Canceling the Optimizer Is Safe, and the Run Is Discarded

Correction first: canceling a running optimization in EDGEBIC does not hand you a plan. The run stops, the status reads "Stopped, nothing was changed," and the work in progress is discarded. If you have read that stopping early returns the best schedule found so far, along with an applyable result and a proven bound, that describes how solver engines behave in the abstract rather than what the Optimizer tab does. Cancel is an off switch, not a "give me what you have" button.

What is true, and genuinely worth knowing, is that Cancel is completely harmless. EDGEBIC by User Solutions runs the optimizer in memory and writes nothing to your schedule until a planner presses Accept, so stopping a run mid-flight cannot damage the plan, disturb a colleague, or leave anything half-applied. This post separates the safety, which is real, from the anytime behavior, which the screen does not offer. For the wider picture, see the EDGEBIC optimizer guide.

What Cancel Actually Does

Press Run and the status shows "Trying schedules..." while the optimizer builds and measures complete alternative schedules. Press Cancel during that and the run ends immediately with a single message: "Stopped, nothing was changed."

That message is the whole story. There is no proposal, no comparison table, no move list, and no badge. The screen returns to where it was before you started, and your current plan is what you still have. The message is deliberately worded as a reassurance rather than an error, because nothing failed: you asked the optimizer to stop, and it stopped. The same wording is cataloged alongside every other run-state notice in optimizer messages and what they mean.

Why Nothing Survives the Cancel

The reason a canceled run leaves nothing behind is the same reason it is safe. The optimizer is deferred-apply from end to end: a run computes entirely in memory, presents its result on screen, and touches the database only when a planner presses Accept. Discard drops the proposal and writes nothing. Closing the application drops it too, and that is equally safe.

A run that never reaches the point of presenting a result therefore has nothing to leave behind. It held its candidates in memory, it had written nothing anywhere, and canceling released the lot. The absence of a partial plan and the absence of any risk to your schedule are two sides of one design decision, not a shortcoming bolted onto a good idea.

The Time Budget Is a Ceiling, Not a Fixed Duration

Here is the part of the original story that holds up. When you pick a time budget of 10, 30, or 60 seconds, you are setting a maximum, not a stopwatch the optimizer must run down.

The multi-run search uses the time to try more job orderings, and reports the best complete schedule it built, badged as "best of N schedules tried" and clamped so it can never be worse than your current plan. The mathematical solver can go further and stop early: if it proves that no schedule can beat the one it is holding on your chosen goal, it returns straight away, because further searching cannot improve on a proof. That is why a run given 60 seconds can legitimately come back in three, badged "proven optimal" or "proven within X% of optimal."

An early finish and a cancel look nothing alike on screen. The finish shows the whole comparison. The cancel shows only the stopped message. Details on choosing the number are in setting the optimizer time budget, and the badge language is unpacked in how a proven optimality gap builds trust.

What You Give Up, and What You Never Risk

Canceling costs you the run: the seconds already spent searching, and whatever the optimizer might have found in the seconds remaining. That is the entire cost.

What it does not cost you is anything at all in the plant. Your jobs, your capacity, your recorded work, and every other planner's view of them are untouched, because the optimizer never held a lock on them and never wrote a row. The live schedule during an optimizer run is not "being changed," it is being read. That is why a planner can start a run, think better of it, and cancel with no more consequence than closing a window.

Let It Finish to Get a Proposal

If what you want is a plan to look at, the answer is simply to let the run complete. When it does, you get the verdict sentence, the badge, the before and after numbers, the list of exactly which operations move, and the safety strip confirming that nothing has been saved.

Then the normal review applies. Read the move list, and press Accept to apply the plan or Discard to keep the one you have. Every proposal you are shown has already been measured against your current schedule, so it can only be an improvement or an honest report that your plan is already the best of those tried, which is covered in when the optimizer says your plan is already the best. The planner always presses Accept: the optimizer proposes, and never applies on its own.

Choosing a Budget You Can Wait Out

Because a canceled run yields nothing, the practical advice is to pick a budget you are willing to sit through. With 10, 30, and 60 seconds on offer, that is a low bar, and 30 seconds is the sensible default for everyday use.

Reach for 60 when the plan matters and you want the hardest search available, typically the considered planning run at the start of a shift or a week. Reach for 10 when you are trying goals quickly and comparing what each one favors. In all three cases you are setting the point at which the optimizer must report, and in all three cases it may report sooner if it has proven there is nothing left to gain.

The Bottom Line

Cancel stops an optimizer run harmlessly and discards it: the status reads "Stopped, nothing was changed," and no proposal, badge, or gap survives. The safety is real and unconditional, because the optimizer writes nothing before Accept, but there is no partial plan to collect on the way out. The time budget genuinely is a ceiling rather than a fixed duration, so a run can finish early on a proof, and a completed run is the only thing that produces something to review. Pick a budget you can wait out, read the comparison when it lands, and keep the final call where it belongs. Explore the platform at EDGEBIC.