Skip to content

Odoo Production Scheduling and Capacity Planning Guide

Stop Missing Delivery Dates Production Scheduling in Odoo

The Date That Was Never Going to Happen

A customer asks when they can have it. The salesperson asks production. Production looks at the board, forms an impression, and says four weeks. The order is confirmed for four weeks.

Nothing in that exchange involved capacity. Nobody asked what else is already committed in those four weeks, which work centres this job needs, or whether the one machine it must pass through is already full. The date came from an impression, and the impression was formed by someone who had not seen the three orders confirmed that morning.

Most plants that miss promise dates do not miss them because they are slow. They miss them because the date was never derived from anything. The capacity was already gone when the promise was made, and the only thing that changed later was that everyone found out.

This article covers the difference between forecasting and scheduling, the four inputs a schedule actually needs, the finite-versus-infinite capacity decision, how to find the constraint that governs your plant, and what makes a plan survive a week of reality. It is written for planners, operations directors and the sales leaders who make the promises.

Forecasting and Scheduling Are Not the Same Thing

Worth separating immediately, because the two get conflated and they answer different questions.

Forecasting estimates what will be wanted. It looks outward at demand (history, seasonality, pipeline, market signals) and produces a view of future requirement. It is a prediction, and it is wrong to some degree by definition.

Scheduling commits the capacity to make it. It looks inward at what the plant can do (work centres, hours, materials, existing commitments) and produces a plan for when specific work happens. It is a decision, not a prediction.

One feeds the other. A forecast tells you roughly what to prepare for, which drives long-lead purchasing and capacity decisions. A schedule tells you when a specific order will run, which drives the date you give the customer.

Why the distinction matters practically: businesses with a demand forecasting problem and businesses with a scheduling problem need different work. If your issue is that you are surprised by what customers want, that is forecasting. If your issue is that you know what they want and still miss the date, that is scheduling, and no amount of forecasting improvement will fix it.

This article is about the second.

The Business Problem: Dates Set Without Capacity

  • Promises are made from impression, not from load. The person asked has a feel for the plant. They do not have a view of committed hours by work centre.
  • Routing times are unvalidated. Standards set at implementation and never compared to what operations actually take, which means even a capacity calculation would be built on wrong numbers.
  • The constraint is assumed rather than identified. Everyone has an opinion about which machine is the bottleneck, and the opinion is frequently wrong, and in any case moves depending on product mix.
  • Everything is planned at once. No sequencing decision, so the plant works on whatever is most urgent that morning, which is the definition of expediting.
  • Material availability is not part of the date. A job scheduled for a week when a component arrives the week after.
  • Changes are absorbed invisibly. A rush order is inserted and nothing else is rescheduled, so three other jobs quietly become late and nobody tells their customers.
  • Nobody owns the plan. Planning is something that happens, rather than a role with authority to say no.

What Unreliable Dates Cost

  • Customer confidence, compounding. A customer told four weeks and delivered in seven stops believing your dates. Thereafter they add their own buffer, order earlier, or dual-source, all of which cost you.
  • Expediting as routine. Overtime, premium freight, machine changeovers out of sequence, and a supervisor spending their day re-prioritising. These costs are rarely attributed to the planning failure that caused them.
  • Work in progress inflation. When sequencing is driven by urgency rather than plan, more jobs are started than finished, and cash sits on the floor as partly built product.
  • Capacity decisions made badly. Without a load picture, the response to being late is often to buy more capacity, sometimes at a work centre that was not the constraint.
  • Sales constrained by caution. The opposite failure. A plant burned by missed dates starts quoting defensively long lead times, and loses work it could have taken.
  • Changeover efficiency lost. Sequencing by urgency rather than by similarity means more setups than necessary.

The Four Inputs a Schedule Needs

A schedule is only as good as these, and most plants are weak on at least two.

Routings That Reflect Reality

Which operations, in which order, at which work centres, taking how long. If the routing describes three operations and the plant runs five, every calculated date is wrong. Validating routings against measured times is the single highest-value preparatory task, and it is the same work that makes shop floor execution worthwhile, which is why the two projects belong together.

Work Centre Capacity, Honestly Stated

Hours available per day, accounting for shifts, breaks, planned maintenance and realistic utilisation. The common error is planning at theoretical capacity, which assumes no setup variance, no breakdown and no absence. A plant planned at full theoretical capacity is a plant that is always late.

Current Committed Load

What is already promised and not yet finished. Without this, a new order is scheduled into a plant that appears empty.

Material Availability

The date a job can start is the later of capacity being free and material being present. Scheduling against capacity alone produces dates that fail at the stores.

A fifth, which is not data: somebody with authority to decide sequence. A schedule nobody can enforce is a suggestion, and the plant will revert to urgency-driven working within a fortnight.

Infinite and Finite Capacity: The Decision Underneath

This is the choice that determines what kind of plan you get, and most businesses make it without realising they have.

Infinite capacity planning assumes work centres can absorb whatever is scheduled. It calculates when work should happen based on lead times and dependencies, without checking whether the hours exist. It is simple, fast, and it produces a plan that is achievable only if you happen not to be overloaded.

Finite capacity planning respects the available hours. Work is scheduled into capacity that actually exists, and when a work centre is full, jobs move out rather than being piled on. It produces realistic dates and it requires accurate routings and capacity data to be worth anything.

The honest position for most mid-market plants: start by making the load visible rather than by attempting full finite scheduling. A simple view of committed hours per work centre per week, compared against available hours, answers most of the practical questions (can we take this order, what is overloaded, where should the overtime go) without requiring the data precision that full finite scheduling demands.

Plants that attempt sophisticated finite scheduling on unvalidated routings produce precise wrong answers, which are worse than approximate right ones because people believe them.

Capability in Odoo varies by version and edition, and the planning functionality available to you in Odoo’s Manufacturing application should be confirmed with your partner before designing an approach around it.

Finding the Constraint

Every plant has one work centre that governs its output. Identifying it correctly changes how you plan everything else.

  • It is not necessarily the slowest machine. It is the one where work queues. Those are usually the same but not always: a fast machine fed by a slow one may be the constraint if the slow one is actually fine.
  • It moves with product mix. The constraint for one family of products may not be the constraint for another. Plants with a varied mix have a constraint that shifts, which is itself useful to know.
  • Find it by looking at queues, not by asking. Where does work physically wait? That question answers it faster and more accurately than any discussion, and the answer frequently surprises the people who were asked.
  • Then plan around it. Protect it with a buffer of work so it never starves. Schedule it carefully and let non-constraint centres follow. Put quality checks before it rather than after, so it never spends time on work that will be scrapped later. And resist improving non-constraint centres; improvement there produces no additional output and consumes effort that would pay back at the constraint.
  • Measure it. Utilisation, downtime and queue length at the constraint are worth more than the same metrics everywhere else combined.

Making the Plan Survive Contact With the Week

A plan that is remade every morning is not a plan. Three practices keep it stable enough to be useful.

  • Freeze the near term. Agree a horizon (commonly a few days to a week) inside which the schedule does not change except for genuine emergency. The freeze is what allows setup sequencing, material staging and labour planning to mean anything. Without it, every efficiency gain from planning is lost to churn.
  • Define how a rush order enters. It will happen. The question is whether it enters through a defined route (a named authoriser, a visible consequence, a decision about what moves) or informally, in which case three other jobs become late silently. Make the trade-off explicit: inserting this job delays those jobs, and somebody with authority accepts that.
  • Reschedule on a cadence, not continuously. Once a day or once a shift, at a set time, so everyone works from the same version. Continuous rescheduling means nobody trusts what they are looking at.
  • Report schedule adherence, not just on-time delivery. Jobs that started when planned and finished when planned. On-time delivery can be achieved by heroics; schedule adherence tells you whether the plan is any good.
  • Review the causes of lateness by operation. When a job is late, the useful question is where it lost time. That points at routing accuracy, a capacity assumption, a material failure or a sequencing decision: four different fixes.

Planning Approaches Compared

ApproachHow dates are setData requiredBest forMain failure mode
ImpressionSomeone’s feel for the plantNoneNothing beyond a few jobs a weekDates unrelated to capacity
Fixed lead timeStandard quoted lead time per productMinimalStable mix and steady volumeBreaks the moment demand varies
Load visibilityCommitted hours per work centre compared to availableRoutings and capacity, approximately rightMost mid-market plantsRequires someone to act on what it shows
Finite schedulingWork placed into capacity that existsRoutings and capacity, accuratelyPlants with validated data and a planning rolePrecise wrong answers on unvalidated routings
Constraint-basedConstraint scheduled carefully, rest followsRoutings plus a correctly identified constraintPlants with a clear, stable bottleneckConstraint identified by opinion rather than queue

Best Practices

  1. Separate your forecasting problem from your scheduling problem before choosing what to fix.
  2. Validate routings against measured times before building any plan on them.
  3. State capacity realistically, net of shifts, maintenance and realistic utilisation, never at theoretical maximum.
  4. Include material availability in the date, not just capacity.
  5. Start with load visibility rather than attempting full finite scheduling on unvalidated data.
  6. Identify the constraint by where work queues, not by asking.
  7. Protect the constraint with a work buffer and put quality checks before it.
  8. Freeze the near-term schedule and define a formal route for rush orders.
  9. Reschedule on a cadence, at a set time, so everyone sees the same plan.
  10. Report schedule adherence and the operation where lateness occurred, not just on-time delivery.

Common Mistakes

  • Promising dates without looking at load. The root cause of most missed dates, and it is a process failure rather than a planning one.
  • Planning at theoretical capacity. Assumes nothing ever goes wrong, guaranteeing the plan is optimistic from the first day.
  • Building a schedule on unvalidated routings. Produces precise answers that are confidently wrong.
  • Attempting finite scheduling before load visibility works. Sophistication ahead of data quality.
  • Identifying the constraint by discussion. Look at where work waits instead.
  • Improving non-constraint work centres. Effort that produces no additional output.
  • Inserting rush orders informally. Three jobs become late and nobody tells their customers.
  • Rescheduling continuously. Nobody trusts a plan that changes hourly.
  • Measuring only on-time delivery. It can be met by expediting, which hides a bad plan.
  • No planning authority. A schedule nobody can enforce decays into urgency-driven working within weeks.

An Illustrative Scenario: A Make-to-Order Fabricator

The following is a composite illustration drawn from patterns common to make-to-order manufacturing. It is not an account of a specific Techvaria client, and no figures below come from any real engagement.

Consider a fabricator producing to customer order, with cutting, forming, welding, finishing and assembly, quoting lead times on request.

The Trigger

On-time delivery had declined over two years without any single identifiable cause. Sales believed production was slow; production believed sales promised impossible dates. Both had evidence, and neither could settle it.

What the Review Found

Dates were given by the production supervisor from impression. There was no view of committed load by work centre. Routings described the process as it had been designed several years earlier, and two operations added since were not in them, so every job took longer than the system believed, consistently.

The assumed bottleneck was welding, which everyone named without hesitation. Looking at where work physically queued showed it was finishing, where a single booth handled every job and where work waited in a visible pile that had become so normal nobody remarked on it.

Rush orders were inserted by whoever took the call, with no record and no consequence assigned.

What Was Implemented

Routings first, corrected by walking the process: the same exercise that would be needed for any shop floor execution work, and the one that changed the capacity picture on its own.

Capacity was stated per work centre net of shifts, planned maintenance and a realistic utilisation assumption rather than theoretical hours. The resulting load view showed finishing committed well beyond its available hours for several weeks ahead, which explained the pattern nobody had been able to name.

Load visibility came before any attempt at finite scheduling, deliberately. A weekly view of committed hours against available hours per work centre was enough to answer the practical questions, and the planner used it to give dates rather than giving them from feel.

A rush order route was defined: a named authoriser, and an explicit statement of which jobs move as a consequence. The near-term schedule was frozen for the current week.

Finishing was protected: a buffer of work kept ahead of it, and the inspection step moved to before it rather than after, so the constraint never processed work that would later be scrapped.

After Two Quarters

Quoted lead times lengthened slightly and were met considerably more often, which the sales director initially resisted and later described as the better trade. The argument between sales and production ended, not because either had been right, but because the load view made the question factual.

The finding the operations director valued most was the constraint. Capital had been earmarked for a second welding station, which the queue analysis showed would have produced no additional output at all.

Industry Use Cases

  • Make-to-order fabrication and engineering. The core case: variable work content, promise dates on request, and a constraint that moves with mix. See manufacturing solutions.
  • Automotive components. Schedule stability matters more than elsewhere because customer delivery windows are contractual and narrow. See automobile and EV solutions.
  • Construction products and prefabrication. Production tied to site dates that cannot move, which makes the material-availability input as important as capacity. See construction.
  • Healthcare and medical products. Where validation and inspection steps consume constraint capacity and belong carefully in the routing. See healthcare solutions.
  • Trading and distribution with light assembly. Where a small production step governs despatch for a much larger order book. See trading and distribution.

Frequently Asked Questions

Forecasting estimates what will be wanted: it looks outward at demand and produces a prediction. Scheduling commits the capacity to make it: it looks inward at work centres, hours and existing commitments and produces a decision about when specific work runs. If you are surprised by what customers want, that is a forecasting problem. If you know what they want and still miss the date, that is scheduling, and better forecasting will not help.

Probably not as a first step. Start with load visibility: committed hours per work centre per week against available hours. That answers most practical questions without demanding the data accuracy full finite scheduling requires. Attempting finite scheduling on unvalidated routings produces precise answers that are confidently wrong, which is worse than an approximate view people treat with appropriate caution.

Look at where work physically queues, rather than asking people. The answers usually differ, and the queue is right. Bear in mind the constraint can move with product mix, so a plant with a varied mix may have more than one depending on what is running.

Most commonly because the routings do not match how work is actually done (operations added over the years and never reflected in the system) or because capacity is stated at theoretical maximum rather than net of shifts, maintenance and realistic utilisation. Both produce a plan that is optimistic from day one.

Define the route rather than preventing them. A named authoriser, and an explicit statement of which jobs move as a consequence, so the trade-off is accepted by someone with the authority to accept it. Informal insertion makes three other jobs late silently, and their customers find out before you do.

Schedule adherence (jobs that started and finished when planned) alongside on-time delivery, because on-time delivery can be achieved through expediting while the plan itself remains poor. Also record the operation at which a late job lost time, since that points to one of four different fixes.

Planning capability varies by Odoo version and edition. Confirm what is available in your instance with your partner before designing an approach around it, rather than assuming a particular behaviour.

Conclusion

The four-week date was not optimistic. It was arbitrary, because nothing in the conversation that produced it involved capacity.

Fixing that is mostly unglamorous. Correct the routings so they describe how work is actually done. State capacity net of the things that genuinely consume it rather than at theoretical maximum. Include material availability in the date. Make committed load visible per work centre, and give someone the authority to use it when answering the customer’s question.

Then find the constraint by looking at where work queues rather than by asking, protect it, and resist improving everything else. Freeze the near term so planning produces efficiency rather than churn, and define how a rush order enters so its cost is accepted rather than absorbed invisibly by three other customers.

And be clear about which problem you have. Forecasting predicts demand; scheduling commits capacity. If you know what the customer wants and still miss the date, no forecasting improvement will help you, which is worth establishing before anyone buys anything.

Find Out Why Your Dates Slip

Techvaria is an official Odoo Silver Partner with more than 350 implementations delivered since 2016, from offices in Bangalore, Ahmedabad and Dubai.

A production planning review walks your plant and produces four things: where your routings differ from how work is actually done; your real capacity per work centre net of shifts, maintenance and utilisation; your constraint identified from where work queues rather than from opinion; and a load view you can answer customer questions from.

It usually also settles an argument between sales and production that has been running for some time, by making it factual.

Talk to our Odoo consultants, or see how we approach manufacturing implementations.

Find Out Why Your Dates Slip

A production planning review walks your plant and shows where routings differ from how work is actually done, your real capacity per work centre, your true constraint and a load view you can quote dates from. Tell us which promise dates you missed last quarter, and we will tell you honestly where the capacity went.
Mustufa Rahi

Mustufa Rahi is an Odoo Certified Functional Consultant and ERP expert at Techvaria with 15+ years of experience in implementation, automation, and business process optimization, helping organizations scale efficiently.