Every folding carton and label job travels through three cost worlds before it closes. The estimator builds a number the sales team can quote. The shop floor produces a number that reflects what actually happened. The ERP books a number the finance team can report. These three numbers are rarely the same, and the gaps between them are not random noise — they accumulate in predictable places, and together they hide the true margin on every job you run.
This piece traces a single hypothetical job through each stage, names where the two models — MIS and ERP — diverge by design, and explains what that divergence costs you in visibility.
Stage 1: The Estimate
The estimate is the most controlled moment in the job lifecycle. A competent estimator has substrate cost, run speed, expected waste percentage, machine rate, and labor rate all in hand before the quote goes out. The model is internally consistent, and for that hour or day, it is the best picture of job economics anyone will have.
But the estimate is already carrying assumptions that will not survive contact with the shop floor.
Substrate cost is the most immediate one. A containerboard price increase of $120–$140 per ton moved through the market in two waves in early 2026. Any folding carton estimate written before that increase used board cost that was materially wrong by the time the job ran — not because the estimator erred, but because the input changed after the number was locked.
Machine and scrap rates are the slower-moving problem. Small and medium converters often set these once, in a custom quoting program, and revisit them infrequently. The rate attached to a press reflects the business as it was when the rate was last calculated — not the business as it runs today.
Stage 2: The Shop Floor Actuals
The job hits the floor, and the divergence starts accumulating quietly.
Setup overruns.
Make-ready took longer than estimated. That overrun is captured in the time clock if your shop floor data collection is working. But the cause — a plate registration problem, a color call that ran long, a substrate that didn't feed cleanly — is almost never captured systematically. A single overrun is insignificant. The same type of overrun on forty jobs in a quarter is a signal.
Speed loss and micro-stoppages.
The estimate assumed a run speed. The press ran at 85% of that speed because of substrate tension variation and two brief mechanical pauses. Neither event triggered a formal downtime record; neither appears in the job cost report as a discrete variance. The cost is absorbed into the run time, and the run time looks close enough to plan that no one asks the question.
Waste.
The estimate carried a waste allowance. Actual spoilage on this job came in slightly higher — enough to affect the per-unit substrate cost, not enough to flag an alert. Over many jobs, waste variance is where estimating accuracy quietly erodes.
Ink.
Ink cost is a structural approximation even when shop floor data collection is diligent. Ink is sometimes mixed from other inks without a formal inventory booking. When consecutive jobs run the same ink, consumption cannot be cleanly divided by job. The practical workaround — widely used in label converting — is to book the estimated ink cost as the actual and reconcile total ink purchased against total theoretical consumption once a year. That is a deliberate, known approximation built into the cost model. It is not a gap that better data entry closes.
Press substitution.
For digital label converters, there is an additional divergence that analogue plants do not face: a last-minute press switch. Quotes must be built against specific digital equipment. When the job moves to a different press at run time — different click-charge structure, different substrate profile, different ink consumption curve — the cost can deviate materially from the estimate without any shop floor error occurring.
Stage 3: What the ERP Books
Now the job closes, and the ERP records a cost. This is where the models formally diverge.
The MIS job profitability report runs on standard costs — rates and assumptions built into the MIS model. Those standard costs are, explicitly, theoretical. As the Labels & Labeling Label Academy documents: standard costs shown on MIS job profitability reports "are theoretical costs, and are not used in the financial statements." The financial statements use whatever the ERP has booked — a different cost model, different timing, different overhead logic.
That means the MIS and the ERP are running parallel cost worlds by design. Reconciliation between them is always a manual or integration step. It is never automatic.
UOM and waste conventions.
The MIS probably carries substrate in square feet or sheet count. The ERP may carry it in pounds, linear meters, or MSF — whatever the purchasing department negotiated. Converting between those units is not a rounding exercise; it requires explicit, maintained conversion logic. If that logic is missing or stale, the substrate cost that lands in the GL is not the same number the MIS computed, even if both are "right" by their own internal logic.
Labor and machine rates.
The MIS uses the rates the estimator used, updated whenever someone updates them. The ERP applies a standard cost that may have been set at the beginning of the fiscal year, against a different overhead structure, using a different definition of what counts as direct labor on this work center. Setup time that the MIS treats as a job cost may be absorbed into overhead in the ERP — or vice versa. Make-ready hours that appear as direct cost on the job cost report may disappear into a production variance account in the GL and never surface in job-level reporting at all.
Overhead allocation.
The applied overhead rate in the ERP is often stale, and it may also be attached to the wrong parameter. Whether warehouse costs should be allocated per thousand labels, per press hour, or per meter of substrate is a modeling decision that requires ongoing re-evaluation. When a plant adds a new machine, shifts mix toward digital, or begins ganging jobs, the overhead parameter itself — not just the rate — may need to change. The ERP continues running the old logic on a fundamentally different business until someone decides to reopen that model.
WIP timing.
The MIS knows when a job completed each operation. The ERP may book the job's full cost at a different point — when the work order closes, when the goods receipt is posted, or when the invoice goes out. In a busy converting operation, the distance between those events can span accounting periods. A job that ran in the last week of one period may not close in the ERP until the following period, shifting both cost and margin to a different month.
Tooling.
Plates, dies, and embossing tools are their own problem, and they sit outside the estimate-to-actual variance framework entirely. On a reprint, tooling may need to be remade at the converter's expense — and that cost falls in a different job record or a different period than the revenue it supports. The practitioner recommendation is to analyze tooling cost separately from job costing. That is sound advice, but it also means a category of real job cost is never visible in the standard job profitability view.
What the Divergence Hides
Each of the gaps described above is defensible on its own terms. Ink approximation is practical. Parallel cost worlds between MIS and ERP exist by design. WIP timing follows accounting rules. Overhead allocation is a modeling choice made once and revisited rarely.
But collectively, these gaps mean the per-job margin you see in the MIS and the per-job margin implied by the GL are two different estimates of a number you never actually observe directly. The true margin — what the job actually cost the business, fully loaded, correctly timed, with actual material prices — is a reconstruction, not a readout.
Variance analysis on job costing typically happens days or weeks after job completion, by which time the operational context that would explain the variance is gone. The press operator who ran the job has moved on to ten other jobs. The substrate batch that caused the waste spike has been consumed. The reason make-ready ran long was documented nowhere.
What gets hidden is not usually a single dramatic loss on a single job. It is the pattern: the jobs that consistently come in below estimated margin, the work centers where speed assumptions are chronically optimistic, the overhead allocation model that quietly overcharges digital runs and undercharges wide-format. Each one looks like noise at the job level. Together, they are a systematic mis-statement of where the plant actually makes money.
That is what the MIS/ERP boundary costs in visibility: not the data you don't have, but the distortion in the data you think you do.
If you're working through how these cost models interact in your own plant, we're glad to talk through what we've seen — no agenda, just a conversation.