Why Job Hours and Work Center Hours Do Not Add Up

On a routing with synchronized parallel machines, the hours reported per station legitimately add up to more than the job's own total, and neither figure is wrong. They answer different questions, and a price or a comparison built from the wrong one is where the error becomes expensive. EDGEBIC by User Solutions reports both, and carries a setting that decides which lens the job-level roll-up uses.

This post covers one return: not building a number on a measure you did not know you were using. It sits under the EDGEBIC results guide and pairs with how EDGEBIC improves quote accuracy, which depends on getting this right.

The Discrepancy That Looks Like a Bug

Here is the situation that generates a support ticket in almost every plant that runs synchronized machines.

Somebody totals the hours reported against each work center for a job. Then they read the job's own total hours. The station sum is higher. Sometimes much higher. The obvious conclusion is that something is double-counting, and the obvious next step is to file it.

It is not a bug, and the documentation says so directly: routings with parallel work centers log the same clock hours on each parallel machine, so the sum can legitimately exceed the job total.

Understanding why takes one physical example. Three drill heads are locked together machining one large frame, running in exact lockstep for eight hours. Ask two questions about that.

How much of the job's timeline did this consume? Eight hours. The operation started once and ended once.

How much machine capacity did it consume? Twenty-four machine-hours. Three machines were each committed for eight hours and none of them could do anything else.

Both answers are true, both are useful, and they differ by a factor of three. There is no single number that satisfies both questions, which is why the system reports both.

The Mechanism: Two Lenses, One Setting

Per-work-center rows always report full effort. Every machine shows what it was actually committed for, which is the only way a capacity conversation can work: a machine booked for eight hours is unavailable for eight hours, regardless of what it was mirroring.

Job-level roll-ups are governed by a policy in the schedule settings called primary hours only, which excludes parallel. Its own description is precise: when on, job and routing hour totals count primary steps only, so synchronized parallel machines do not double-count. Three mirrored drill heads log twenty-four machine-hours, and the job total reads eight.

Two properties of the setting are worth knowing before anyone touches it. It is display and roll-up only, changing no schedule, so nothing moves on the Gantt and no operation is rebooked. And it ships off, which means the default job total includes mirrored hours.

Only One Kind of Parallel Multiplies

This matters because the two kinds of parallel machine behave completely differently, and only one creates the discrepancy.

Dependent parallel, the synchronized mirror. The mirror copies the primary's exact start and end, in lockstep. Hours booked on the mirror are the primary's hours multiplied by the parallel factor, where 1.0 means equal. The mirror does not shorten the operation; it records that a second machine is committed for the same window. This is the case that multiplies.

Independent parallel. All machines start together and each then runs its own share, capacity-checked, with its own setup and per-unit hours. The work is divided rather than duplicated, so the operation genuinely finishes sooner, and totals behave the way intuition expects. This is how parallel machines clear a big lot faster.

The guidance in the product documentation is a good rule to remember: dependent is physics, independent is throughput. Choose a synchronized mirror only when machines are physically locked together, and choose independent when the aim is to shorten an operation. If somebody configured a mirror where they wanted throughput, the roll-up confusion is a symptom of that mistake rather than the problem itself.

The Causal Chain to a Bad Number

Three specific errors come out of not knowing which lens you are reading. Each one is ordinary and each one has cost somebody money.

1. A cross-job comparison. Two jobs are compared on total hours. One has mirrored machines and one does not. The mirrored job reads higher and gets treated as more labor-intensive, when it may simply be more machine-intensive over the same elapsed window. Any conclusion drawn from that comparison, about pricing, about which product to push, about where effort goes, is built on two different measures.

2. A reconciliation that cannot succeed. Somebody sums station hours to rebuild a job total, finds a gap, and either files a defect or, worse, adjusts something until the numbers agree. Forcing agreement between two lenses destroys the one that was answering the capacity question.

3. A quote built on the wrong lens. If the estimate is priced from machine time, full effort is right, because a mirrored machine really was tied up. If it reasons about the job's own hour content, the primary path is right. Using one where the other belongs either overstates the price and loses the order, or understates it and wins an order you should not have taken, which is the more expensive of the two on fixed-price work.

The fix for all three is the same and it is not technical. Decide which lens your organization uses for which question, write it down, and state the lens whenever an hours figure leaves a desk. That is the same discipline as labeling where a number came from, applied to a measure rather than a source.

Measuring the Exposure in Your Own Plant

An hour of work, and the answer is often that this affects very few of your jobs, which is a legitimate finding.

1. List your routings that carry dependent parallel steps. In many shops this list is short. If it is empty, you have no exposure and you can stop here.
2. For one such job, read its total both ways. With the primary hours only policy off, then on. Record the gap. That gap is the size of the ambiguity for that routing.
3. Sum the per-station hours for the same job and confirm the sum exceeds the job total under the primary-path lens. Confirming it is expected is what stops the next person filing it.
4. Audit your last quotes on affected products. Which lens produced the hours in the estimate? If nobody can say, that is the finding.
5. Check for accidental mirrors. Any dependent parallel step where somebody wanted an operation to finish sooner is misconfigured, because a mirror does not shorten anything. Fixing those removes the discrepancy at its source rather than explaining it forever.
6. Choose one convention and hold it. A plant that flips the setting between reports has two incompatible histories, and comparisons across the switch date are meaningless.

For documented outcomes rather than typical ones, the User Solutions and RMDB lineage includes the USS Nimitz refit coordinating more than 26,000 tasks, and Cummins deploying this scheduling approach across 33 locations. Those belong to their engagements and are quoted as heritage, not as forecasts.

Where This Does Not Apply, and One Real Warning

Four limits, and the last one is a genuine caution rather than a limit.

Serial routings are unaffected. With no parallel steps, the per-station sum and the job total agree and both settings produce the same figure. For most jobs in most shops there is nothing here.

Independent parallel does not create the gap. It splits work rather than duplicating it. If your parallel machines are all independent, your totals behave conventionally and the setting will change little.

The setting changes presentation, not reality. It does not reduce machine commitment, free capacity, or alter a single date. Anybody hoping to relieve a loaded station by changing how hours roll up has misunderstood what the switch does.

Mirrors are booked without a capacity check, and that is a real risk to watch. A dependent mirror is committed on the extra machine regardless of that machine's own calendar and other load. If a machine used as a mirror also carries independent jobs, it can end up committed twice in the same window and read above 100 percent on the load view. The documented practice is to keep machines used as synchronized mirrors clear of other load in those periods and to watch the anomaly report for over-allocation findings. That is a scheduling caution rather than a reporting one, and it is worth more attention than the roll-up question that brought you here.

The takeaway

When station hours sum to more than a job's total, nothing is broken: a synchronized mirror records the same clock window on a second machine, so full effort and primary-path hours are two correct answers to two different questions. The cost is never the discrepancy, it is the comparison or the quote built without knowing which lens produced the figure. Find the short list of routings that carry mirrors, decide one convention for which lens answers which question, and check that every mirror on that list is there for a physical reason rather than because somebody wanted an operation to go faster. To see both lenses against your own routings, book a demo of EDGEBIC, and if you are running the older platform, the move from RMDB to EDGEBIC brings the reporting with it. Read this next to how parallel machines clear a big lot faster and how EDGEBIC improves quote accuracy.