Overtime as a Decision, Not a Surprise

Most overtime is not bought to add capacity. It is bought to recover a schedule that turned out to be wrong, and it is authorized on the day, at premium, with no time left to compare it against anything cheaper. That is the difference between planned overtime, which is a legitimate capacity decision, and unplanned overtime, which is a payment for late information.

EDGEBIC by User Solutions attacks the second kind two ways: it shows the overload weeks before it bites, and it removes hours that never had to be worked. This post covers both mechanisms, the arithmetic on documented examples, and the substantial part of the overtime bill that no scheduler can touch. It sits under the EDGEBIC results guide.

Three Kinds of Overtime, Three Different Fixes

Before any mechanism, a classification, because the fixes do not transfer.

Shops track overtime as one line on a payroll report, which is why the third bucket stays invisible. It is usually the largest, and it is the only one that can be removed rather than merely managed.

The Arithmetic of Avoidable Overtime

The documented paint booth case shows the third bucket in a single day.

Three jobs on one booth, one 8-hour shift. Scheduled in due-date order (white, black, white) with real changeover times loaded, the day runs 08:00 to 16:45: 8.75 hours in an 8-hour shift. That is 45 minutes of overtime, or a job pushed into tomorrow, and in a plant with a promised ship date it is usually the former.

The same three jobs resequenced light-before-dark run 08:00 to 12:45. Total changeover falls from 330 minutes to 90, a 73% reduction, and the day finishes with more than three hours to spare. Every changeover hour recovered that way is a production hour back, and an overtime hour never bought.

The work did not change. The hours worked changed, because the order changed. Every hour of that difference would have been paid at premium and recorded as a capacity problem, and it was a sequencing problem.

The mechanism is the setup matrix, which charges the real changeover for each transition instead of a flat number, and then makes sequencing a lever worth pulling. At plant scale, the optimizer carries a least-setup goal alongside on-time-first and reports both sides of the trade before you commit to either.

Seeing the Overload While Options Are Still Cheap

The second and third buckets both shrink when you find out early, because the option set is time-dependent.

EDGEBIC computes each day's capacity explicitly rather than assuming a shift length. The documented worked example, for a machining center on a Wednesday:

08:00 to 16:30                        = 8.5 gross hours
less 30-minute break                  = 8.0
less 1.0 h recurring maintenance      = 7.0 net
x 2 machine instances                 = 14.0
x 85% planned utilization             = 11.9 usable hours

Committed work is placed into those hours, so an overloaded day exists in the plan the moment the order causing it is scheduled. The utilization report's daily heatmap breaks the next 14 days into one row per work center per day, and its rating bands mark anything above 100% as critical. In the documented plant example, one resource shows 420 scheduled hours against 400 available while the plant average reads a comfortable 68.1%.

Six weeks of notice buys options that two days of notice does not: move a job to an equivalent machine, resequence, negotiate a date, or plan the overtime properly at a normal rate with a full crew. See reading a red day for the diagnosis routine.

Making Planned Overtime a First-Class Input

When overtime is the right answer, it should be in the plan rather than in a conversation.

A daily capacity override does that. The documented case: a planner approves a four-hour Saturday run on a specific work center and shift for a rush order, records the reason, and the scheduler treats those four hours as real capacity on a Saturday it would otherwise skip. The planner enters the total hours the machine will genuinely be available, and the engine uses that figure verbatim rather than multiplying it by instances or efficiency.

Two things follow. The schedule downstream of that Saturday is now correct, because the extra hours are visible to every other job's dates. And the overtime is documented with a reason, so the monthly conversation about why the premium hours were spent has an answer attached to each occurrence.

A monthly override handles the other shape: a retooling week where the plant can only guarantee a fixed number of machine hours across the period. See how EDGEBIC calculates work center capacity for the resolution order between daily overrides, monthly overrides, and the shift formula.

Overtime Caused by Recovery

The second bucket is disruption. Something breaks, the plan is invalid, and overtime buys back the difference.

The lever here is how good and how fast the recovery calculation is. In the documented breakdown case, a mill fails Wednesday morning on a job with 10.5 of 31 planned hours logged. The reschedule preserves the completed work, replans the outstanding 20.5 hours from Thursday, and cascades the two downstream operations. Final completion moves from July 21 to July 23, against a due date of July 25.

No overtime required, because the schedule had room and the recalculation found it. Without that calculation, the same event usually produces a weekend, because nobody can prove the job still fits. See the breakdown walkthrough and how completed work is preserved.

A Warning Worth Having Switched On

One diagnostic deserves mention because it catches overtime hiding in the data rather than on the floor. EDGEBIC's anomaly checks include a warning when total daily bookings on a work center exceed the shift-design ceiling (the sum of each shift's maximum hours per day, times instances), and a critical check when they exceed the physical limit of 24 hours times instances.

The first one is the useful one for this topic. A plan that quietly books more hours than the shift design allows is a plan that will be executed as overtime whether or not anyone authorized it. Catching it in the anomaly report is cheaper than catching it on a timesheet. See what the anomaly checks look for.

Measuring It Properly

Overtime as a single payroll line cannot be reduced, because you cannot tell which part of it was avoidable. Split it before you try.

For every authorized block of overtime, record four fields: the work center, the date, the hours, and one word for the cause (capacity, recovery, or sequence). Two months of that produces the only chart that matters, and it usually shows one machine and one cause carrying most of the total.

Alongside it, track setup hours on your two worst work centers separately from run hours. This is the number the sequencing mechanism moves, it is almost never measured on its own, and in the documented paint case it is the entire difference between an 8.75-hour day and a 4.75-hour one.

Then re-measure after the changes land. A shop whose sequence-caused overtime falls while its planned capacity overtime stays flat has removed waste rather than shifted it, which is the outcome worth defending in a review. For the cost framing across every category, see manufacturing cost reduction.

What the Software Cannot Do Alone

It cannot remove genuine excess demand. If the constraint needs 420 hours in a month that holds 400, somebody is working late or something is not shipping. The schedule makes the gap precise and early. It does not close it.

It cannot authorize anything. The daily override records a decision a human made. Whether Saturday is worth the premium is a business judgement with a cost on both sides.

It cannot control attendance. Absenteeism, sick days, and turnover create real capacity gaps that no plan anticipates. Scheduling shows the consequence quickly, which helps, and that is all.

It cannot enforce every shift rule for you. Some fields, such as a shift's overtime-allowed flag, exist as information for planners rather than as constraints the engine enforces. Treat the ceiling checks in the anomaly report as your enforcement point, and do not assume a setting is blocking something it merely records.

It cannot fix an unrealistic utilization figure. If your work centers are configured at 100% planned utilization, every day's capacity is overstated and the schedule will look fine right up to the point where the floor works late. That number is a planning assumption, and it should reflect what your plant actually sustains.

Want to know which of the three buckets your overtime falls into? Bring a month of authorized overtime and the schedules for those weeks to a demo, and we will sort them.