The corridor looks good. It is also the one area the site supervisor will certainly have gone over again before an announced visit. Anyone who gives fourteen days' notice of an inspection and then walks the route from the main entrance to the meeting room is inspecting the contractor's preparation. About the service itself they learn nothing.
That is why so many quality meetings go round in circles. Both sides have impressions, neither side has data, and in the end the winner is whoever is louder or has been there longer.
Two questions that get conflated
Is it clean, and was what was owed delivered? These are different questions, and the usual inspection answers only the first.
A room can be clean without anyone having been in it that day, because it is barely used. It can be cleaned in accordance with the contract and still look unsatisfactory, because the method does not suit the surface or because the interval was set too low. And it can appear clean while lacking precisely the services you cannot see, the sanitary deep clean or wiping beneath the furniture, for instance.
Robust performance monitoring therefore separates the observed result from the contractual check and brings the two together only at the end.
The sample decides everything that follows
Whoever inspects also selects. And the selection determines the outcome more than the standard applied.
Three rules turn a walk-through into a sample.
First: draw only rooms that ought to have been cleaned. The room schedule shows which days of the week a room is serviced. Complaining on a Thursday about a room cleaned on Mondays and Wednesdays establishes nothing. What is usable are rooms that fell due today or yesterday.
Second: draw by a rule and keep the selection concealed until the start. The two belong together, and the two are frequently confused. A list known in advance leads to a prepared condition; that is not dishonourable, anyone would do it, and it markedly reduces the informative value. But a selection the inspector makes spontaneously on site is not thereby unbiased. People walk the short routes, enter the known problem rooms and avoid the second floor. It becomes robust only with a documented random or selection rule over the rooms falling due, stratified by room group, drawn shortly before the start and disclosed to the contractor only on site.
Part of this is what has happened since the cleaning. A room where an event took place last night is no evidence today of poor cleaning. Access restrictions, building works, weather and exceptional use have to be recorded before the assessment, otherwise user behaviour gets counted as a defect.
Third: do not set the size by feel. The basis used in the sector is acceptance sampling by attributes, set out in DIN ISO 2859-1. It provides tables from which, given a lot size and an inspection level, the sample size and the permissible number of defective units follow, and it recommends the general inspection level unless otherwise agreed.
That does not yet give you a cleaning inspection procedure, though. The standard comes from goods inspection, and transferring it to buildings is a methodological adaptation in its own right, one that has to be described in the contract. At minimum the following need deciding: what constitutes the lot, whether the unit is the room or the combination of room and service type, what counts as a defective unit, which defect classes exist, which AQL value and which inspection level apply, how the draw is made, and which edition of the standard is to be used. That last point is not a formality: internationally a new ISO edition appeared in 2026, while in Germany the 2014 version continues to be current. Fail to name the edition and you will argue about it later.
One basic condition is almost always passed over: a lot should consist of comparable units. A room schedule containing offices, sanitary areas, corridors, kitchens and plant rooms is not a homogeneous population. Without stratification by room group you are adding unlike to unlike.
Then the most common application error. An AQL plan answers a yes/no question: is the lot accepted or rejected. The AQL value itself is admittedly a percentage figure, namely the still acceptable proportion of defective units. The defect rate observed in a sample is not, however, a quality grade for the whole building, and it does not serve as a school mark. Anyone wanting to work with a graded percentage needs an assessment procedure that recognises gradations by service type and weights them across room groups. Both are possible and both are defensible. You simply have to choose one and attach the deduction rule to the system you chose, rather than deriving a percentage from an acceptance plan that does not yield one.
Whatever tool is used, one rule applies, and many systems fail at it: the population, the interval status, the selection rule and the assessment standard of every inspection must be fixed immutably. If the room schedule changes later, that must not alter a completed inspection retrospectively. Without that fixing, two inspections from different years are not comparable, and with that the trend disappears, which is the thing that actually matters.
What the ancillary rooms show
A floor area can be tidied up before a visit. A cleaning store can too, but far less successfully, and what a trained eye sees there concerns things that do not change overnight.
What gets checked there is quickly listed. The condition of the store itself. Whether the machines in use are looked after, whether the pads are clean and the right ones, whether mop covers are actually laundered rather than merely rinsed. Whether current safety data sheets are available for every product held in the building. Whether hazardous substances are stored and labelled as required. And whether the four-colour system for cleaning cloths is observed, separating sanitary areas, sanitary ware, general surfaces and kitchen areas.
The colour system is common industry practice rather than a standard, but it is a reliable indicator. Where all the cloths are the same colour, or where only one colour is left in stock, nobody is working to it, whatever the operating instructions may say.
The second place worth looking is purchasing, and here caution is called for. Order quantities over the contract term are hard to fabricate after the event, which makes them a serviceable plausibility check. Proof they are not. Material may come from existing stock, be purchased centrally for several buildings, be sourced through a subcontractor or be misallocated in the accounts. And on factory-finished floors a deep cleaner may not be envisaged at all. Anyone finding conspicuous quantities has a reason to ask, not yet a result.
The contractual side
Alongside the observed cleanliness stands the question of whether the contractual obligations you cannot see have been met.
One question decides half the inspection: what does the contract actually owe? Output-based contracts owe service types, intervals and a quality level, and leave it to the contractor how to organise them. Time- or staffing-based contracts owe hours, attendance times or a minimum staffing level. In the first case a tenderer's internal costing is not an owed staffing level, and an inspection counting hours there is checking against the wrong thing.
Once that is settled, six areas cover the essentials.
Scope and times. Are agreed cleaning times being kept, where they are agreed? In buildings with restricted access windows in particular, this shifts over the years without anyone objecting.
Personnel and qualification. Does the staffing match what the contract requires? Is briefing carried out, and is it documented? Turnover counts here too, because it consumes the knowledge of the building the client is paying for.
Products and equipment. Are the products named in the tender, or equivalents, actually in use? In procurements with environmental criteria this was an award criterion, and a later deviation is then more than an oversight.
Communication and documentation. Is there a named, contactable site supervisor? Are defect notices dealt with traceably? Is a key and access record kept?
Subcontractors. Check whether subcontractors deployed comply with the notification, evidence and consent requirements set out in the tender documents and the contract, and whether they meet the same requirements. What applies is not in the statute; it is in your own documents.
Hygiene and occupational safety. Is the hygiene plan adapted for this facility being observed, rather than the state's framework plan? Are disinfectants correctly dosed? Is personal protective equipment provided and worn?
Single finding and pattern
A stain is a stain. Documenting it is right, but it supports no statement about the service.
One exception first, and it is important: some single findings have to be dealt with immediately regardless of frequency. Contamination with bodily fluids, an unsecured hazardous substance, an acute slip hazard, an incorrectly applied disinfection procedure. Such defects belong in a class of their own, otherwise they dilute to invisibility in a large sample.
Beyond that, it becomes interesting when the same finding recurs. Where the same service type is criticised in a substantial share of the rooms inspected within one room group, and the sample is large enough to make chance unlikely, there is no cleanliness problem but an organisational one. Behind it lies one of a small number of causes: the service is not included in the round schedule, the equipment is missing, or nobody was ever briefed on it.
From what share, and from what number of rooms inspected, you are willing to call something a pattern should be settled in advance. Otherwise the threshold emerges after the fact, out of the result.
That distinction is the difference between a defects list and a usable finding. A defects list leads to rework at the places listed, and three months later the same items appear again. A pattern finding leads to a change in the process.
The precondition is that rooms in which a service type does not occur at all drop out of the analysis rather than counting as "unremarkable". Otherwise every irregularity dilutes until it is invisible.
The trend is worth more than the single figure
An inspection result of 87 says little on its own. Three consecutive results of 92, 88 and 81 say more, and they do so before any threshold is crossed.
More does not mean certain, though. Every sample fluctuates, and whether a movement lies beyond the range of fluctuation depends on the sample size, the season, the intensity of use and whether the same room groups were inspected. A trend is a reason to ask about causes, not a verdict.
For a trend to exist at all, the inspections have to be comparable: the same room group structure, the same assessment levels, the same selection rule. Two series side by side make sense here. The building trend runs on across changes of contractor and shows how the building develops over the years. The contract trend starts afresh with each new contractor and shows what is attributable to them. Keep only one series and you lose either the building's history or the attribution.
What you can do yourself on the next walk-through
Take the room schedule with you and pick three rooms that ought to have been cleaned today according to the interval. Choose them on site, not beforehand.
Ask to see the cleaning store. You need no specialist knowledge for this; for a first impression the general picture is enough. For a formal finding you then need specific criteria and documentation.
Ask about hazardous substances. What matters is not solely whether a folder of safety data sheets sits in the store, since it can equally be provided digitally. The question is whether a hazardous substances register exists, whether the necessary information is current and accessible to staff, whether comprehensible operating instructions are available, and whether task-related instruction has been given. If nobody can answer that, you have a finding that has nothing to do with cleanliness and counts nonetheless.