Hotel Cleaning: What Robots Take Over — and What Stays with Housekeeping

The useful way to think about hotel cleaning is by area, not by shift. The guest rooms are their own world: they follow the departure and stay-over rhythm, they are entered one at a time, and they are cleaned by hand from bed to bathroom. The guest-floor corridors are the second: long, repetitive, mostly carpeted, and in most houses the surface that quietly gets skipped when the shift is short. The public areas are the third — lobby, restaurant and breakfast room, conference and event floors, back-of-house corridors — mostly hard floor and, unlike everything else, actually clearable at defined times.

The clock is the other half of the picture, and a hotel has an unusually awkward one. There is no hour at which the whole building is empty. Reception is staffed around the clock, late arrivals and early departures overlap, and the lobby never really closes — it only gets quiet. What does close: the restaurant after service, the breakfast room between the last plate and the first setup, the conference floors between bookings. The housekeeping window sits in the middle of the day, roughly between the departure deadline and the arrival peak, and that is the only stretch in which the guest floors are both accessible and legitimately noisy.

Who does the work varies more than the work itself. Some houses run housekeeping in-house and buy in the public areas; some contract everything to a service provider; many operate a mix that shifts with occupancy. What barely varies is the failure mode: when a shift is short, the rooms get done because rooms are sold, and the corridor and the lobby wait. That is exactly the pressure a machine can take off, and it is why the honest question is never "can a robot clean a hotel" but "which surface can I hand over so that the people keep the rooms". Our hotel overview sets out the machines we deploy there; the structural background is in our piece on staff shortage in commercial cleaning.

This deserves to be said plainly, because the category markets it badly. Cleaning a hotel room is stripping and making beds, the bathroom in full, surfaces and contact points, glassware, restocking, dealing with a guest's belongings in a stay-over room, and a final walk-through that decides whether the room is sellable. None of that is floor area. The floor is a small and easy part of a room clean, and it is the only part a machine could even address.

The geometry closes the case. Pudu's official manual specifies a minimum passage width of 70 cm for the CC1 and recommends at least 75 cm for a long passage; the machine surmounts 20 mm of height and climbs 8 degrees when it is not cleaning — in cleaning mode the manual caps the gradient at 3 degrees. A hotel bathroom threshold, the gap between bed and wall, a luggage rack, a desk chair, a bathrobe on the floor — a guest room is not a surface, it is an obstacle course with a door that is locked for good reasons. Add the access question, which is not technical: a machine that enters occupied rooms is a conversation with your guests you do not want to have.

So the cost question that gets typed into search engines — what does hotel room cleaning cost — has an answer that no robot changes. It is driven by the ratio of departure to stay-over rooms, room size and bathroom fit-out, the linen-change rhythm, occupancy and arrival peaks, your quality-control regime, and whether you staff it yourself or contract it. A robot moves none of those levers. What it does move is whether the floor work outside the rooms has to come out of the same shift as the rooms — and in a short-staffed house, that is the lever that actually decides whether the corridor gets vacuumed today.

Guest-floor corridors, carpet: the SoftBank Whiz. SoftBank's launch release states it works best on carpets and hard flooring; the data sheet gives 465–557 m²/h at three hours per charge, and the release adds up to 1,500 m² per charge — which lies inside the span three hours at that rate produce. The dust bag holds 4.0 litres, filtration is HEPA at 99.97 % of PM2.5 particles including pollen and dust, and routes are taught by walking the machine once, up to 600 of them. Two practical notes: the machine weighs 30 kg plus a 5 kg battery, so nobody carries it up a floor on the side, and both the start and the battery swap are human work — Whiz Connect monitors, it does not start cleaning runs. In a hotel that is a feature rather than a defect, because housekeeping is on the floor at that hour anyway. The class comparison is in our guide to autonomous carpet cleaning.

Lobby, restaurant and back-of-house, hard floor: the Pudu CC1. Officially 700–1,000 m²/h at a 500 mm cleaning width, 15-litre fresh and waste water tanks, five hours of scrubbing or up to nine hours of silent mopping per Pudu's manual, under three hours to charge, and operating noise below 70 dB(A). The line that matters most in a hotel is the application scope: Pudu lists terrazzo, marble, tiles, epoxy resin, sandstone, artificial stone and low-pile carpet. Parquet, laminate and vinyl are on no official Pudu list, and a wet program on a hotel parquet is a damage claim, not a cleaning result. That is why the floor survey happens before the machine is planned, not after. Carpet, where the CC1 meets it, is vacuumed dry with the carpet attachment Pudu sells separately — never scrubbed.

Small carpeted rooms — lounges, meeting rooms, boardroom-sized floors: the Nexaro NR 1700. Its data sheet gives 49 dB(A) in eco mode and 57 dB(A) in balanced and performance, measured on carpet, and it navigates entirely without a camera, which shortens the works-council and data-protection conversation considerably. The manual limits it to carpets with a free fibre length of up to 15 mm and advises fencing off high-pile carpet — the only published pile-height limit among the vacuum machines we run. Which vacuum belongs where is worked through in our vacuum robot comparison.

And the machines that do not belong in a hotel, which is the shorter and more useful list. The large industrial platforms we assess — the Pudu BG1 series, the CenoBots S5, the Adlatus CR700 and SR1300 — are out on two counts at once: physical size in guest-facing space, and the fact that none of their manufacturers publishes a noise figure. The compact Pudu ET1 is a closer call: it is a small wet scrubber that would suit a budget-format lobby during the day, but its noise level is likewise unpublished, which rules out any placement near guest floors, and its CE documentation was still unpublished when we last checked — details in our ET1 profile. Where a manufacturer publishes no decibel value, we do not supply one and we do not place the machine near sleeping guests.

The rule is short and it is not negotiable: guest floors are day work, public areas are night and off-hours work. It is a placement rule, not a decibel rule. Vacuuming is loud by construction — the working principle is a fan, and a fan makes noise. SoftBank's data sheet gives the Whiz 62 dB in normal mode, which is unremarkable in a corridor at eleven in the morning, when housekeeping would be running an upright vacuum anyway, and which is a complaint mechanism at three in the morning outside a room door regardless of what the sheet says. That is the whole reasoning, and it does not change for a quieter machine.

The night side belongs to the scrubber-dryer, and the areas that genuinely close: the restaurant after service, the breakfast room before setup, conference and event floors between bookings, back-of-house corridors. The lobby is the special case, because it does not close — it gets quiet, usually between the last late arrival and the first early departure, and that quiet window is what the route is planned around rather than through. The second constraint is physical: a scrubber leaves a thin wet film that needs a few minutes to dry, so the last run has to finish before the first guests walk it. A wet lobby floor at seven in the morning is a slip hazard and a complaint, not a clean floor. What else has to be settled for unattended night runs — access, keys, escalation when a machine stops — is in our article on night operation.

The day side is the housekeeping window, and the Whiz fits it because it needs a person anyway: housekeeping sets the route once, starts the machine in the morning block, swaps the battery when it runs down, and works the rooms while the corridor is being covered. Multi-floor houses need one more decision, because the machine does not take the lift by itself. Either a machine stays on its floor and the route is per-floor, or a member of staff moves it, or you use lift integration — which for our hard-floor machines means the CC1 Pro, whose official product page lists lift control and electronic gate control as IoT compatibility. The trade-offs between the two CC1 generations are in our CC1 versus CC1 Pro comparison.

One distinction is worth keeping straight in any quotation, because both properties are sold under the same word. Autonomous is not the same as unattended. Pudu documents that the CC1 returns to its docking station at low battery and resumes the task at the point of interruption — that is a machine capable of getting through a night on its own. The Whiz is the opposite case: it is started at the machine, its battery is changed by hand, and no docking station appears in its data sheet. It is autonomous during a run and not unattended across a night — which is precisely why it belongs in the daytime window and not in the night plan.

The breakfast room and the restaurant are not just floor area; they are a food business, and the EU food hygiene regulation addresses their floors directly. Regulation (EC) 852/2004, Annex II, Chapter I, states that food premises are to be kept clean and maintained in good repair and condition. Annex II, Chapter II, adds for rooms where food is prepared, treated or processed that floor surfaces are to be maintained in a sound condition and be easy to clean and, where necessary, to disinfect. Article 5 requires food business operators to put in place procedures based on HACCP principles, with documentation appropriate to the nature and size of the business.

Now the part that vertical marketing usually skips. None of that prescribes a cleaning technology. There is no legal requirement to use a robot, no legal requirement to keep robot logs, and nothing in the regulation that a machine satisfies on your behalf. What the text does establish is that the condition and cleanability of the floor is a matter someone can and will look at, and that your procedures are supposed to be documented in proportion to your operation. We are describing the regulation, not giving legal advice — how your HACCP concept is built and what your food-safety authority expects of it is a conversation with them and with whoever wrote your concept.

Against that background, what a machine genuinely contributes is a record. Every run produces date, time, area covered and the route as driven, and that turns a hand-kept cleaning schedule into machine-generated data. The CC1 Pro goes one step further with a rear-facing AI camera that inspects the floor just cleaned, triggers targeted re-cleaning, and feeds cleaning heatmaps — the difference between "the route was driven" and "the floor is verifiably clean". For chain standards, franchise audits and service-provider SLAs, that is the useful part.

Two honest limits belong next to it. Such a report documents floor area cleaned to a verified result; it does not document a hygiene status, and it says nothing about the surfaces, contact points and sanitary areas that make up most of a hygiene concept. And these machines clean — they do not disinfect. Where disinfection is required, it stays a manual, specified, human task, and it should be written into the specification next to the robot rather than quietly read into it.

The guest reaction is the part hoteliers worry about most and the part that has given us the least trouble. Our experience across occupied sites is consistently positive: a working robot in a busy building draws attention, and that attention is friendly — people, children especially, respond to the machine. The movements are slow and predictable, which is what keeps it unremarkable. That said, it is worth planning with the reaction rather than against it: a route through the check-in queue at four in the afternoon is a bad route not because the machine cannot handle it, but because it makes a machine the obstacle in a moment that should feel effortless.

On the staff side, the deciding factor is who carries the machine's own upkeep. Filling and emptying water, changing brushes and squeegees, dust bags and filters, keeping the map current when a banquet layout changes, and going to the machine when it stops in front of something it does not understand — every one of those is a working hour. If that hour lands on the same short housekeeping shift the machine was meant to relieve, the deployment moved the problem instead of solving it. That is the entire reason we rent our machines in a full-service model: maintenance, wear parts, remote monitoring, fault handling and map upkeep sit with our service, not with your team.

The rest of the honest list. Edges, corners, skirting, chair and table legs, and the area under low furniture stay manual. Stairs are the boundary of these machines' world, not a task. Spot treatment and periodic deep cleaning of carpet are wet work on a textile covering and remain a scheduled human job. Sanitary areas, glass, surfaces and contact points are not floor area and were never in scope. And the rooms stay with housekeeping. What the machines take is the flat, open, repetitive share, so that scarce hours go to the rest — which is a smaller promise than this category usually makes, and one that survives contact with a real duty roster. If you are working out which machine fits which building, our robot selection guide walks the decision by property type.