Why Two Utilization Numbers Disagree, and Which One to Trust

Two reports can show different utilization for the same station in the same window, and both are computing correctly. The difference is the denominator: available hours can be approximated quickly or derived from the real shift calendars, and only one of those is safe to put in front of somebody who is about to authorize a machine. EDGEBIC by User Solutions labels which is which, and knowing the difference is worth whatever your last capacity decision cost.

This post covers one return: not making an expensive decision on a convenient number. It sits under the EDGEBIC results guide and is the data-quality companion to how capacity evidence answers a new machine request.

Utilization Is a Ratio, and the Denominator Is the Soft Part

Utilization looks like a simple number. It is loaded hours divided by available hours, expressed as a percentage.

The numerator is usually easy. Loaded hours come from the schedule, and the schedule knows exactly what it booked where.

The denominator is where the judgment lives. Available hours means the hours this station could work in this window, and answering that properly requires knowing which shifts it runs, on which days of the week, which holidays apply, how many instances it has, and whether anybody has entered a per-day capacity override because Tuesday was down for maintenance or Thursday got an approved overtime block.

That is a real computation. Doing it across every station over a long window costs something. So there is a fast approximation, and there is a calendar-true version, and they do not agree.

The Mechanism: Two Reports, Two Denominators, Both Labeled

EDGEBIC ships both, and it tells you which is which rather than leaving you to discover it.

Work Center Performance is the light, fast utilization trend per work center over a window. It carries scheduled hours, actual hours, available hours, utilization percentage, a rate-performance percentage comparing actual to scheduled, jobs touched, and a bottleneck label. Its available-hours figure is documented as an approximation that ignores shift calendars.

Work Center Utilization is the heavyweight capacity report: a KPI strip across plant utilization and scheduled, actual, and available hours, with counts of bottleneck, overloaded, and underloaded stations, over four tabs covering per work center, a fourteen-day daily heatmap, top jobs by load, and plan against actual. Its available hours come from the real shift calendars and overrides, which is why it is the authoritative capacity view.

The documented rule is unambiguous: when the two disagree, Work Center Utilization wins. That is not a hedge, it is a stated authority.

Why the Gap Turns Into a Wrong Decision

The chain from a soft denominator to a bad purchase is short, and each link is ordinary.

1. Somebody needs a utilization figure. Usually for a meeting, usually today.
2. They open whichever report loads fastest and reads most simply, which is the approximate one.
3. The percentage lands somewhere near a threshold. The rating bands are critical above 100 percent, high from 85 to 100, good from 60 to 85, low from 30 to 60, and idle below 30.
4. The band, not the number, becomes the argument. People quote labels. "It is running critical" and "it is running good" produce entirely different meetings from percentages a few points apart.
5. The decision follows the band. Approve overtime, approve a machine, refuse an order, or do nothing.

The failure is at step three, and it is invisible. An error in available hours moves the ratio proportionally, and the direction is not predictable, because it depends on how far the approximation happens to sit from your actual shift pattern at that station. A shop running one uniform shift on every station may see a small gap. A shop with a second shift on two stations, a Saturday pattern on one, and a per-day override history will see a larger one, and it will vary by station, so you cannot mentally correct for it.

The cost is not the reporting error. It is the decision the error changed.

The Second Disagreement: Averaging Hides a Bad Day

There is a related mismatch that trips people up for a different reason, and it is worth separating clearly because the fix is completely different.

The daily heatmap colors each cell by that day's load against that day's capacity. A red cell means oversold: more hours booked than the day holds. The per work center tab averages across whatever range you selected.

So a station can read comfortably in the good band across a month and still be red on one Tuesday. Both are correct. They answer different questions:

• The averaged view answers do we have enough capacity here, which is a capital and staffing question.
• The day-level view answers is this specific day plannable, which is a scheduling question.

Reading them as competing is the mistake. An averaged overload means you are short of capacity and no amount of resequencing fixes it. A single red day on a healthy average means work needs moving, and moving work is free. Confusing the two either buys a machine to solve a Tuesday or spends a week resequencing around a genuine shortfall.

Measuring the Gap in Your Own Plant

Half an hour, three stations, and you will know whether this matters to you.

1. Pick three stations with different shift patterns. Include your constraint and include one with a second shift or a weekend pattern if you have one, because that is where the approximation drifts most.
2. Pull both utilization figures for the identical window. Same from and to dates, no other filters.
3. Record the two percentages and the gap. One row per station.
4. Check the bands. For each station, does the gap move it across a rating boundary? A gap that keeps a station in the same band is academic. A gap that moves it from good to critical is the whole point of this exercise.
5. Open the day-level heatmap for the same window. Note whether any single day is red on a station whose average looked healthy. That is the averaging effect, quantified.
6. Audit your last three capacity decisions. For each one, which number was in the room? Re-run it against the calendar-true figure and see whether the conclusion holds. If any of them flips, you have just priced this post for your own plant.

For documented outcomes rather than typical ones, the User Solutions and RMDB lineage includes GE Railcar moving from 30 percent to 90 percent on-time delivery, and Cummins deploying this scheduling approach across 33 locations. Those belong to their engagements and are quoted as heritage, not as forecasts.

Where This Distinction Does Not Matter

Four honest limits, because "always use the heavy report" is not the lesson.

It does not matter for trend and rank. If the question is which stations are running hottest, or whether this station is busier than last month, the fast report answers it correctly. Direction and ordering survive an approximate denominator. Absolutes against a threshold do not.

It matters less if your calendars are uniform. A plant where every station runs the same single shift with no overrides will see a smaller, more consistent gap. Measure it once and you will know whether to care. Do not assume you are that plant without checking, because a single station on a second shift breaks the assumption.

Neither number tells you whether the work is worth doing. A station can be at 95 percent building an order that should never have been accepted. High utilization is not a virtue and it is not a target, as the 100 percent utilization trap covers. The schedule answers whether the load belongs there; utilization only answers how full it is.

No utilization figure decides anything by itself. It sizes the problem. Whether the answer is overtime, an alternate machine, an outside process, a moved date, turning the order down, or capital is a management judgment the number informs. The reason to insist on the authoritative denominator is not that it decides for you. It is that the wrong one can make a decision look obvious when it is not.

The takeaway

Utilization is a ratio and the denominator is the part that takes work, so a fast figure and a calendar-true figure will differ on the same station in the same window, and neither is broken. What breaks is the decision made when the fast number carries a rating band into a room where somebody signs. Agree once that capacity conversations use the calendar-true report, keep the fast one for trends where it is genuinely better, and separate the averaged capacity question from the day-level scheduling question, because a red Tuesday and a short quarter need opposite responses. Measure the gap on three of your own stations before you trust either. To see both numbers against your own calendars, book a demo of EDGEBIC, and if you are on the older platform, the move from RMDB to EDGEBIC brings the capacity reporting with it. Read this next to how capacity evidence answers a new machine request and how much capacity are you already losing.