Skip to content
allanninal.dev/build
Collaborate

Part 3 of 7 · Housekeeping dispatcher series ~5 min read

Why the route matters as much as the order

Housekeeping software tends to optimise the order of rooms and ignore the distance between them, which produces lists that look excellent and take twenty per cent longer to work through.

Key takeaways

  • A floor change costs several minutes each time, and they add up fast.
  • Allocate contiguous blocks and order within them.
  • Trading a little priority for a lot of route is almost always right.
  • The trolley is a constraint: restocking is a journey too.
  • Measure the walking by counting floor changes; it is the easiest proxy.

What the walking costs

One housekeeper's morning under strict priority and route-first orderingA stacked bar chart with two bars in minutes. Two series: minutes cleaning in green, and minutes moving between rooms in red. Strict priority: four hundred and two minutes cleaning and eighty-eight moving. Route first: four hundred and two cleaning and thirty-one moving. A note says same rooms, same work, and fifty-seven minutes of one person's morning.0200400600800~490Strict priority~433Route firstMinutes cleaningMinutes moving between roomsSame rooms, same work. Fifty-seven minutes of one person's morning.
Fig 1. One housekeeper’s morning under two orderings. The cleaning time is identical; the difference is entirely movement.

Across a team of six that is most of a person-day a week, and it appears nowhere: the rooms all got cleaned, everybody worked their hours, and the difference is absorbed as the day running late.

Floor changes as the measure

Counting floor changes is a good enough proxy for the walking and it needs no floor plan, no distances and no modelling. A list that involves three floor changes is considerably better than one involving nine, and that comparison is available immediately.

Within a floor, ordering by room number is usually close enough. Corridors are linear, room numbers follow them, and the marginal gain from anything cleverer is small compared to the effort of maintaining a floor plan.

Blocks, then order

How rooms are grouped into blocks and ordered within themA vertical chain of five steps entered by a box labelled The day's rooms with required-by times. Step one groups by floor, and by wing where relevant. Step two allocates whole blocks, one person to one area. Step three orders within the block by required-by time. Step four asks whether any room is at risk of not being done in time; if so it exits to Move that one room rather than the whole plan. Step five produces a list per person, contiguous and timed. A note says priority moves individual rooms and does not reorder the whole day.AWS ACCOUNTThe day's roomswith required-by timesGroup by floorand wing where relevantAllocate whole blocksone person, one areaOrder within the blockby required-by timeAny room at risk?cannot be done in timeMove that one roomnot the whole planyesA list per personcontiguous, timedPriority moves individual rooms; it does not reorder the whole day.
Fig 2. How the two constraints are combined. Priority acts as an exception mechanism rather than as the primary sort, which is the whole trick.
  • Compute
  • App integration
  • Management
  • Analytics
  • Front-end & mobile
  • People

How much priority to trade

The practical rule is that a room should only break the route if it would otherwise miss its required-by time. A room needed by one o’clock that would be reached at half past twelve on the route order does not need to be pulled forward, and pulling it forward costs a floor change for no benefit.

That single rule removes most of the conflict between the two constraints, because most of the time the route order is already good enough for the deadlines.

The trolley

How trolley capacity creates additional journeys during a shiftA horizontal row of five boxes. A trolley holds about twelve rooms. Then a restock: eight minutes, to the store. Where is the store: one floor, usually. Plan the break, not mid-corridor. Or a second trolley, cheaper than the walking. A note says two restocks a shift at eight minutes each is another sixteen minutes per person.THE CONSTRAINT NOBODY MODELSA trolley holdsabout 12 roomsThen a restock8 minutes, to the storeWhere is the store?one floor, usuallyPlan the breaknot mid-corridorOr a second trolleycheaper than the walkingTwo restocks a shift at eight minutes each is another sixteen minutes per person.
Fig 3. The trolley capacity constraint. It is invisible in every allocation system and it produces a predictable journey a couple of times a shift.
  • Compute
  • Machine learning
  • Management
  • Front-end & mobile

Knowing roughly how many rooms a trolley covers lets the list be broken at a sensible point — at the end of a corridor, near the store, between blocks — rather than wherever the supplies happen to run out.

It also quantifies an easy decision. If restocking costs sixteen minutes per person per shift across six people, a second trolley or a satellite store on another floor pays for itself quickly, and that is an argument with a number in it.

What not to build

A full routing optimiser with floor plans, distances and travel-time matrices is available and is not worth it. The gain over floor blocks and room number ordering is small, and the maintenance cost of keeping a floor plan current through refurbishments is real.

Next: sharing the work out.

All posts