Preparing the published article.
How to Scope a Cleaning Robot Trial by Zone First
Build a zone register of traffic, floor type and soil load before any machine is demonstrated on site.
Build A Zone Register Before You Look At Machines
The most common procurement mistake in cleaning robotics is to accept a demonstration first and to define the trial afterwards. Reverse that order. Before a machine is brought on site, walk the building and produce a zone register: one row per physically distinct area, with the same fields completed for every row. A register turns a demonstration into a measurable trial, because it fixes what is being tested and where. Start with the fields you can observe without any equipment: floor finish, area in square metres, the hours when the zone is occupied, who currently cleans it, how long that task takes today, and where soil visibly collects. Then add the two fields that drive most machine selection, which are traffic volume and soil load. Record both in simple, repeatable terms rather than in adjectives, so that two people walking the same corridor reach the same entry: define what counts as light, medium and heavy traffic for your site, and attach a plain description to each level. Note the access route into each zone as well, including the narrowest point the machine would have to cross, because a zone that cannot be entered is not a trial zone. A register with twenty honest rows is more useful than one with two hundred estimated rows.
Separate Footfall From Soil Load
Traffic and soil load are not the same measurement, and treating them as one is how a trial gets scoped wrongly. A main entrance mat and a food court servery may see similar footfall, yet the entrance delivers grit that is dry and abrasive while the servery delivers grease, food debris and spilled liquid that need different chemistry and different dwell time. Ask, for each zone, what the soil actually is, where it comes from, and whether it is wet or dry. Then group zones that share a soil type and an access pattern, because those groups are what a single machine configuration can realistically cover. Aim for three to five groups rather than twenty individual cases. For each group, note the cleaning frequency the zone genuinely needs, not the frequency it currently receives, and state in one sentence what a good result looks like there. Write that sentence in observable terms, such as the absence of visible debris along a defined inspection line, rather than as a general impression of cleanliness. If you cannot describe the result in observable terms, the zone is not ready to be part of a trial, because no later measurement will settle the question.
Decide What Success Looks Like Before Deployment
Before deployment, agree three things in writing: which zones are in scope, what the baseline is, and how the comparison will be made. Measure the baseline yourself while the zone is still cleaned by hand, at the same hour of day and on the same weekday, so the comparison is not distorted by a weekend or a public holiday. Use a method you can repeat, such as a marked inspection point, a fixed inspection time, and a short form that records what was found. Keep a named person responsible for each measurement, and record that person alongside the result, because an unattributed observation is difficult to defend later. Then decide which zones are outcome questions and which are feasibility questions. An outcome question asks whether the floor reaches the agreed standard; a feasibility question asks whether the machine can reach the zone, turn in it, and work without blocking people. Keep the two separate in your reporting, because a zone that fails on access has not failed on cleaning quality, and a mixed result hides that distinction. Finally, decide in advance what would make you stop the trial, and write it down next to the zone list.
Photo credits: the images in this article are licensed reference photographs of the relevant robot class, credited with author and licence in each caption; they are not photographs of the specific equipment discussed. Licence and author details are recorded with each image.