Operator and Skill Scheduling in EDGEBIC: Labor as a Real Constraint

EDGEBIC by User Solutions can treat people as a finite resource, not an assumption: a routing step that names a required skill is scheduled only into a window where a certified, rostered, present operator still has unbooked hours. That single rule closes the oldest gap in machine-only scheduling, where a plant with one certified welder can be shown five simultaneous jobs on five machines and nobody notices until the shift starts.

This post covers what the labor constraint is and when you need it. For the buttons, see setting up operators, skills, and rosters. For the arithmetic, see how EDGEBIC books operator hours.

The Failure This Exists to Kill

A machine-only scheduler asks one question: does this work center have free hours? If Weld-A and Weld-B both have eight free hours on Monday, two Inconel jobs both get Monday. The schedule prints, the supervisor prints it, and at 08:00 one certified welder discovers he is expected in two places.

Nothing in that sequence is a software defect. The scheduler answered the question it was asked. The problem is that the question was incomplete, and the cost shows up as silent slip: the plan degrades quietly, dates move without anyone deciding they should, and after a few weeks the shop stops trusting the printed schedule and runs from memory instead. Skilled-labor availability is one of the constraints that separates a plan from a wish, and it is a standard concern in the job shop labor scheduling literature.

With the labor constraint configured, EDGEBIC asks both questions at once: is the machine available, and is there a qualified person with hours left to stand at it.

The Model in Six Pieces

Two design choices are worth naming because they explain most of the behavior downstream.

People carry hours, not presence. Each operator holds a budget of hours per shift per day. Not a yes-or-no availability flag, and not a pool of hours per skill. A person certified in three skills still has one budget: booking them through TIG work leaves fewer hours for their inspection work, exactly like reality. A per-skill pool would sell the same person into three pools at once, which is the same double-count in a new costume.

The roster is a weekly pattern with absences as exceptions. That mirrors how machines are modeled, with weekly shift hours and holidays as exceptions, and it mirrors how plants actually roster. Per-date rosters would mean thousands of rows per planning horizon and no better answer.

Machine and Human Compose on One Step

A routing step can carry both a machine target and a labor requirement, and they answer different questions:

• The work center, or the work center group, answers which machine.
• The required skill answers which qualified person.

They are deliberately separate dimensions. A "Certified Welders" machine pool is the classic modeling mistake here: a pool has no concept of certification, roster, or leave, so a people-shaped pool constrains nothing at all.

When both sit on one step, machine selection accounts for staffing. A pool member whose shifts cannot be staffed for the step's skill loses its bid instead of winning on machine capacity and then sliding when no operator can be found. Picking the machine first and discovering the staffing problem afterwards produces schedules that look right on the machine calendar and fail on the floor.

Attention: Not Every Machine Needs a Person All Day

The one field that surprises people is operator attention, the fraction of a person a running machine consumes. Setup always books one full person, because somebody has to set the machine up. The fraction describes only what happens after that:

This matters more than it first appears, because it is what lets one setter start a cell of seven machines without the schedule pretending they are in seven places. The full explanation, including the seven-machine worked example, is in the attend fraction.

The default deserves one warning here: it is 0.0, not 1.0. A step with a required skill and untouched attention gates operators on setup only. If your welders must stand at the machine for the whole run, set attention to 1.0 explicitly. The default was chosen so that a forgotten field under-books visibly rather than silently overselling people, which is the safer direction to be wrong in.

Naming the Person: The Pin

Leaving the operator field on automatic lets the scheduler staff from the whole qualified pool and load-balance across it. Setting it to a named person is a hard pin with deliberate consequences:

• The step schedules only when that person has hours. The rest of the qualified pool is never a fallback.
• If the pinned person is on leave, the step slides through their absence to their first day back instead of quietly going to a colleague.
• Clearing the skill from the step clears the pin.

That sliding behavior is the feature, not a limitation. A pin exists for a reason: a customer requirement, a sign-off authority, the one person you trust with that fixture. When such a person is away, the honest answer is a visible schedule slip, not a silent substitution nobody approved. Use pins sparingly and the automatic setting most of the time.

The Rule to Internalize: Unstaffed Capacity Is Unusable

For a skill-gated step, a shift with nobody qualified rostered is not partially available. It is unavailable.

The documented walkthrough makes it concrete. A weld step needs 6.5 hours on Weld-1 on Thursday. Weld-1 has eight free machine-hours Thursday. But the step requires TIG-Weld, and the only certified welder is on leave that day. The step does not get Thursday. It moves to Friday 08:00 to 14:30, and Thursday's Weld-1 hours go to a job with no skill requirement. Nothing flags the machine as idle, because for this step it was not genuinely free.

Planners who come from machine-only scheduling read this as the schedule "wasting" a machine day. It is the opposite: the machine day was never real for that operation, and the plan now says so before the shift rather than after it.

Where the Crew Plan Shows Up

One clarification prevents a whole category of misunderstanding: these are plan data. Crew assignments are regenerated wholesale on every scheduling run, so they answer who is planned to do this work, not who actually did it. Recorded work still comes from the shop-floor kiosk and actual dates, which are the surfaces that never move.

Opt-In by Design

The labor constraint is per step. With no operators configured, or on any step with no required skill, scheduling behaves exactly as it does without the feature. That property is worth using deliberately during rollout: start with the two or three operations where certification genuinely gates the work, get the roster right for those, and expand only when the first set is trustworthy. A plant-wide skills model built in one weekend from half-remembered certifications produces slipping dates nobody believes.

The optimizer takes the same conservative stance. Jobs whose routings carry a skill or a pin are reproduced exactly as the standard engine scheduled them rather than rearranged by a model that does not represent labor, which is the rule described in locked jobs and the never-worse guarantee. An optimizer that does not model a constraint should never move work subject to it.

What the Labor Constraint Deliberately Does Not Do

Three boundaries are worth knowing up front, because each one is a design decision rather than a gap waiting to be filled.

It does not replan by itself. Entering a vacation on Friday changes nothing until somebody runs a schedule. EDGEBIC never moves a live plan in response to a master data edit, because silent replanning is how planners stop trusting a tool. Enter the leave, run the schedule, read the result.

It does not schedule people to exact minutes. Operators are budgeted in hours per shift per day, the same fidelity the engine uses for machines. Two of one person's bookings can overlap on the wall clock inside a shift, and no day can ever oversell their hours. Promising exact clock intervals three weeks out would be precision nobody can execute against.

It does not change the shop-floor kiosk. Punching is untouched, and crew assignments are the plan rather than a record of who worked. Keeping planned crew and recorded labor in separate places is what stops a planning artifact from quietly becoming a payroll input.

Getting Started

The practical order is: create the skills you certify against, add operators, certify them with real expiry dates, roster them on the global shifts they work, record known leave, then set the required skill and attention on the routing steps that need them. Setting up operators, skills, and rosters walks each screen. The traps worth knowing before you start, from expiry surprises to unstaffable custom shifts, are in operator and skill mistakes.

For where labor sits among the other constraints, the complete EDGEBIC guide covers the full pipeline, and what production scheduling is connects it to the general discipline. Scheduling skilled labor is a recurring theme in defense and aerospace work in particular, where certification records and schedule records have to agree: EDGEBIC by industry covers those cases.