The short answer
Most cleaning rotas break on constraints: a site that can only be accessed between certain hours, a key that one person holds, a task needing a certification, a rule that nobody works alone after a time.
Encode the constraints first. A schedule that respects them and is slightly less efficient beats an optimal one that cannot be worked.
The constraints that actually bind
- Access windows, which differ per site and sometimes per day
- Key holding and alarm codes, held by specific people
- Certifications for specialist tasks
- Lone working rules, particularly out of hours
- Travel time between sites, on real routes
- Contracted hours and rest requirements
Travel time is the one most often estimated badly. Straight-line distance between two postcodes bears little relation to a van in traffic at seven in the morning.
Prediction feeds the optimiser
| Prediction | Used by the schedule |
|---|---|
| How long this site actually takes | Slot length |
| Travel time on this route at this hour | Sequencing |
| Likelihood of absence | Cover requirements |
| Whether access will be available | Ordering and fallbacks |
The first row is the largest single improvement in most operations. Specification times and actual times differ consistently per site, and using actuals fixes a rota that always runs late.
Build in the fallback
Schedules meet reality: someone is off, a site is locked, a job takes twice as long. A system that produces a perfect plan and no way to adjust it gets abandoned within a fortnight.
- Make manual adjustment easy and quick.
- Show the consequence of a change, such as which site now runs late.
- Keep a list of who can cover which site, with the constraints.
- Record why a change was made, so patterns become visible.
- Re-optimise from the current position rather than from the original plan.
Respect the people constraints
Staff have preferences, caring responsibilities and travel limits. A schedule that ignores them produces turnover, and turnover in this sector is expensive and already high.
Capturing preferences as soft constraints, honoured where possible, tends to improve retention more than small pay differences do.