Robot Fleet Utilization Rate, The Metric That Sets Your Real Cost Per Clean
At a glance: Two facilities can buy the same six robots and end up with cost per square metre figures that differ by half. The difference is almost never the machine, it is how much of each machine's available time is actually converting into cleaned floor. This guide defines fleet utilization so it cannot be gamed, gives benchmark bands to judge your own number against, and shows the five places fleets lose hours without anyone noticing.
Utilization Is Not the Number Vendors Quote
Almost every fleet proposal includes a utilization figure, and almost every one of them is measured against a denominator that flatters the result. The honest definition is narrow and it is the only one worth tracking.
Fleet utilization rate = total productive unit-hours delivered in a period, divided by total unit-hours the fleet was available to work in that same period.
Productive unit-hours are hours a robot spent performing a scheduled cleaning or delivery task. Available unit-hours are the hours the robot was deployed, powered, connected and not in planned maintenance. Both halves of that fraction need to be pinned down before any number means anything.
The Denominator Problem, Three Ways to Lie by Accident
Most inflated utilization figures come from choosing a generous denominator. There are three forms of this and they all look reasonable until you write them down.
| Denominator used | What it produces | Why it is wrong |
|---|---|---|
| 24 hours × days | Utilization of 8 to 12% | Includes hours the building is closed and no cleaning is schedulable. Punishes you for a constraint you cannot remove. |
| Calendar hours the site is staffed | Utilization of 20 to 40% | Includes the traffic window when robots cannot safely share floor space with the public. |
| Deployed hours, robot switched on and schedulable | Utilization of 45 to 75% | Correct. This is the denominator to contract on. |
The practical rule is to fix the denominator as deployed unit-hours, written into the fleet agreement alongside a definition of what counts as deployed. When a vendor declines to state the denominator, the percentage they quote is decorative.
Benchmark Bands, What a Healthy Number Looks Like
Utilization depends heavily on building type, because the schedulable window is set by occupancy. The bands below are drawn from facilities that have run autonomous fleets for at least two full quarters and measure against the deployed-hours denominator.
| Facility type | Typical schedulable window | Healthy utilization band | Constrain it |
|---|---|---|---|
| Office, single shift | 9 to 10 hours per night | 55 to 70% | Traffic clearance and lift availability |
| Hospital, continuous | 18 to 20 hours | 60 to 75% | Corridor congestion, isolation protocols |
| Airport or transit concourse | 4 to 6 short windows | 35 to 50% | Passenger flow, stopped-area permits |
| Warehouse or distribution centre | 16 to 20 hours | 65 to 80% | Aisle availability, pallet obstruction |
| Retail, open hours | 2 to 4 hours before opening | 30 to 45% | Shelf restocking, customer floor time |
Read the table as a constraint ceiling, not a target. A retail site running at 45% is not underperforming; it is running at its ceiling, because the building only gives it a short window. A warehouse at 45% is leaving half its fleet idle in a building that was open for twenty hours.
The same logic is why fleet size and schedule should be derived together rather than separately. If you are still sizing the fleet before you have measured the window, our note on right-sizing a fleet from four measured inputs is the place to start, and if you have already sized it, the utilization number is what tells you whether the sizing held up.
The Five Places Fleets Lose Hours
Once you measure honestly, the lost hours cluster into five recurring categories. In the fleets we have reviewed, these five account for 25 to 40% of available unit-hours before any hardware fault is involved.
- Charging queue collapse. Robots return to charge at the same time because they were dispatched at the same time. Two units wait while one charges. This is a scheduling problem, not a battery problem, and the fix is staggered dispatch or a second docking point.
- Lift and door waiting. Multi-floor routes lose time at every vertical transition. A robot that waits 90 seconds at each of six lift calls in a night loses nine minutes per unit, and nine minutes across a six-unit fleet is nearly an hour of capacity per shift.
- Deadheading between zones. Travel time that does not clean anything. A badly sequenced route can spend 15% of a session in transit between areas that could have been adjacent.
- Manual intervention for false obstructions. A robot that stops for a shadow, a wet-floor sign or a misread pallet needs a human to release it. Each intervention is typically 4 to 12 minutes of unit time plus supervisor time.
- Task completion without verification. A robot that reports a zone complete when a corner was skipped is not losing hours directly, it is corrupting the utilization number, because the hours count as productive while the floor is not clean. This is the most damaging loss because it is invisible in the metric.
Why Utilization Drives Cost Per Square Metre
Cost per square metre is the number a facility manager actually has to defend, and utilization is the lever that moves it most. The relationship is close to inverse: if you double productive hours from the same fleet, you roughly halve the fixed cost carried by each square metre.
Worked example, a six-unit cleaning fleet on a single shift:
| Line | Low utilization case | Optimised case |
|---|---|---|
| Deployed unit-hours per night | 6 units × 4 h = 24 | 6 units × 4 h = 24 |
| Utilization rate | 45% | 68% |
| Productive unit-hours per night | 10.8 | 16.3 |
| Coverage per productive hour | 1,100 m² | 1,100 m² |
| Floor cleaned per night | 11,880 m² | 17,930 m² |
| Fleet monthly cost, all-in | $9,600 | $9,600 |
| Cost per m² per month | $0.031 | $0.021 |
Nothing changed on the machine side. The same six robots, the same batteries, the same cleaning heads. A 23 percentage point improvement in utilization cut the cost per square metre by a third, which is a larger swing than most hardware negotiations deliver.
This is why utilization belongs next to the other financial instruments rather than in a maintenance report. It sits alongside the payback period calculation as an input, and it is the mechanism that most often explains why a total cost of ownership model built on optimistic assumptions misses on the downside.
How to Measure It Without Instrumentation Debt
You do not need a data platform to measure utilization. Four numbers per unit per shift will do it, and if your fleet management system already logs task state you can extract them without new hardware.
- Session start and end time per unit, from the robot's own log rather than the operator's memory.
- Productive task time, the sum of intervals where the robot was executing a scheduled task and moving or cleaning, not paused.
- Intervention events, count and duration, with a one-word reason code.
- Verified coverage, area actually completed against area scheduled, so an over-reported task cannot inflate the productive total.
Track it weekly, not monthly. Monthly utilization hides the pattern of which night of the week collapses and why, and the fix is nearly always schedule-shaped.
If you have more than one building, the measurement has to be consistent across sites or the comparison is meaningless. The same definitions applied per site is what makes a portfolio number trustworthy, and our guide to fleet management architecture covers where that data should live when sites are run by different operators.
What to Put in the Fleet Agreement
Utilization is easy to game and easy to defend against once it is defined. The following clauses keep the number honest through the life of the contract.
- State the denominator. Deployed unit-hours, with deployed defined as powered, connected and schedulable.
- Define productive time as active task execution only, and state explicitly that paused, returning-to-dock and standby time are excluded.
- Report intervention events separately from productive hours, with reason codes and durations.
- Tie a service credit to a utilization floor measured over a rolling month, not a single good week.
- Require verified coverage alongside the utilization figure so a high percentage cannot be manufactured by over-reporting tasks.
A vendor confident in their scheduler will accept these terms. A vendor who wants to keep utilization vague is telling you where the weakness in their fleet orchestration sits, and that is useful information before you sign rather than after.
The Takeaway
Utilization is the bridge between what you bought and what you get. It is a simple ratio with a contested denominator, which is exactly why it needs to be written into the contract in plain terms. Fix the denominator, benchmark against a band that reflects your building's real window, attack the five leak points in order, and the cost per square metre takes care of itself. The metric is not a maintenance statistic, it is the price you are paying per clean.
