How Much Capacity Are You Already Losing?

Most shops are losing capacity in five specific places, and none of them appear on a utilization report: sequence-driven changeover waste, capacity assumptions that do not match the machine, idle gaps between operations, rate erosion nobody tracks, and rework that consumes hours without ever being scheduled. Every one of those is measurable in your plant this month with a clipboard and a spreadsheet. This guide gives you the arithmetic for each, so you can size what is actually on the table before anyone quotes you software.

EDGEBIC by User Solutions is a finite capacity scheduling system, and its most useful trait during an evaluation is unglamorous: it computes capacity the same way twice, once for the schedule and once for the dashboard, so the number the planner sees and the number the engine uses cannot disagree. That consistency is what makes hidden loss visible. But you do not need the software to start finding it.

First, Compute What Capacity Actually Is

Before you can find lost capacity you need a defensible definition of the capacity you have. Most shops do not have one, and the gap between the number in the plan and the number on the floor is itself the first source of loss.

The arithmetic is short. For one work center on one shift on one day:

gross hours = shift end - shift start
net hours   = gross - breaks - scheduled downtime - partial holiday
capacity    = net hours x number of machine instances x utilization %

Work one through with real values. A CNC center runs 08:00 to 16:30, with a 30-minute break and a one-hour preventive maintenance block every Wednesday. It has two machines and runs at the standard 100% utilization.

That 14.0 is the number a job must fit into. If a job needs 10 hours it fits; if it needs 16 it spills to the next day. Want the machine planned tighter than that? Take the hours out where they can be seen: shorten the shift to the productive window, add a downtime event, or override the day. This is the calculation EDGEBIC runs before allocating a single hour, and it is the same calculation whether you are looking at the Gantt or the capacity dashboard. See how work center capacity is calculated for the full resolution chain, including per-day overrides for approved overtime.

Now do this by hand for your three busiest work centers and compare each result to what your current plan assumes. The differences you find are loss source number two, and they are usually the biggest surprise.

Loss Source 1: Changeover Consumed by Sequence

This is the largest recoverable block in most shops with mixed product runs, and it is invisible because planning systems typically charge a flat setup allowance regardless of order.

The documented paint booth example makes the size of the effect concrete. Three jobs, one booth. A flat 30-minute allowance charges 90 minutes of setup in total. The real transitions are 60 minutes going from a light color to a dark one and 240 minutes going from dark back to light (a full solvent purge). The floor lives through 510 minutes.

Two separate losses are hiding in that one comparison. The first is honesty: the plan was understating the work by roughly half a day. The second is recoverable: once the true from-to times are loaded, the same three jobs cost 330 minutes in due-date order and 90 minutes in like-to-like order. That is a 73% cut on one machine on one day, achieved by re-ordering rather than by working faster. The paint shop walkthrough traces every minute, and what a setup family is explains how the from-to matrix stays manageable when you have hundreds of products.

Size it yourself: two weeks of clipboard logging at your worst machine, then a paper re-sequence of one representative week. The difference is your number.

Loss Source 2: Capacity Assumptions That Do Not Match the Machine

Three inputs get set once during a configuration and then quietly drift out of date. Each one distorts capacity in both directions.

Instance count. How many of a work center's machines does the plan think exist? If one has been down for three months, or one was added last year and never entered, every capacity figure for that work center is wrong by the fraction of one machine. See machine instances explained.

The productive window. Somewhere the plan has to account for the difference between clock time and productive time: breaks that overrun, tool changes, minor stops, the last ten minutes of a shift. The utilization percentage in the formula does not do it for you, because work centers run at 100% and the value is not editable on the work center screen. So the haircut belongs in the shift hours, in a downtime event, or in a per-day capacity override, and the right size for it comes from measurement rather than a guess. How to configure instances and utilization covers the trade-off.

Shift calendar. Does the plan know about every plant holiday, every shift that does not actually run, and every hour of standing maintenance? That last one is the leak, because there is no maintenance record to fill in: a weekly PM hour reaches the calendar only if someone nets it out of that machine's shift hours or enters a per-day capacity override on the affected dates. An hour a week the plan does not see is 52 hours a year per machine that you have already spent.

Size it yourself: recompute one work center by hand with the formula above and compare it to what your plan assumes. Do this for three work centers. The pattern will be obvious.

Loss Source 3: Idle Gaps Between Operations

A job that finishes at Cutting on Tuesday and starts at Milling on Thursday has spent a day and a half doing nothing, and both work centers report fine utilization for the week. Gaps of this kind come from queue time set too generously, from dependency sequencing that does not reflect how work really moves, or from a machine that was booked for something else in between.

This class of loss is specifically checked for. EDGEBIC's anomaly report includes an instance idle-gap family that flags capacity waste between allocations, alongside checks for dependency violations where a downstream step is scheduled before the step it depends on. See what the anomaly checks actually look for and, if you are already seeing this pattern, why a schedule has idle gaps.

There is also a legitimate way to close many of these gaps rather than eliminate them: overlap the operations. If a downstream step can start on the first transfer batch rather than waiting for the whole lot, the gap disappears without either machine working faster. That is lot streaming, and it is one of the few mechanisms that genuinely shortens a job's elapsed time without adding capacity.

Size it yourself: pick ten completed jobs. For each, total the time between the end of one operation and the start of the next. Divide by total elapsed job time. That percentage is your queue and gap share.

Loss Source 4: Rate Erosion Nobody Tracks

Your routing says a part takes 0.20 hours at Cutting. Does it? Standards get set once and then the tooling changes, the material changes, the operator changes, and nobody revisits the number.

Rate erosion is insidious because it does not cause a visible failure. The schedule simply runs a little late every day, everywhere, and the cause is distributed so evenly that no single job looks like the problem.

The fix is comparison, and it requires actuals. When operators log real hours and pieces against the plan, planned against actual becomes a report rather than an argument. EDGEBIC's shop floor terminal captures punches by state (setup, run, idle, down, rework, teardown) with reason codes, which is what makes the comparison specific enough to act on. See how actuals flow into the schedule and reading the Gantt planned vs actual.

Size it yourself: pick five high-volume parts. Compare routed hours to actual hours across the last twenty jobs. Any part where actual exceeds routed by more than 15% is quietly eating capacity every time it runs.

Loss Source 5: Rework That Is Never Scheduled

Rework consumes machine hours, operator hours, and setup time, and in most shops none of it appears on the schedule. The job is already "done" in the system when it comes back.

The result is a work center that is 100% booked on paper and 115% consumed in reality, which shows up as everything running late for reasons nobody can name. Capturing rework as a distinct punch state, with a reason code, is what turns it from a mystery into a line on a report.

Size it yourself: ask your quality lead for rework hours by work center over a quarter. Divide by that work center's total available hours. That percentage is capacity you have already spent and never planned.

Putting the Five Together

The "typical size" column deliberately gives no percentages, because the honest answer is that the distribution differs by shop. What is consistent is the shape: shops with sequence-dependent changeovers find most of their loss in the first row, and shops with long multi-step routings find most of theirs in the third. A single OEE percentage behaves the same way, and only becomes actionable once you know which of its three losses is costing you the most.

For the broader framework on capacity as a planning layer distinct from material planning, the ASCM body of knowledge is the standard reference, and NIST MEP publishes assessment guidance that pairs well with the measurements above.

What Recovering It Looks Like

None of the five requires new equipment. All five require a plan that knows what the machines actually are and sequences work against that reality. That is the entire premise of finite capacity scheduling, and it is why capacity gains from scheduling tend to appear quickly: you are not building capability, you are stopping waste that was already happening.

In the documented User Solutions record, Technical Glass Products reported a 4% capacity gain along with a two-week reduction in lead time. Treat that as evidence that the category of recovery is real at a specific plant, not as a forecast for yours. Your number comes from the five measurements above.

Once you have sized the prize, what an unreliable schedule actually costs you prices the rest of the ledger, and the measurable results guide walks through the mechanism behind each documented outcome.

To see your worst work center scheduled against its real capacity, contact US with one routing, one shift calendar, and two weeks of changeover observations. We will run it through EDGEBIC and you can compare the honest capacity number to the one your plan uses today.