Start with what the crew is being asked to prove
Most failed kitchen trials are not cleaning failures. They are question failures. A crew sent out with no written question comes back with an opinion, and an opinion cannot be reviewed. Before any equipment is discussed, one person writes down the sentence the trial must answer, in the form of a claim that could turn out false: that the method removes the wax-like film building on the underside of the canopy; that the filter pack can be cleaned in place during a service window rather than taken to a wash bay; that the duct entry lip can be reached by hand without opening the duct. Write one claim per surface, not one for the system as a whole, because a single claim hides the fact that three of four surfaces may have gone well while one did not.
Each claim needs an owner, a date, and the person who will read the result. In a chain that person is usually the facilities manager or a regional manager who did not watch the trial. If the crew cannot say who reads the record, the trial is a demonstration and it is fine to run it as one; just do not later treat it as evidence.
Have the crew name the deposit before they name a method
Restaurant extraction surfaces carry more than one kind of residue, and a prepared crew sorts them on the surface itself, with a gloved finger and a scrap of white paper, before any decision about heat, dwell, or detergent. Grease film freshly deposited from a fryer line behaves one way. Grease that has baked hard and polymerised on a hot canopy panel behaves another. A wax-like condensate that travels further up the run and settles as a dull, soft skin behaves differently again. Loose dust and flour bound into the residue near the kitchen end is a fourth case. A crew that can tell these apart on the panel can explain afterwards why one panel cleared and the neighbour did not.
Give each crew member the same three words for what they feel: soft and wiping, hard and flaking, or waxy and smearing. That vocabulary travels between sites; a longer chemical description does not, and it is not something a kitchen team should be asked to verify.
Set the boundary list, with an alternative for everything left out
A prepared trial states what it will not touch. For extraction work the exclusions usually run to the filter media itself where the pack is damaged or sagging, to any electrical enclosure and its cable entries on the canopy and fan, to bearings and the fan shaft area, to damper linkages, to painted panels that have been re-coated locally, and to any inspection hatch seal that has already gone brittle. Each exclusion must carry a fallback in writing, otherwise the crew quietly stops at the boundary and the record shows a clean gap that nobody planned.
Fallbacks are ordinary decisions: this panel goes to the manual scrub team, this filter pack gets replaced rather than cleaned, this fan housing is assessed on the next shutdown. A trial that has named its exclusions and their fallbacks is easier to scale than one that claimed the whole system, because the second time you only repeat what actually worked.
Mark two test patches and photograph them before the first pass
The most useful thing a crew can do before a trial is put two or three patches on the worst panels, tape them, and photograph them under the same light, at the same angle, on the same day. Marked patches turn a vague impression into a comparison. Photographs taken after the cleaning, without a matched before shot, are worthless to anyone who was not standing there.
Put patches where the crew expects difficulty rather than where it is easiest to see a result. A patch on a clean, flat, well-lit panel flatters the method and misleads the reader. A patch on the underside of the canopy, where the deposit has had months to build, answers the question you actually asked. Keep one patch on a surface that is in view of the dining side, since that is the one area managers look at first.
Decide in advance how heat and dwell will be told apart
Preparation is not complete until the crew knows how the trial will report what did the work. Heat alone, heat held against the surface for longer, or the mechanical action of the jet and brush all move residue, and an unprepared trial reports them as one result. That matters because scaling copies the wrong variable: a site buys a different machine when what was missing was thirty seconds of patience.
Write the two patches to differ on one variable only. Same panel, same deposit, one patch cleared promptly, one held before clearing. The crew then records which patch cleared and how, in one sentence each. If both clear equally, say so; a trial that reports a null result is worth more than a trial that claims a cause it did not test.
Walk the room before the room is under service pressure
A staff who has walked the kitchen the day before the trial does not spend the trial hunting for a socket or a hose. The walk covers reach to a water point, the route from that point to the canopy, whether the machine can be moved into place through the service door or has to be carried, how far the supply line reaches along the extraction run, and whether the walk-in or the preparation area has to be closed while the work happens. The walk also settles where people will walk during the trial, because a wet floor under a servery hatch is an accident waiting for the first customer queue.
Whether the equipment sits fixed in one place or is moved along the line is read from this walk, not from a rating sheet. A system that runs the length of a kitchen with a fan at the far end is a reach and routing problem first and a machine question second.
Put the run-off in the hands of someone with a name
Emulsion and wash-down water coming off a canopy has to go somewhere, and on a live site this is the detail that ends trials. Preparation means walking the same route the water will take and getting the site to agree it before the trial: which drain or gully receives it, whether floor channels or ponding will appear under the line, whether the floor outside the kitchen is used for customer circulation afterwards, and who signs the floor off as safe at the end. A trial that produced an impressive result and a flooded scullery teaches the reader the wrong lesson.
Ask the site whether the kitchen can be released before the floor is fully dry. A written answer here changes the trial into something the kitchen can accept; an unwritten answer turns into an argument at ten at night.
Bring the questions you will need answered in writing, as a list
A crew on site will be asked what the machine is capable of, and a prepared crew does not answer from memory. The site pages for the industrial steam cleaner range position the mobile machines around indoor garages and tight outdoor parking, and the fixed machines around vehicle, interior and engine-compartment work. Neither page sets out a duty for extraction systems. That gap is precisely why the trial is run, and it is why the specification has to be collected in writing for the exact machine quoted.
So the crew arrives with a short list, to be filled in with figures for the unit on site rather than read off a brochure: the steam temperature stated for that machine, the working pressure stated for that machine, the heating load and supply voltage the kitchen must provide, the water tank capacity and what that means for refill pauses during a single run, the water inlet arrangement and whether it fits the kitchen's tap connection, the footprint and how the machine moves, and which protections the machine lists. A supplier who will not put these in writing for the quoted machine has answered the question anyway. On a site where the utility supply is itself the constraint, ask specifically about Fixed Steam Cleaning Machine PowerDe HW02 as the fixed option on the site's own page, and read its stated rows back to the kitchen's electrician before anything is booked. If the extraction run, the service windows, and the run-off route do not fit any catalogue entry, that is the point at which customized steam cleaning equipment becomes the sensible enquiry rather than an early one.
Fix the stop conditions before the first pass
A prepared crew knows what will make them stop. Typical written stops for extraction work: residue that has driven into panel seams or rivet lines and does not lift; filter media that swells, sags, or sheds after contact; any sign of water reaching a cable entry, motor, or bearing; a panel whose finish lifts or clouds; floor channels that begin to back up; or a service window that has run out before the patches are finished. Each stop is a decision recorded at the moment it happens, not reconstructed afterwards.
Name one person who can call the stop and one who can release the kitchen back to service. Two roles, not one. The person holding the wand should not be the person deciding whether the floor is safe to open.
Write the one-page record the crew completes on site
Preparation ends with the form. One page, one row per patch: site, date, surface, deposit word from the shared vocabulary, what was tried, whether heat or dwell was varied, what came off and what did not, where the water went, what was excluded and with which fallback, and the name of the person who inspected the patch. The inspector is not the operator. A photograph reference for each patch belongs in the same row.
That page becomes the standard for the next site if the trial is accepted. Written in this form, it also survives staff change, which in restaurant back-of-house is not a risk at the end of the year but a certainty during it.
What a prepared trial still does not answer
Preparation cannot make a kitchen release a surface it has never released, and it cannot prove a frequency. A trial shows what happened on marked patches, on named panels, on one date, under one set of conditions. It does not show what the duct interior carries, and it does not substitute for whatever the site's own fire and extract-cleaning obligations require, which are set by the site and its insurer, not by a cleaning trial.
A prepared crew says that plainly in the record. That is also what makes the accepted parts of the result believable when the same crew returns with a second site asking whether the method holds.
