Office Cleaning: What Unterhaltsreinigung Actually Covers

Unterhaltsreinigung is the recurring cleaning that keeps a building usable during normal operation — the cycle, not the project. In an office it typically covers the floors of circulation and common areas (vacuuming, sweeping, damp mopping), waste and recycling bins, sanitary rooms, kitchenettes and break areas, handprints on glass doors, handrails and light switches, plus clear horizontal surfaces at hand height. What is explicitly not part of it: deep cleaning (periodically stripping and resealing hard floors, wet-extracting carpet), glass and frame cleaning above a defined height, and special cleaning after construction, water or fire damage.

The term is not an invention of the providers. The trade carries it in its own collective agreement: the Bundesinnungsverband des Gebäudereiniger-Handwerks, the German building cleaners’ trade association, divides the sector into nine wage groups plus apprenticeship pay and describes wage group 1 as "Innen- und Unterhaltsreinigung" — interior and maintenance cleaning — and wage group 6, the journeyman group, as "Glas- und Fassadenreinigung/Gesellinnen und Gesellen" — glass and facade cleaning. Both are simultaneously the universally binding sector minimum wages, in force nationwide for all commercial employees in building cleaning since 1 February 2025. The dividing line that so often blurs in a specification is drawn cleanly in the wage structure: the recurring interior business on one side, glass and facade on the other.

For an office building that means the contract you sign governs almost exclusively Unterhaltsreinigung. It describes a cycle (daily, three times, twice weekly), a scope per room type and a time window. Everything that happens rarely in an office — deep-cleaning the carpet, cleaning the windows twice a year — sits beside it and is calculated separately. Comparing both layers in one number compares nothing. Our office page shows how we structure this in practice.

Within Unterhaltsreinigung there are two depths of service, and the German trade names them Sichtreinigung and Vollflächenreinigung. Clean-on-sight means what is visibly soiled and the frequented, visible zones — the rest is skipped or served on a longer cycle. Full-area cleaning means the defined surface completely, regardless of its visible state. Both are practice terms of the trade, not a legally defined standard. Precisely for that reason they belong in the written specification, not in the prose of a quote.

The consequence is easy to underestimate. Two offers with an identical cycle — "floors three times a week" — can mean vacuuming the open-plan aisles and leaving everything under the desks, or vacuuming every reachable square metre including the furniture islands. Neither is wrong; they are different services at different prices. The fix is unglamorous and works: define per room type what is cleaned every time, what is cleaned on sight, and what runs on a longer interval.

What the legislator requires is a standard, not an interval. § 4 (2) of the German Workplaces Ordinance (ArbStättV) states that the employer must ensure "dass Arbeitsstätten den hygienischen Erfordernissen entsprechend gereinigt werden" — that workplaces are cleaned in accordance with hygienic requirements — and that soiling and deposits which may lead to hazards are removed without delay. The ordinance’s annex separately requires that floor, wall and ceiling surfaces be easy and safe to clean. There is no statutory cleaning frequency for offices; it follows from use, occupancy and your risk assessment. This is not legal advice — the interpretation in your specific case belongs with your occupational safety function and your legal department.

An office floor splits into two worlds, and the split decides everything else. The first is connected circulation area: corridors, open-plan aisles, reception, lobby, canteen, meeting rooms, stair landings. Same surface, same path, every evening — routine in the literal sense. The second is furnished islands: under desks, between pedestals, around cable trays and server cabinets. Per square metre, the second group costs disproportionately more time, and it is the part that gets quietly dropped first when a shift runs short.

On top of that comes the floor covering. Carpet tile in the work areas, hard floor in lobby, corridors, canteen and sanitary rooms — in many buildings both lie on the same storey. That is the real reason a single machine rarely serves a whole office: carpet is vacuumed dry, hard floor is scrubbed wet. A vacuum robot can drive across hard floor, but it does not scrub it, and a scrubber-dryer has no business on carpet. What autonomous carpet care can and cannot do — dry vacuuming on a cycle, with wet extraction staying periodic manual work — is set out in our article on autonomous carpet cleaning.

And then there is everything that is not floor at all: sanitary rooms, kitchenettes, bins, hand-height surfaces, glass. Robots clean floors. They do not wipe desks, they do not empty bins — and cleaning is not disinfection. We do not present any of these machines as a disinfection procedure; where a surface must be disinfected, that is done by people following a defined method. That is also why the honest framing of automation in this trade is not replacement but relief: the machine takes the connected floor so the shift has time left for the parts where a person is genuinely needed. The structural background — why office cleaning shifts are hard to staff in the first place — is in our piece on the staffing shortage in building cleaning.

The Nexaro NR 1700 is our first choice in an office, and the manufacturer targets this environment itself: the official datasheet describes it as flexibly deployable in commercial use "u.a. in Lobbies, Showrooms und Großraumbüros" — among others in lobbies, showrooms and open-plan offices. It is small and flat enough to work between desk clusters and through meeting rooms and reception, it returns to its charging station on its own, and it navigates without any imaging sensor. What it actually achieves per hour and per charge in our deployments — and where the datasheet is more optimistic than the field — is documented in our field report on the NR 1700. Plan it by area per night, not by square metres per hour.

The SoftBank Whiz is the area machine for carpet, and its operating concept fits an office better than it fits most other verticals. SoftBank Robotics states officially that it works best on carpets and hard flooring, that it covers up to 1,500 m² in 3 hours per charge — the four hours quoted alongside are the charging time, not a second runtime figure — and that it can record up to 600 cleaning routes: cleaning staff drive the route once by hand on the first day, and from then on the robot repeats it autonomously. Offices are exactly the environment where that pays off — the layout is stable, the routes stay valid, and the official product page lists workplaces among its target environments. One operational caveat that matters for scheduling: the Whiz is started by a person at the machine, not remotely.

For the hard-floor half of a mixed building, the Pudu CC1 is the counterpart. Pudu lists office buildings among its official application areas; the machine sweeps, scrubs, vacuums and mops in one device, carries a 15-litre fresh and a 15-litre waste water tank, is officially rated at 700–1,000 m²/h and specified at under 70 dB(A). Two things to plan for: a scrubber leaves a thin, briefly drying film, so it runs before people arrive rather than in front of them, and it belongs on lobby, corridor, canteen and kitchen hard floor — not on the carpet where the vacuum robots work. Which class fits which building at all is the subject of our guide to choosing a cleaning robot and, for this class specifically, the scrubber-dryer guide.

Just as important is what does not belong in an office. The heavy industrial classes — large sweeper-scrubber combinations and self-emptying industrial sweepers built for warehouses, production halls and car parks — are a poor fit here regardless of how impressive the throughput looks: mass, turning circles, dock infrastructure and unpublished noise figures all argue against them on a carpeted floor with people at desks. Two machines sit at the edges of the office case, and both are available through us but newly onboarded. Pudu’s compact ET1 is interesting for small offices; trade reports describe camera-based VSLAM navigation, while the official product pages do not specify the sensors at all. If it holds, it brings exactly the data-protection conversation the NR 1700 avoids — our ET1 profile covers it. The Adlatus CR700 — developed and final-assembled in Ulm — lists office buildings among its official target industries and navigates camera-free, but needs a plumbed docking station, which turns it into an installation project; our assessment of it is documentation-based and is written up in the CR700 market overview.

An office differs from a supermarket in one respect that decides procurement: the floor being cleaned is a workplace, and the people on it are employees with a works council. § 87 (1) no. 6 of the German Works Constitution Act (BetrVG) gives the works council a co-determination right over the "Einführung und Anwendung von technischen Einrichtungen, die dazu bestimmt sind, das Verhalten oder die Leistung der Arbeitnehmer zu überwachen" — the introduction and use of technical devices intended to monitor employee behaviour or performance. Whether a particular cleaning machine falls under that provision is a question of the individual case and not one we decide; we do not give legal advice. The practical point is a different one, and it is the one that moves projects: a machine with no imaging sensor removes the subject of the discussion.

This is where sourcing discipline matters, so here is exactly what we can and cannot say. Nexaro’s official datasheet lists the NR 1700’s sensors as a long-range laser distance sensor (LR-LDS), a bumper sensor, eight cliff sensors, two magnetic-field sensors, a wall-following sensor, incremental sensors and an inertial measurement unit — no imaging device among them. Nexaro does not publish an explicit "without camera" sentence anywhere, and the list itself ends open-endedly with "und weitere". So we ground the claim in the published sensor architecture and in our own fleet record, not in a manufacturer promise. That is a weaker-sounding but far more defensible statement than the marketing line, and it is the one that survives a works-council meeting.

Not every machine in our fleet is camera-free, and we say so. The Whiz lists a LIDAR sensor and a 2D/3D camera in its official datasheet; camera-based perception is the norm in this machine class, and it is not per se a problem — it is a question you settle in writing before signature: which sensors, what is stored, where, for how long, who is the processor, and what the works agreement says. The questions worth putting on paper, machine class by machine class, are collected in our article on cleaning robots and data protection. For a building where the works council co-decides, the camera-free machine is not a nice-to-have — it is often what makes the project happen at all.

The German status quo is well documented by the sector itself. In its 2025 industry report, the Bundesinnungsverband writes that, unlike in the Scandinavian countries, German operators frequently organise their work so that most people notice little of the service — which is why cleaning is often done by part-time staff in the early morning or late evening hours, and why the part-time ratio is correspondingly high. The same report records a countertrend: the pandemic triggered a first rethink towards Tagesreinigung, daytime cleaning, in schools and offices, and the sector frames it as an opportunity for both hygiene and recruiting, since considerably more people would be interested in a building-cleaning job if contiguous, family-friendly daytime hours were available. The report closes that section with the sentence that matters for planning: in the end the customers decide the extent and the timing.

For machines, the office night is the good case — and it is worth being precise about why. Vacuum robots are loud; that is why we never place them near sleeping guests at night, and why a hotel corridor at 2 a.m. is the wrong deployment. In an office nobody sleeps, so the empty building at night is exactly the window where a noisy machine is harmless and the map stays stable: no chairs moved between two runs, no people to drive around, no interruptions. The NR 1700 fits that pattern directly since it recharges autonomously and runs to schedule; the Whiz, which a person has to start at the machine, fits the late-evening window at the end of a shift better than the middle of the night. What unattended operation genuinely requires of a building — alarm systems and motion detectors, doors, lifts, and a fault escalation path — is worked through in our article on robots in night operation.

Which brings us to cost, and to the reason we publish no price. What drives the figure in an office is a small set of factors, and none of them is the machine on its own: the depth of service (clean-on-sight versus full-area), the cycle, the room-type mix and above all the share of sanitary rooms, the furnishing density, how large the connected sections of floor are, the floor covering split, and the time window — the sector’s framework collective agreement regulates night, Sunday, public-holiday and hardship premiums, so the same work carries different labour costs at different hours. The benchmark you compare against is the status quo, and the honest reference points for that, including the verified sector minimum wage, are laid out in our article on what a cleaning robot costs. If you want the calculation done on your own floor plan rather than in the abstract, that is what a site assessment is for.