A single variance figure at the end of a work order is a familiar output and a weak one. It confirms that expectation and reality diverged, and stops there. The plant manager cannot act on it, because it does not say whether material was wasted, whether the process took longer than it should, whether an outside operation cost more than quoted, or whether nothing happened on the floor at all and the expectation was simply set wrong.
Work-order costing becomes useful at the point where the variance can be decomposed. Everything below is in service of that.
What work-order costing is
Work-order costing establishes what a production job was expected to cost, accumulates what it actually consumed as it runs, and exposes the difference. The work order is a practical unit of comparison because it has both a defined production intention and a recorded outcome.
The two sides are built differently and fail differently. The expected side is a model, and its risk is that it stops reflecting how the item is really made. The actual side is a record, and its risk is that something real never gets captured. A costing system is only as credible as its weaker side.
The expected side: two legitimate bases
There is a persistent assumption that expected cost must be derived from the bill of materials and routing. It is one correct answer, not the only one.
- Derived from BOM and routing. Expected cost is computed from the structure and the operation sequence the item genuinely follows. Its strength is consistency: when the structure changes, the expectation moves with it. Its consequence is that the standard is not stable across a period.
- Separately maintained. An expected cost is held independently of the current structure. Its strength is stability, which is what financial reporting across a period usually wants. Its consequence is that it must be reviewed deliberately, or it silently ages.
Manufacturers commonly need both behaviours at once, for different product groups. A mature product with a settled structure and a product still undergoing engineering change do not want the same treatment. The question to ask of any system is therefore not which basis it uses, but whether it can support both without one of them being maintained in a spreadsheet.
The two bases carry different maintenance obligations. Where expected cost is derived from the BOM and routing, its reliability depends directly on the accuracy of that structure: if the material content is wrong, no amount of care on the actual side will produce an interpretable variance. Where it is maintained separately, it has to be reviewed and reconciled deliberately, so that it does not drift away from current manufacturing reality while appearing stable. Either way the expectation has to remain traceable to how the item is actually made, which is why material structure and multi-level BOM is a prerequisite subject rather than a parallel one.
The actual side: what has to be captured
Actual cost is assembled from components that arrive at different moments and from different sources. Any component that is not captured leaves the recorded product cost incomplete. How those events are recorded and made visible while the job is still running is covered in work-order execution and shop-floor tracking. Where the machine or process time itself is captured, and the expectation it can be read against, belongs to capacity, routing and process time.
| Component | What it represents | What its absence hides |
|---|---|---|
| Material | Relevant material consumption against the job | Yield loss, scrap and substitution at real prices |
| Machine or process time | Actual time on the machine or process step | Slow running, setup overruns, equipment condition |
| Labour or production time | Time recorded against production | Manning levels, skill mix, rework effort absorbed silently |
| Manufacturing overhead | Indirect cost applied on a defined basis | Products that consume very different indirect resources looking alike |
| Subcontract / job work | Outside processing on the item | The true cost of an outsourced step, and any make-versus-buy comparison |
| Rework | Additional cost caused by correcting output | Quality cost, which is otherwise absorbed into normal production |
Machine time and labour time are not interchangeable
They are frequently collapsed into one figure, and collapsing them loses information the variance later needs. An operation where a machine runs for six hours with brief attendance and an operation where a person works for six hours have entirely different cost structures and entirely different improvement paths. Capturing them separately is what allows an overrun to be attributed to the right one. Both actual machine or process time and production labour time can contribute to actual work-order cost, and treating them as distinct inputs is the point.
Overhead: the allocation question
Overhead is one of the areas where product costing can become misleading if the allocation basis does not reflect what actually drives the cost. The total is real; how it is spread across jobs may not be. A single plant-wide rate applied to every job can systematically understate some products and overstate others, and neither error is visible in the total.
Configurable approaches supported in exactllyERP include, as applicable:
- A fixed amount, where the indirect cost genuinely does not vary with the job
- A percentage, applied to a defined base
- A machine-hour rate, where equipment drives the indirect cost
- A labour-hour rate, where people drive it
- Operation-wise rates, where different steps have genuinely different indirect profiles
- Combinations of the above, where a single basis would misrepresent the operation
Having several bases available does not mean every manufacturer should use several. The useful test is whether the chosen basis moves with the thing that actually causes the cost. If it does, one basis is enough. If it does not, adding precision elsewhere will not repair it.
Rework: two treatments, two different questions
Rework can be handled in two ways, and exactllyERP supports both. They are not competing options so much as answers to different questions.
- Additional cost on the original work order. The batch carries its own correction cost, so the recorded cost of those units reflects what they genuinely cost. This is useful when the objective is to retain the full cost of that batch or job.
- A separate rework work order. Rework is recorded as an activity in its own right, countable and comparable across periods. This is useful when the objective is to make rework visible and comparable as its own activity.
The more damaging case is neither of these: rework absorbed into normal production time with no marker at all. The plant then appears to be running to expectation while quietly spending capacity on correction, and the cost surfaces only as a productivity question nobody can source. Rejection recorded at the stage where it occurs, with the rework treated deliberately, is what prevents that. How that rejection is captured operationally, and how a rework order stays linked to the job it came from, is covered in rejection and rework tracking.
Subcontract and job work in the work-order cost
When part of production happens outside the plant, the cost of that step belongs to the item it was performed on. Subcontract or job work cost can participate in actual work-order costing, which keeps the work order a complete picture rather than a picture of the in-house portion only.
This matters beyond accuracy for its own sake. Excluding subcontract or job-work cost from the work order understates the cost of the outsourced processing, and can distort make-versus-buy and make-versus-subcontract comparisons. Which way it distorts them depends on how the comparison is constructed, which is precisely why the underlying figure needs to be complete before the comparison is made. The operational side of that arrangement, covering how the material is tracked while it sits with the processor and how a return is reconciled, is covered in job work and subcontracting.
Reading variance by cost driver
Once the components above are captured separately, the comparison between expected and actual can be resolved into contributions rather than a single figure. Standard-versus-actual variance can be analysed across relevant cost drivers rather than only as one undifferentiated total, and that decomposition is what makes it usable.
| Where the variance sits | What it typically points to | Who can act on it |
|---|---|---|
| Material | Yield, scrap, substitution, or purchase price movement | Production and purchasing |
| Machine or process time | Setup, running speed, equipment condition, or an unrealistic expectation | Plant and maintenance |
| Labour or production time | Manning, skill mix, or effort absorbed by correction | Production supervision |
| Overhead | Often the allocation basis rather than any event on the floor | Finance |
| Subcontract / job work | Rate movement, or more outside processing than planned | Purchasing and planning |
| Rework | Quality at a specific stage | Quality and production |
The last column is the practical argument. A total variance has no natural owner, so it tends to be discussed rather than resolved. A variance attributed to a driver has an owner, and a next step. Where a material variance points to yield or scrap, the quantities underneath it are a separate question, covered in scrap, by-product and yield tracking.
A variance is a difference between two numbers, and either can be at fault. Before treating an overrun as a production problem, it is worth confirming that the expectation was set against the structure and process the job actually used. A large and persistent variance in one direction across many comparable jobs is a signal to revisit the standard as well as the production process, before assuming the cause sits on the shop floor.
What to evaluate in ERP work-order costing
These are the costing-specific questions. The surrounding selection process — replacement triggers, requirement definition, demo scenarios and scoring — is covered in the ERP selection and evaluation guide.
- Both standard bases. Can expected cost be derived from BOM and routing and maintained separately, or does the system assume one model for everything?
- Separate capture of machine and labour time. Are they distinct inputs, or collapsed into a single process cost?
- Overhead flexibility. Are fixed, percentage, machine-hour, labour-hour and operation-wise bases available, and can they be combined where the operation needs it?
- Rework treatment. Are both the original-work-order and separate-work-order approaches supported, so quality cost can be made visible when required?
- Subcontract participation. Does outside processing cost reach the work order it belongs to?
- Variance decomposition. Can variance be read by cost driver, or only as a total?
- Timing. Is cost visible as the job progresses and at closure, or only after a period-end routine?
- Traceability. Can a variance be followed back to the issues, time records and allocations that produced it?
How exactllyERP supports standard-versus-actual costing
In exactllyERP, actual cost is accumulated per work order from material issued against the job, process cost from the routing steps it passes through, and overhead applied on the configured basis. Actual cost per unit is compared against the expected cost for the item, and the variance together with its source is available at work order closure rather than as a later reconciliation.
Expected cost can be derived from the BOM and routing or maintained separately, so a product group whose structure changes frequently and one that must hold a stable standard across a period can both be handled properly. Actual machine or process time and production labour time can each contribute to the cost of the job. Overhead supports configurable bases including fixed amount, percentage, machine-hour, labour-hour and operation-wise rates, and combinations of these, as applicable.
Rejection is recorded at the specific production stage where it occurs, with a reason, and rework can be charged to the original work order or handled as a separate work order depending on which question the business needs answered. Subcontract and job work cost can participate in the actual cost of the work order, so an item that leaves the plant for an operation still returns with that operation in its cost. Rejection rate is available by stage, by product and by period, which is what allows yield loss to be located rather than inferred.
Because these components are captured as distinct inputs, standard-versus-actual variance can be examined across the cost drivers above rather than only as one combined number.
This page describes capability, not configuration. Which overhead basis suits a given operation, how standards should be reviewed, and where rework should be charged are implementation decisions that depend on the plant and on how the business intends to read the numbers. They are worth deciding explicitly, because each one changes what the resulting variance means.
Common questions
What is the difference between standard cost and actual cost on a work order?
Standard or expected cost is what the job was predicted to consume before it ran. Actual cost is what it did consume, accumulated as material is issued, time is recorded and overhead is applied. Neither number is useful alone: the expected figure is an assumption until it is tested, and the actual figure is a fact without a reference point. The comparison between them is the output that matters.
Should expected cost be derived from the BOM and routing, or maintained separately?
Both approaches are legitimate and exactllyERP supports both. Deriving expected cost from the BOM and routing keeps it consistent with the structure production actually follows, and it updates as that structure changes. A separately maintained standard is more stable, which suits costing that must hold steady across a financial period regardless of engineering change. Manufacturers often need both behaviours for different product groups, so the ability to choose matters more than which default is offered.
How should manufacturing overhead be allocated to a work order?
By whichever basis reflects what actually drives the cost. exactllyERP supports configurable approaches including a fixed amount, a percentage, a machine-hour rate, a labour-hour rate, operation-wise rates, and combinations of these, as applicable. A machine-intensive operation and a labour-intensive one rarely share a sensible basis, which is why one plant-wide rate tends to flatter some products and penalise others. Not every operation needs every method.
How does rework affect the actual cost of a work order?
It depends on how rework is recorded, and exactllyERP supports both treatments. Additional rework cost can be charged against the original work order, which keeps the true cost of that batch in one place. Alternatively rework can be handled as a separate work order, which makes the cost of rework visible in its own right and comparable across periods. The first answers what this batch cost; the second answers what rework is costing the plant.
Does subcontract or job work cost belong in the work-order cost?
Yes, where the outside operation is part of producing that item. Subcontract or job work cost can participate in actual work-order costing in exactllyERP. If it is excluded, the work order understates what the item cost and any make-versus-buy comparison is being made against an incomplete figure, because the outsourced step disappears from the item it belongs to.
Why is variance by cost driver more useful than a single total variance?
Because a single total identifies that something went wrong without identifying what, and different causes require different responses from different people. A material overrun points to yield, scrap or purchase price. A machine-time overrun points to the process or the equipment. An overhead difference may point only to the allocation basis rather than to anything on the floor. Standard-versus-actual variance can be analysed across relevant cost drivers rather than as one undifferentiated total, which is what makes it actionable.