Cleaning in Food Retail: What HACCP Really Asks of the Floor

A supermarket floor is not one cleaning problem, it is four, and they arrive on different schedules. The entrance and the area a few metres behind it carry almost everything that comes in from outside: grit, sand, road salt in winter, and meltwater that turns the same dirt into a film. The sales floor itself is mostly dry soiling — dust, trolley tracks, scuff marks from shoes and pallet trucks — spread thinly over a large area. The checkout zone concentrates more footfall per square metre than anywhere else in the building. And the fresh-food areas are their own category: produce, the deli and meat counters, and the bakery station produce organic, wet and greasy soiling right where hygiene expectations are highest. Our supermarket page sorts the same building by zone.

On top of that sits the part no schedule can hold: the unplanned spill. A dropped jar, a leaking pack, a burst bottle. These do not wait for the night shift, they need a response within minutes because they are a slip hazard and, in the fresh-food areas, a hygiene event. So retail cleaning splits cleanly into two jobs that are usually treated as one — the plannable full-area clean, and the immediate local response during opening hours. They call for different answers, and confusing them is why so many robot deployments in retail disappoint.

The time windows make it harder. Trading hours in German food retail are long, and the night is not empty either: shelf-filling teams, deliveries and roll cages occupy the aisles for a good part of it. The window in which a floor is genuinely free is usually short and sits either side of the shelf-filling work. Who cleans varies as well — many chains buy in a contractor for the early-morning full clean while store staff handle everything that happens during the day. Both sides are under the same pressure, which our article on the staffing shortage in commercial cleaning describes in detail.

The relevant instrument is Regulation (EC) No 852/2004 of the European Parliament and of the Council of 29 April 2004 on the hygiene of foodstuffs. Two of its provisions matter for the floor. Article 4(2) obliges food business operators after the primary-production stages to comply with the general hygiene requirements in Annex II. Annex II, Chapter I — headed as general requirements for food premises other than those in Chapter III — opens with one sentence that does most of the work: "Food premises are to be kept clean and maintained in good repair and condition." A supermarket is a food premises, and the sales floor is part of it.

The more specific floor requirement is narrower than it is usually quoted. Annex II, Chapter II applies to rooms where foodstuffs are prepared, treated or processed, expressly excluding dining areas and the premises of Chapter III. Its point 1(a) states that "floor surfaces are to be maintained in a sound condition and be easy to clean and, where necessary, to disinfect", requiring impervious, non-absorbent, washable and non-toxic materials unless the operator satisfies the competent authority that other materials are appropriate. In a supermarket that hits the bakery station, the counter preparation rooms and the produce prep area — not automatically every square metre of the sales floor, which is governed by the general duty in Chapter I.

Then comes the part everyone abbreviates to "HACCP". Article 5(1) requires food business operators to put in place, implement and maintain a permanent procedure or procedures based on the HACCP principles; Article 5(3) confines that duty to stages after primary production, which places food retail squarely inside it. Among the principles listed in Article 5(2), subparagraph (g) is the documentation one: establishing documents and records commensurate with the nature and size of the food business, to demonstrate the effective application of the measures. Note what that sentence does and does not say. It requires records proportionate to your business that show your own measures are being applied. It does not prescribe a floor-cleaning log, a format, a frequency, or a technology.

Where does cleaning sit in that structure? The Commission notice on the implementation of food safety management systems (OJ C 278, 30 July 2016) answers it: cleaning and disinfection is one of the prerequisite programmes in its Annex I, listed at section 2.2, and prerequisite programmes are described as the foundation on which effective implementation of the HACCP principles rests. That is the honest position of floor cleaning in a hygiene concept — foundational, documented as part of the prerequisite layer, and not a critical control point in its own right. Retail chains layer their own audit regimes on top of the legal minimum, and those private standards can demand considerably more documentation than the regulation does. We do not quote requirements from those schemes here, because their texts are not publicly verifiable in the way EUR-Lex is. Everything in this section is orientation from the published legal text, not legal advice — for your own hygiene concept, the responsible authority and your own advisers decide.

Two machines from our fleet cover the retail formats we see most. The CC1 is the compact one: Pudu states 700–1,000 m²/h, operating noise below 70 dB(A), maximum suction of 17,000 Pa, 15 litres each of fresh and waste water, and per-mode runtimes of 5 hours scrubbing, 5 hours sweeping plus vacuuming plus mopping, 4 hours carpet vacuuming and 9 hours silent mopping, all declared as laboratory values at the lowest gear setting. Its manual specifies a minimum path clearance of 70 cm, which is what makes it viable between shelving. It is the only one of the two with official dry modes, and that matters at the entrance, where dry grit is the actual problem — but switching modes means changing components by hand, not pressing a button.

The L3 is the wider-throughput compact scrubber: 400 mm scrubbing width with dual disc brushes at 18 kg contact pressure, 25 litres per tank, 700 mm minimum passable width, a 3D LiDAR and an Nvidia chip rated at 100 TOPS. On the LiDAR, CenoBots' own documents disagree: the official product page and the L3 brochure state a 96-beam unit, the 2026 full-line catalogue states 128 beams — most plausibly a model-year change, and worth asking on delivery which unit is fitted. CenoBots advertises up to 2,016 m²/h and labels the row "Max. Theoretical Productivity" in its own datasheets — that figure is working width times top speed times 3,600, with no turning, overlap or refill stops in it, so do not plan a store from it. Retail and grocery stores are among the application segments the manufacturer itself names for this machine. Its WS3 workstation handles charging, water refilling, drainage and detergent filling in one unit, which is what turns a four-hour machine into a nightly routine that nobody has to attend. How the compact class compares in detail is the subject of our comparison of compact scrubber robots.

One floor-type warning belongs here, because it is the most expensive detail to get wrong in retail. Pudu's official application scope for the CC1 lists terrazzo, marble, floor tiles, epoxy resin, sandstone and artificial stone as hard floors, plus short-pile carpet in vacuum mode only. PVC, laminate and wood flooring do not appear in any official Pudu material — though that list ends with an open "etc.", so their absence is weaker evidence than it reads. A great many German stores have exactly those floors, so the floor covering is the first thing to establish on site — before throughput, before area, before anything else.

And the honest boundary of the whole exercise: these machines clean, they do not disinfect. Chapter II of Annex II speaks of floors that must be easy to clean and, where necessary, to disinfect — the disinfection step is a separate, deliberately controlled operation and we make no claim that a robot performs it. What else stays manual is predictable: edges and shelf plinths that no round brush reaches, the inside of counters, sanitary facilities, and broken glass, which must be swept up by hand first because shards damage brushes. The robot can re-mop the area afterwards.

Every autonomous machine of this class logs that it drove, when, and over which map. That is an execution record, and it answers a real question — was the planned clean actually carried out on the night in question. It does not answer the question an auditor is more likely to ask, which is whether the floor was clean afterwards. The distinction between those two records is the whole subject of our article on cleaning verification, and it is worth internalising before anyone quotes you a reporting feature.

One machine in the Pudu line closes that gap explicitly. The CC1 Pro carries what Pudu describes as the world's first rear-facing AI camera for real-time monitoring of cleaning performance: it inspects the floor it has just cleaned, identifies remaining spots and produces heatmaps after each task. Pudu's official German page describes two heatmap types — one showing heavily soiled areas for cleaning planning, one flagging stubborn stains that need manual inspection — alongside operational displays for task completion and maintenance frequency. That is the difference between "the route was driven" and "the result was checked", and it is the reason the Pro exists at all; the full comparison sits in our CC1 versus CC1 Pro article.

Now the part that has to be said plainly, because the market says the opposite. A robot report is supporting documentation for the hygiene concept you have written. It is not a hygiene concept, it does not create one, and no authority certifies a machine as HACCP-compliant. The regulation asks for records commensurate with your business that demonstrate your measures are applied; a coverage heatmap can be one of those records for the floor areas the machine actually drives. It says nothing about the areas it does not drive, nothing about disinfection, and nothing about the rest of your prerequisite programmes. Anyone selling you a machine on the promise that it satisfies a legal duty is describing a product that does not exist.

The full-area clean belongs in the closed store, for the reason that governs every autonomous deployment: stability. These robots drive a learned map and correct it live against their sensors, so every obstacle that was not there yesterday costs a detour, a re-plan and, in the end, an uncleaned patch. In retail that obstacle has a name — the roll cage. A night in a supermarket is not automatically an empty floor, which is why the run has to be scheduled around the shelf-filling work rather than merely "at night". The general requirements for unattended operation, from charging logic to alarm systems, are covered in our article on night operation.

Daytime operation is possible and is a placement decision, not a decibel decision. Pudu documents the CC1 at below 70 dB(A) in cleaning operation. That is a real value and it is scrubber-typical, but it is not conversation-quiet — it sits in the range of a conventional commercial vacuum cleaner, and any claim that it is quieter than a normal conversation is wrong. Practically, a scrubber running in an open sales area during trading hours is unproblematic where customers are moving anyway and problematic where they linger. The second constraint is the floor itself: scrubbing leaves a thin drying film, so route planning and signage matter more during opening hours than the noise figure does.

For the unplanned spill, the honest answer is that a full-coverage machine is the wrong tool — running an entire sales floor to deal with two square metres of spilled juice makes no sense. This is precisely the gap the CC1 Pro's AI spot cleaning addresses: Pudu states 1,500–3,000 m²/h for that mode, in which the machine patrols, recognises common soils such as coffee and sauces by image recognition, and cleans where it finds them, giving a stated coverage of 5,000 to 8,000 m² per run. One caveat belongs beside it, and we repeat it wherever the Pro appears: on noise, Pudu contradicts itself. The May 2025 product deck says below 70 dB, the current official German product page says below 75 dB. Never treat the Pro as the quieter machine.

Retail is not one vertical for cleaning purposes, and the difference is not size. The supermarket carries the food hygiene layer described above, plus fresh-food areas, plus the longest trading hours — it is the format where documentation matters most and where the split between night full-clean and daytime spot response is sharpest. The specialty store drops the food hygiene layer almost entirely: mostly dry sales floors, display aisles, an entrance zone, and a cleaning job that is about appearance and dust rather than hygiene evidence. The same compact machines fit both, for different reasons.

The DIY store is where the pattern breaks, and it is worth being direct about it. The dominant soiling there is coarse and dry — screws, sand, wood chips, packaging film, cardboard — and a compact wet scrubber is not the right first machine for it. Debris of that kind belongs to a dry sweeping pass, with wet cleaning as a second stage on the smooth floor areas if at all. Recommending the same machine into a DIY store that works in a supermarket is the most common mismatch in this segment.

What decides the economics, in every one of the three, is not the machine price. It is the share of total floor area a machine can actually reach unaided, the length of the uninterrupted window it gets, whether there is a water connection or a station near the route, and how much edge and detail work stays with people afterwards. Those factors, and how they interact, are the subject of our article on what cleaning robots cost; whether a scrubber is the right class for your floor at all is settled first in the scrubber-dryer guide.