates on a shared physical reality and breaks silos between production, sales, and finance. Analyses from McKinsey on operational transformation performance confirm that execution rigor remains the differentiator in production steering.
3) Methods and routines that work in real life
A useful method must work in a noisy shopfloor with teams focused on throughput. For a Director of Operations, it must also be deployable: the same routine, the same decision, a measurable flow effect—even when plants do not share the same history. It links a routine to a decision, and a decision to a flow effect, to strengthen production steering.
Daily routines: SQCDP and QRQC as decision engines, not decorative boards
SQCDP stands for Safety, Quality, Cost, Delivery, People. QRQC stands for Quick Response Quality Control (rapid-response quality control). These routines work when they trigger dated actions with an owner and a deadline. They fail when they become a snapshot of results with no trade-off and no escalation, and production steering runs out of steam.
Priority management and shopfloor sequencing: simple rules, immediate impacts
Sequencing is used to decide “what to produce now” without constant debate. A priority rule must remain rare, stable, and understood; otherwise every emergency bypasses it. A common trap is accelerating the “most delayed” batch on a non-constrained station: you improve a local metric, then overload the constraint downstream, and overall lead time increases. Simulation enables comparing multiple sequencing rules and their impact on WIP and lead time, and equips production steering.
Real-time monitoring: what to measure, who to alert, and what reaction time to target
Real-time production monitoring exists to reduce reaction time, not to increase the amount of data. You must measure what changes a decision within the hour: stop, rate drift, quality at station, material shortage, queue saturation. A stop on the constraint must trigger action in under 10 minutes. A quality drift must trigger isolation and qualification in under 30 minutes; otherwise production steering becomes reporting.
4) Disruption management: the expected decision chain, minute by minute
One hour of downtime on a constrained line can represent several tens of thousands of euros in lost output in automotive or aerospace.
Managing production disruptions is not about heroics; it is about a short, understood procedure. For a multi-site Industrial Director, the issue is also to standardize this decision chain: same thresholds, same words, same reflexes—otherwise each plant reacts differently and comparison becomes impossible. A poorly qualified disruption often triggers the wrong decision, then cascading rescheduling, which degrades production steering.
Detection, qualification, arbitration, rescheduling, communication, back to standard
Detection identifies the event and its context: station, order, part number, time, and duration. Qualification decides severity: safety, quality, constraint, critical customer. Arbitration applies a known priority rule, then decides on sequence, assignment, and logistics. Rescheduling must remain minimal; otherwise it breaks production steering execution.
Communication announces the new plan, impacts, and expectations—without debate. Returning to standard closes the incident, then triggers a root action if the event repeats. Simulation can verify whether the arbitration rule truly reduces the loss on overall flow and quantify the value of a targeted buffer stock in production steering.
The recurring trap: accelerating locally and slowing globally
A site often accelerates a non-constrained station to “catch up,” then inflates WIP in front of the next constraint. This increases waiting time, therefore lead time, even if local pace increased. A countermeasure is to steer by the constraint's throughput and limit releases. Simulation quickly shows the effect of these rules because it reproduces queues and blocking—and strengthens production steering.
5) Minimum viable indicators: 8 to 12 KPIs that trigger action
A Director of Operations does not need a cathedral of indicators; they need a shared language across plants and functions. Each KPI must trigger an action; otherwise it remains an observation. The goal is to connect performance, flow, and working capital requirement in a visible way for production steering.
For each KPI: associated decision and typical drift to avoid
KPI | Triggered decision | Typical drift to avoid |
|---|---|---|
OEE (Overall Equipment Effectiveness, TRS) | Prioritize maintenance on the constraint, then decide a reliability plan | Optimizing OEE on a non-constrained station with no impact on overall throughput |
FPY (First Pass Yield, first-pass quality) | Stop, isolate, qualify, then correct the dominant cause | Accepting rework as “normal,” then losing capacity without seeing it |
WIP (Work In Progress, work-in-process) | Limit upstream releases, then protect the constraint | Increasing WIP “to never run out,” then exploding lead time |
Plan adherence | Escalate a gap, then decide minimal resequencing | Rebuilding the plan every day and breaking execution |
Schedule adherence | Reduce exceptions, then stabilize priority rules | Letting every emergency bypass the sequence and creating permanent chaos |
Lead time | Reduce waiting, treat the bottleneck, limit queues | Confusing lead time with cycle time and acting in the wrong place |
Rework | Decide a cause-reduction plan, then review controls | Fixing faster without reducing defects, then losing twice |
Stops / downtime | Classify, treat the dominant cause, adjust response | Multiplying categories and solving nothing |
Changeovers | Review lot policy and sequence, then reduce losses | Changing too often “to serve everyone,” then saturating the constraint |
Link performance and working capital: WIP, inventory, and tied-up cash
Working capital requirement increases when WIP and inventory rise faster than cash collection. Improving useful throughput on the constraint often reduces WIP, freeing cash and floor space. The Industrial Director can arbitrate between a targeted buffer stock and a global buffer, and justify a reliability investment through working capital impact, not only through OEE. Simulation makes this conversion robust because it tests disruption effects on WIP and lead time in production steering.
6) Tools and systems: who decides what, with which data
Comparison table: required data, supported decision, typical limit
Family | Required data | Supported decision | Typical limit |
|---|---|---|---|
ERP (Enterprise Resource Planning, integrated management software) | Orders, bills of material, inventory, lead times | Mid-term plan, purchasing, macro capacity | Low shopfloor granularity; lead times often fixed |
MES (Manufacturing Execution System, production execution system) | Routings, stations, times, scrap, traceability | Execution, quality, order tracking | Does not compute optimal scheduling on its own |
APS (Advanced Planning and Scheduling, advanced planning and scheduling) | Constraints, calendars, times, changeovers | Constraint-based scheduling, sequences | Highly dependent on parameter quality |
SCADA (Supervisory Control And Data Acquisition, supervision and data acquisition) | Machine states, sensors, alarms | Short-horizon supervision, technical reactions | Little product and logistics context |
Shopfloor tracking | Simple events, progress, alerts | Fast shopfloor decisions, escalations | Risk of drifting into reporting |
Flow simulation | Disruption distributions, rules, proven capacities | CAPEX tests, ramp-up tests, logistics stress tests | Requires disciplined modeling |
7) Flow simulation and digital twin: the 5-step method
Flow simulation is worthless if it remains a lab exercise. It must produce a decision, a rule, then an execution routine. For a COO, it is a way to decide without an “opinion contest” between functions, and to compare scenarios on a shared basis. The digital twin serves here as a governance support, not a showcase, for production steering.
Start from a measurable business objective: +20% capacity, -15% lead time, -10% conversion cost, or reduced carbon footprint. Translate the objective into flow variables: constraint, WIP, rate, changeovers, quality.
Set the model boundaries: labor constraints, safety, quality, certifications. This step defines what the model is allowed to change. You avoid a “optimal” but unusable scenario.
Build a minimal model: represent the constraint, upstream and downstream flows, and the main sources of variability. What-if scenarios (simulation “what if”) test rules, buffers, resources, and lot sizes.
Validate assumptions and run stress tests: the model must reproduce the current state within acceptable margins, then test frequent breakdowns, absenteeism, supplier delays, and demand peaks. A stress test often shows that a “profitable” scenario collapses as variability increases.
Translate results into arbitration rules and execution routines: sequencing rules, buffer levels, alert thresholds, and escalation rules. The shopfloor must understand what changes, and why. Then test implementation at small scale and expand.
8) Three mini-cases: what changes when you test before acting
Mini-cases: what changes when you test before acting
Mini-case | What | How | Impact (production steering) |
|---|---|---|---|
The client needed to transform a metro bogie maintenance plant dating from the 1930s into a “factory of the future.” The goal was to double production capacity while integrating next-generation technologies, after a first simulation by a third party had failed. |
Use of DispoX simulation to model flows, identify blockages, and test the effectiveness of a flexible operator pool rather than investing in new machines. . | Production targets exceeded, massive stock reduction, and adoption of the Dillygence method as the group standard for every new project. | |
2 — Stabilized production ramp-up (production rate increase) | A product launch requires a fast ramp-up (production rate increase), with risk of saturation and customer delays. | The digital twin tests learning-curve scenarios, OEE (Overall Equipment Effectiveness, TRS), scrap rate, and staffing, then identifies the first station to tip into congestion. | The site adjusts multi-skilling, buffers, and sequence, with visible stability gains in plan adherence and reduced rework. Without simulation, the site sometimes invests in the wrong place and simply moves the constraint. |
3 — Simulated logistics disruptions | A supply chain (supply chain) experiences delays, transport variability, and occasional shortages. | What-if scenarios (simulation “what if”) test supplier delays, stockouts, substitutions, and buffer-stock policies, then evaluate impact on OTD (On-Time Delivery, on-time delivery) and WIP. | The site reduces unnecessary inventory and increases targeted stock on critical references, improving service without inflating working capital everywhere. Purchasing, production, and finance share a common criticality logic. |
9) Reading grid: 5 traps and countermeasures
Indicators without action
Trap: you display KPIs, but no decision comes out, so the team disengages. Countermeasure: link each KPI to a standard action, a threshold, and an owner. Limit KPIs to those that trigger a decision within the week.
Real time without decision
Trap: you collect real-time data, but nobody knows what to do when an alert arrives. Countermeasure: define an escalation matrix with response times and levels, then test it on an incident. Reduce the number of alarms, increase their quality, and enforce a short review of “noise” alerts to eliminate them.
Frozen scheduling and constant exceptions
Trap: you announce a sequence, then break it ten times a day and lose all stability. Countermeasure: enforce a short freeze, for example 24 to 48 hours, and treat only major exceptions. Formalize a single priority rule, measure adherence, then reduce the causes of exceptions—often material and quality.
Bottleneck ignored or “nomadic” and not tracked
Trap: you steer everywhere, so you steer nowhere, and the constraint shifts without being tracked. Countermeasure: identify the weekly constraint, then protect it with a release rule and maintenance. If the constraint remains nomadic, segment by product family. Simulation helps explain why the constraint moves when mix changes.
Uncontrolled automation and disputed data quality
Trap: you automate an unstable flow, then you automate chaos, and teams dispute the numbers. Countermeasure: first stabilize decision rules and event definitions. Enforce minimal, useful data entry, then make value immediately visible to the shopfloor. Simulation becomes a filter: it helps choose where to invest and what to measure.
Production steering: simulate before investing, rescheduling, or moving the bottleneck
Production steering is not won with more meetings, but with a short loop that produces executable decisions. If your indicators do not lead to action, if your schedule changes ten times a day, or if you invest only to move the bottleneck, you are not steering—you are reacting. Flow simulation brings physics back into debates, shows congestion effects, and makes it possible to test arbitration rules before imposing them on the shopfloor. Takeaway: robust production steering turns uncertainty into scenarios—and scenarios into simple rules that are understood and followed.
At Dillygence, our teams combine industrial expertise and a digital twin to test your decisions (CAPEX, production ramp-up, flows) before they become expensive on the shopfloor.
FAQ — Production steering
What are the limits of intuition versus algorithms?
Intuition integrates human signals, social tensions, and negotiation elements that algorithms do not quantify well. However, intuition often gets congestion, WIP, and variability effects wrong because these phenomena are non-linear. The limit is therefore not the human—it is the lack of proof when the system becomes complex.
How do you accelerate a production ramp-up (production rate increase)?
To accelerate a ramp-up (production rate increase), you must identify the likely bottleneck before it saturates, then protect it with resources, reliability actions, and release rules. A simulation stress test makes it possible to test OEE, scrap, learning, and staffing scenarios. Acceleration mostly comes from stability, not heroics.
Will simulation replace the COO?
No, because the COO provides leadership, culture, talent management, and organizational diplomacy—and these dimensions are not modeled well. However, simulation changes accountability standards for heavy decisions: CAPEX, ramp-up, and logistics robustness. The role shifts toward more experimentation and fewer bets. A COO who commits CAPEX without simulation now takes an avoidable risk.
What is the new governance standard in 2026?
The standard becomes “test before decision” for high financial-risk choices, with scenario validation and stress tests. Simulation tools and the digital twin become a co-pilot because they align production, sales, and finance on a shared physical reality. Dillygence applies this expectation daily thanks to its digital twin, to test rate, capacity, and flow scenarios before prioritizing high-impact investments.
