Applying Changes

This is the page most easily skipped and the one where value is most often lost. An optimization result is a plan. Nothing improves until it is executed in your own systems and with your own people.

Inside the platform

Adopting a result updates the regions the platform uses, so that:

• monitoring and alerts are evaluated against the new structure
• future optimization runs start from it
• reports reflect it

If you want to keep the old structure available for comparison, save it as a configuration in the Preview sandbox first.

For an API-connected team, remember that metric values continue to come from your source system on every sync. What you have changed is which region each area belongs to, not the numbers within it.

Outside the platform

1. Update your scheduling system

The territory definitions in your scheduling or HIMS platform have to change to match. Until they do, your source system and Optimal Workforce disagree about your structure, and every subsequent sync will keep telling you so.

2. Reassign clients and staff

Work through the movement table. For each area that moves, decide which specific clients transfer and which worker picks them up. This is where the judgment that the algorithm does not have gets applied:

• language and cultural fit
• required certifications and skills
• continuity for clients where a change of worker is genuinely harmful
• existing relationships worth protecting

It is entirely reasonable to exclude specific clients from a move. A plan that is 95% executed and works is better than one that is 100% executed and does not.

3. Tell people, in the right order

Staff first. Supervisors before their teams; the people whose patch is changing before anyone else. Explain what is changing, why, and when. People accept a boundary change they understand and resist one that arrives as a schedule surprise.

Clients next, where their worker is changing. Give notice. A handover conversation between the outgoing and incoming worker is worth the time it costs.

Then everyone else — scheduling, payroll, anyone whose work depends on who covers what.

4. Phase it if it is large

A change touching a hundred areas does not have to happen on one Monday.

• Start with the regions that were worst off — that is where the benefit is.
• Move a region at a time, or a set of adjacent areas at a time.
• Leave a gap between phases to see the effect and to fix what surprises you.
• Avoid holidays, known seasonal peaks, and any other change your organization is already absorbing.

5. Watch what actually happens

Set a review date before you start, and hold it.

• Week one — travel time and schedule adherence. This is where problems show first.
• Month one — the metrics you optimized for. Did the projected improvement materialise?
• Month three — overtime, missed visits, turnover. The outcomes you actually cared about.

If reality does not match the projection, that is information. Usually it means the model was missing a constraint that matters. Feed it back in — as a free-text constraint, a different weighting, or a corrected region structure — and it will be closer next time.

Documenting it

Generate a report at each stage:

• Before — a balance report, so you have a dated baseline.
• At decision time — an optimization report on the run you adopted.
• After — a trends report showing what actually changed.

Together those three make the case for the next round far easier to make. See Reports.

Common ways this goes wrong

Adopting in the platform and stopping there. The dashboard now shows a balanced structure that does not exist in reality. Nothing has improved and your monitoring has stopped telling you the truth.

Executing everything at once. Large simultaneous change produces enough disruption to swamp the benefit, and you cannot tell which part caused what.

Skipping the supervisor conversation. The people running the regions know what will break. Ask them before you commit, not after.

Not measuring the outcome. Without a before-and-after you cannot tell whether it worked, and the next rebalancing proposal has no evidence behind it.