How to Explain an Optimizer Proposal in EDGEBIC

The Explain dialog in EDGEBIC by User Solutions turns an optimizer proposal into a written account: the goal you chose, how the search ran, the outcome measure by measure, the moves that matter, and what the optimizer refused to touch. It is the difference between telling a supervisor "the software moved it" and telling him exactly what improved and by how much. This is how to open it, read it, and use it.

The wider optimizer workflow is covered in how to choose an optimizer goal. Every task in this library is mapped on the EDGEBIC how-to hub.

Before You Start

• Open manufacturing orders exist and are scheduled. The optimizer improves a plan that already exists.
• You have run the optimizer at least once, so a proposal is on screen. Explain describes a result, not a possibility.
• You know which engine is selected in Settings, because the language of the badge and the presence of locked jobs both depend on it. See how to choose the optimizer engine.

Step 1: Run and Stop Before You Accept

On the Drive Schedule tab click Optimize Schedule, or open the Optimizer tab directly. Pick a Goal and a Time budget, then click Run.

When the run finishes, the comparison appears and nothing has been saved. The safety strip along the result says so in three fixed checks: nothing saved yet, completed work untouched, and never worse than the current plan. Read before you decide, because the proposal lives only on this screen.

Step 2: Open Explain

Click Explain on the result. The dialog opens under the heading Why this plan? with five sections:

The How we searched section is the one planners underrate. It names the winning ordering, for example due date order with the earliest due date first, which often explains the whole result in five words.

Step 3: Read the Certificate Honestly

The badge on the result and the certificate line inside Explain say how hard the answer is, and the two engines say different things.

• Best of N schedules tried in X seconds. The multi-run search built N complete real schedules and kept the best. That is an honest search with no claim of proof.
• Proven within X% of optimal, or proven optimal, in X seconds. The mathematical solver additionally computed how far any schedule could possibly go, so it can bound the gap.

Both engines carry the same never-worse guarantee. Neither is licensed to say the plan is always optimal, and the screen never does.

Step 4: Read the Locked Jobs

With the mathematical solver selected, Explain also lists the jobs that were locked, with the reason spelled out per job. Typical reasons name a routing capability the solver does not yet rearrange: a multi-instance work center, a one-per-day work center, transit days, FlowStep lot streaming, transfer-batch lot streaming, parallel processing, alternate work centers, or the sequence-dependent setup matrix.

A locked job is not skipped. It is reproduced exactly as the scheduler planned it, same machines and same times, and the other jobs are optimized around it. That is a correctness decision: the optimizer will not guess about a feature it does not fully model. When every job in scope is locked you get the badge that says there is nothing to optimize, and the plan comes back unchanged.

Step 5: Copy the Account

Click Copy. The whole story goes to the clipboard as text, ready to paste into an email or a shift note. This matters more than it looks: the text you copy is the same text stored in the audit record if you press Accept, so the conversation today and the audit lookup in three months quote one source.

Step 6: Accept, or Discard

Accept saves the proposal through the same pipeline as a normal scheduling run, with the same protection for recorded work, refreshes every schedule view, and writes the audit record. Discard drops it and writes nothing.

If the world moved while you were reading, for example a colleague rescheduled or kiosk actuals landed, Accept refuses rather than overwrite. A banner appears saying changes arrived while you reviewed, with a Re-run button that solves again on the latest data. That is never a data-loss situation.

What Changes When You Accept

Only the operations in the move list change position. Completed and in-progress steps pass through unchanged, work-center load re-forms around the new plan, and the Gantt, Job View, Drive Schedule grid, and dashboards all refresh. One audit record is written. On Discard, none of that happens.

How to Check It Worked

Open a job that appears in the move list and click View Audit Trail. The newest event is the optimizer record, and it holds the same explanation you read on screen. Then check the Drive Schedule grid: the measure Explain claimed improved, usually days late, should now read what the Outcome section predicted.

Common Mistakes

• Accepting without reading the move list. The list is deliberately changes-only, so it is a 30-second read. Unmoved operations are never listed.
• Quoting a proven gap from the wrong engine. Only the mathematical solver produces the proof. The multi-run badge counts schedules tried, and saying more than that overstates it.
• Treating a locked job as a fault. It means reproduced exactly, correctness first.
• Assuming Explain reflects the live plan later. It describes one proposal at one moment. After you Accept, the stored audit record is the durable copy.

See how the optimizer fits the rest of the planning day on the EDGEBIC product page.