Introducing a Cleaning Robot: How a Rollout Actually Runs

Phase one is the site walk and a demonstration on your own floor. That distinction is the whole point: a robot running cleanly on a trade-fair floor proves nothing about a shop floor with a threshold at the entrance and a display island that moves every fortnight. What we do in a demo appointment is free and non-binding — consultation, a look at the actual floor covering, a live run of the machine and a recommendation for your building. What is being decided in this phase is not whether robots work. It is which class of machine your floor and your layout allow at all, and where a charging position could plausibly sit.

Phase two is mapping and route setup, and it is technician work rather than customer work. Our team maps the premises, programs the cleaning routes and briefs your staff on the functions they will actually touch. Mapping is fast relative to cleaning: Nexaro publishes a mapping speed of a maximum of 600 m²/h for the NR 1700, against a maximum cleaning speed of 100 m²/h. On the CC1, maps are not a black box either — the machine’s settings menu contains a dedicated map function for creating new floor plans and editing existing ones, which is what makes later layout changes a service task rather than a return to the factory. Treat the resulting map as a decision document rather than a technical artefact: it fixes which zones are cleaned, which are deliberately excluded, in what order the machine works and where it parks. Those are operational decisions with consequences for your shift plan, and they are far cheaper to argue about while a technician is standing in the room than three weeks later.

Phase three is the supervised trial window, and it exists for exactly one reason: to find the states of the building that a walk-through cannot show. A site walk sees the building on one day, in one configuration. Real buildings have a delivery morning, a pallet that stands in the aisle only on Tuesdays, chairs stacked on tables after an event, a mat that gets moved when it rains. Every one of those is an exception the machine will meet without you there. Finding them under supervision is a conversation. Finding them later is a fault message.

Phase four is unattended operation, and the CC1 manual states its handover condition more precisely than most sales decks manage. Scheduled tasks run automatically at the set time — provided the robot is correctly positioned, has sufficient battery charge and an appropriate water level. Read that sentence as an operating requirement, because that is what it is: three conditions that somebody has to guarantee every single night. Who that somebody is, and what happens when one of the three is not met, is the subject of the service model. If your window is the night shift, our piece on what unattended night operation really requires goes through the building side of it in detail.

Start with the dock, because it is the one piece of infrastructure that has to be permanent. Per the CC1 operating manual, the docking or charging station is optional equipment, and automatic charging together with automatic water refill and drainage is supported only in conjunction with it. Without a dock the machine still cleans; it just needs a person for the things the dock would otherwise do. The dock in turn needs a power socket and a footprint that stays free — a spot that gets repurposed for pallets during the Christmas season is not a dock.

For scrubbers the obvious follow-up question is water, and the answer is better than most buyers expect. Pudu’s official CC1 page states plainly that the docking station combined with the mobile water station requires no plumbing alterations, while supporting automatic charging, water supply and drainage and the dosing of cleaning agent. So a building without a convenient water connection near the cleaning area is not disqualified — it is a configuration question. Which class of machine you need in the first place, and what a scrubber does to your floor, is the subject of our scrubber-dryer guide. One operational note belongs here regardless of dock configuration: a scrubber leaves a thin wet film that needs a few minutes to dry, so the route has to be planned around foot traffic rather than through it.

Then the geometry, and this is where the site walk earns its keep. The CC1 manual specifies a minimum passage width of 70 cm, with at least 75 cm recommended so the robot can travel a long passage smoothly. Thresholds, steps and floor sockets may be at most 8 mm high while the robot is cleaning and 20 mm while it is only driving across; the maximum surmountable gap is 35 mm. The maximum permissible gradient is 8° when not cleaning, and cleaning is specified for level surfaces only — Pudu explicitly advises against cleaning on visible slopes of more than 3°. The manual also names an operating environment of 1 to 40 °C at up to 85 percent relative humidity. None of these are exotic, and every one of them is a hard gate that a floor plan on a screen will not reveal.

The network question has a different answer per machine, and it is worth asking your IT department rather than your cleaning team. The Nexaro NR 1700 needs no customer Wi-Fi at all: it connects over a built-in 2G/LTE-M mobile link, with the mobile costs included in the HUB licences. The trade-off is that the dock position needs mobile coverage. Nexaro also publishes the layout limits openly — a maximum cleaning area of 1,000 m² per job, dependent on floor plan, furnishing, floor covering and soiling, with a maximum distance to the charging station of 35 m and a maximum open area without characteristic objects of 12 × 12 m. The CC1, by contrast, supports 4G, Wi-Fi and Bluetooth per its specification, so it can use your network but does not have to.

A robot changes the shape of the cleaning job before it changes any headcount. The flat, open, repetitive share goes to the machine; edges and corners, sanitary areas, surfaces and touchpoints, glass, exception handling and the judgement call about what is actually dirty stay with people. That is not a marketing framing, it is what the sector’s own data supports: citing the Job-Futuromat of the Federal Employment Agency, the 2025 industry report of the German building cleaners’ trade association puts the automatable share of tasks in commercial cleaning at a minority — for a retail salesperson, by comparison, the report gives 7 of 8 core tasks. We go through the figures and what they mean for staffing in our piece on staff shortage in commercial cleaning.

That redistribution only produces relief if somebody owns the machine’s own upkeep, and this is the question most rollout plans leave implicit. Water goes in, waste water comes out, hoppers and dust containers get emptied, brushes and squeegees wear, filters need handling, and the map needs maintaining when the layout changes. In our full-service model those tasks sit with us: a technician comes at least once a week, checks the robot, cleans brushes and filters, refills cleaning agent and replaces wear parts as needed. In the self-service model, where the machine is bought, they sit with your team — and then they belong in the shift plan by name, not as an assumption. The daily-cycle items in particular need a named person in either model.

The last piece is acceptance, and it is won during the trial window rather than announced at a staff meeting. Teams that were involved while the routes were being set tend to treat the machine as their machine — they know why a zone is excluded, and they report the pallet in the aisle instead of working around it. Our experience in customer-facing deployments has been consistently positive on the public side too: a working robot draws attention in an occupied venue, and the attention is friendly — people, and children in particular, react to the machine. In venue types where the machine runs in front of an audience — a supermarket sales floor as much as an office reception area — that reaction is worth planning for rather than fearing.

The first and most common failure mode is a cluttered floor, and the manuals are blunt about it. Pudu instructs that cables lying on the floor be removed beforehand so the robot does not get tangled, states that the machine is designed for level surfaces and is not suitable for environments with steps, steep gradients or cramped spaces, and advises against running it on heavily greasy floors. None of that is a defect. It is a description of the operating conditions — and a robot that achieves partial coverage because of obstacles looks like a machine problem in the report and is a housekeeping problem in the building. This is the single item most worth fixing before signing anything, because it is also the cheapest.

The second is glass, mirrors and rearranged furniture, which is really one problem: the machine navigates by what it can see, and both of those change what it sees. Pudu documents it explicitly — purely black surfaces such as skirting boards, mirrored surfaces such as walls, or fully transparent environments at a height between 14 and 22 cm can disturb the robot’s LiDAR reflection and cause abnormal movement, in which case the manual recommends consulting technical support and possibly modifying the environment, for example with reflective stickers. Nexaro states the same class of limit from the other direction with its cap of 12 × 12 m of open area without characteristic objects: a map needs features, and moving the features degrades the map. Our explainer on how cleaning robots navigate goes through what the sensors actually see and where the documented limits sit.

The third is runtime expectation, and it usually comes from reading a single number off a spec sheet. Runtime is a mode figure, not a machine property: the CC1 manual gives per-mode runtimes and adds the footnote that they were measured under laboratory conditions in the lowest gear setting, starting at 100 percent charge and running down to 10 percent. A deployment planned on the headline figure and operated in a different mode at a different speed will look like an underperforming machine and is in fact a planning error. Our article on reading battery runtime figures correctly takes the published numbers apart.

The fourth is the wrong machine class for the floor, which is the most expensive mistake because it survives every other fix. Pudu’s official application scope for the CC1 is specific: terrazzo, marble, floor tiles, epoxy resin, sandstone, artificial stone and low-pile carpet. A floor that is not on the list is not a tuning problem. If you are still at the stage of deciding what your building needs, which cleaning robot fits which operation walks that decision through by venue type, and it is a much better place to be than three months into a rollout.

The service model decides who carries the failure modes above, and that is a bigger difference than any spec sheet comparison. In our full-service model, one monthly rate covers the robot, all spare parts, cleaning supplies, weekly service visits, all repairs, software updates, installation and onboarding — and a free swap unit if the machine fails. The point of the swap is not generosity. It is that a cleaning plan built around a machine has to survive that machine breaking, and the only version of that promise which holds is one where the replacement is somebody else’s logistical problem. In the self-service model, where the machine is purchased, spare parts and service bookings are available but the risk stays on your side.

Route and map changes work the same way. When the layout changes — a refit, a new shelving run, a relocated entrance — the map has to follow. In practice that is either a change your team makes in the app or something we handle at the next service visit; either way it is a scheduled task rather than an incident, provided somebody notices the layout changed. Building that noticing into a routine is worth more than most of the technology in the machine.

Which leaves the readiness check. Before you sign anything, you should be able to answer these without guessing. Where does the dock stand, is there a socket there, and will that spot still be free during the busiest week of your year? Does your narrowest regular passage clear 70 cm, and preferably 75? Is every threshold on the route within the machine’s specified limits, and is the route level? For a scrubber: how does water get in and waste water out, and does the plan need the mobile water station? For network: does the machine need your Wi-Fi, or does it bring its own connection — and is there coverage where the dock stands? Who gets in at night, and what does the alarm system do when something moves at three in the morning? Which zones are deliberately excluded from the map, and who decides that? And finally, the one that decides whether the deployment produces relief at all: who empties, fills and checks the machine in the daily cycle, and is that person named in the shift plan?

If those answers exist before the machine arrives, the introduction is an appointment. If they do not, the introduction becomes a series of discoveries — and discoveries happen on the machine’s schedule, not yours.