What Is an Optimizer Proposal in Scheduling?

An optimizer proposal is a complete alternative schedule the optimizer has computed and is holding on screen for review, existing in memory only until a planner presses Accept. In EDGEBIC by User Solutions this deferred-apply design is the whole reason the optimizer is safe to run casually: while you read the verdict, the KPI comparison, and the move list, the database is exactly as it was, the floor's printed plan is still valid, and pressing Discard leaves nothing behind at all.

This entry is part of the EDGEBIC glossary series; for the broader vocabulary of production planning, see the manufacturing glossary.

How an Optimizer Proposal Works

The sequence is deliberately three separate acts, and keeping them separate is the point.

Compute. You choose a goal preset and a time budget, then press Run. The optimizer builds many complete alternative schedules and measures each one. Cancel stops the run harmlessly with the message that nothing was changed.

Review. The result appears as five readable pieces. The verdict is one honest sentence, either describing the improvement or stating that your current plan is still the best. The badge is the certificate of how hard the answer is. The KPI table compares measure by measure with the direction spelled out in words rather than color alone. The move list names only the operations that actually move, one row each. And the safety strip repeats three fixed guarantees: nothing saved yet, completed work untouched, never worse than current plan.

Decide. Accept writes; Discard does not. There is no third path and no partial application. You cannot take four of the six moves.

Two properties make the held proposal more than a preview. It is a full schedule rather than a diff, so accepting it does not require replaying anything. And it carries a stamp of the schedule tables as they stood at computation time, which is what lets Accept detect that the world moved underneath it.

A Worked Example

From the documentation's sample plant, Monday July 20. Three jobs run through Saw-1 and then CNC-Mill-1, and the current plan leaves JOB-2026-0123 showing Days Late 1.

The planner picks the on-time-first goal and a thirty-second budget, and runs. The result comes back with a verdict reading that the plan makes one more job on time and reduces lateness by several hours, and that nothing already started was touched. The KPI table shows late jobs going from 1 to 0 and makespan dropping. The move list holds six rows, each naming a job, a work center, its current start, its proposed start, and a short reason such as starts four hours earlier.

At this moment the database still shows Days Late 1. Nothing has happened.

The planner presses Accept. Thirty seconds later the Drive Schedule grid reads Days Late 0 on all three jobs, the Gantt bars for five operations sit in their new slots, and the audit trail holds the explanation in plain English. Had they pressed Discard instead, the grid would still read Days Late 1 and nothing else would have happened at all.

How EDGEBIC Uses Optimizer Proposals

Several behaviors follow from a proposal being held rather than applied:

• Closing the application discards it. A proposal lives only on that screen, which is always safe precisely because nothing was saved.
• Accept refuses on a stale stamp. If someone rescheduled or actuals landed while you were reading, Accept will not overwrite. A banner reports that changes arrived while you reviewed and offers a Re-run that solves again on the latest data. The held proposal is not destroyed, so you lose nothing but the time.
• Accept goes through the normal pipeline. It is not a special write path. The same protections that apply to recorded work on a regular schedule run apply here, which is how the completed-work guarantee is kept honest.
• One audit record captures the reasoning. The text stored is the same text the Explain dialog showed, so the dialog and the months-later audit tell literally the same story.
• The never-worse guarantee sits underneath it. A proposal only reaches the screen if it measurably beats the current plan under the goal you asked for. See the never-worse clamp.
• The badge tells you what kind of answer it is. Multi-run search reports the best of a number of complete schedules tried, an honest search with no claim of proof. The mathematical solver additionally reports how close the result is to the theoretical best. See an optimality certificate in scheduling.

The habit worth building is to run the optimizer more often than you accept it. Because a proposal costs nothing, trying a second goal preset to see how the trade-offs move is a cheap way to understand your own plant's constraints, and discarding both is a perfectly good outcome. The optimizer guide covers the whole screen, how to run and read an optimization walks a first run, and an objective preset in scheduling optimization explains the goals you can choose between.