Two-Week Floor Robot Trial Before Any Lease: A Pilot Guide

What should a two-week floor robot trial prove?

A two-week floor robot trial answers one narrow question: can a machine handle one zone on one overnight shift well enough to justify a lease quote? It should not try to settle whether robots belong across an entire building. Keep the scope small, measure it carefully, and agree in advance on what result counts as a yes.

The category is growing. According to Supply Chain Digital, summarizing IFR's World Robotics 2026 service robot report, global professional service robot shipments rose 24% to almost 250,000 units in 2025. Supply Chain 247 reports that professional cleaning ranked third, with a 12% share.

Growth does not prove fit. A machine that runs beautifully at a sister site can stall at your dock door, and only your floor, your schedule and your staffing can show that. A disciplined trial is how you collect that proof.

How do you choose the zone and the overnight shift?

Choose one zone with a known footprint, a predictable floor type and steady overnight traffic. A long open aisle or a corridor loop teaches more than a cluttered back room, because the trial should test cleaning and navigation, not your team's ability to clear obstacles. Measure the square footage before the robot arrives so every later figure has a denominator.

Pair the zone with one overnight shift and name a person who knows that space. That lead becomes the observer and holds authority to pause the unit. Without a named observer, interruptions go unrecorded and the results look better than they are.

Write the baseline down too. How many labor hours does the zone consume today, who does the work, and what usually gets missed? A robot can only be judged against a number someone actually captured.

What rules should be fixed before day one?

Put the scope on one page: the zone map, start and stop times, cleaning method, exclusion areas and a plain definition of failure. Agree on who may stop the robot and under what conditions. Keep these rules fixed for all fourteen nights so that night-to-night comparisons mean something.

Expect the first night to surface mapping problems. Stacked pallets, low lighting, a wet-floor sign left in the path and a propped door all change what the robot sees. Log each one as it happens instead of quietly correcting it, because those site conditions are exactly what an assessor will need later.

Which metrics belong on the scorecard?

Keep the scorecard short. Five measures cover most of what a manager needs to decide.

Stamp every stop with its time. A tally alone hides whether problems cluster at one hour, such as when a delivery door opens or a cart crew starts early.

• Coverage: square feet cleaned against the zone total, confirmed by a manual walk at the end of each shift.
• Autonomous runtime: hours the unit ran unattended each night, and the share of its active time that stayed autonomous.
• Interruptions: stops per shift, each tagged by cause such as an obstacle, a sensor fault, charging or a staff pause.
• Staff touch time: minutes a person spent setting up, clearing, recovering or re-running the zone.
• Missed spots: areas the morning walk finds dirty, tagged by location.

What do vendor pilot figures prove, and what do they not?

Published case studies are reference points, not forecasts. A field test run with C&W Services at an international airport, using Avidbot Neo 2 machines, reported 42 days of testing, 2.4M sq ft cleaned, more than 230 hours of cleaning and productivity of 10,735 sq ft per hour. Those figures come from the service provider, so read them as evidence of what a well-run test can record.

A 2023 SoftBank Robotics case study reported that S50 scrubbers running autonomously overnight over about 50,000 sq ft cut hours spent scrubbing by 50% in a healthcare building. A Tennant case study with Domus reported typical autonomous cleaning of 20 hours per week, with 99.2% of runtime in autonomous mode. Tennant's manufacturer rating for the T380 is 33,440 sq ft per hour, with four to six hours of runtime per charge. Ratings describe ideal conditions, so your own numbers will differ.

Some trials report outcomes without hours. Hilton at Heathrow Terminal 4 ran a six-week on-site trial of TASKI Phantas, reported no issues and moved to rollout, with no hours figure given. A three-week pilot at the University of British Columbia with A&K Robotics produced qualitative feedback only. Treat these as reasons to measure, not as proof.

IFR's sales figures also carry a caveat. They come from a sample of producers and are not projected to the whole industry, so keep that wording in any board memo.

How does the trial feed a lease quote?

A lease quote follows a site assessment, and the trial hands the assessor facts no one can guess: real runtime, real stoppages, real charging windows and the true edges of the zone. Service Robot Co. leases robots on a three-year term, with one fixed quote after that site assessment, so a cleaner trial log produces a more precise quote.

Record charging behavior in detail. Note where the unit docks, how long recharging takes and whether the dock blocks an exit or a cart lane. Those observations decide where a permanent dock belongs.

Log service events too. If a sensor faults or a brush needs changing, record the time, the fix and who performed it. Every unit we lease is serviced under RoboCare, our service plan, backed by regional service engineers in the states we serve, and that response history is what makes coverage conversations concrete.

How do you make the go or no-go call?

Set the thresholds before the robot arrives, write them down and have the zone manager sign them. Base them on your baseline, not on a brochure. A go decision might require full coverage of the agreed zone on every trial night, with staff touch time below the baseline labor hours. Choose the numbers yourself so the decision stays honest.

Use three outcomes. Go means you move to a site assessment and lease quote for this zone. Extend means the results are promising but one cause, such as a dock door or poor lighting, needs a fix and a second week. No-go means a pre-agreed criterion failed, and you record the reason so the next evaluation starts further along.

Avoid treating one bad night as a verdict. Read the trend across all fourteen nights, and separate failures caused by the site from failures caused by the machine. We are OEM-neutral, so if the trial points to a different model than the one first considered, that finding belongs in the decision.