Service Robot Uptime SLA, What the Percentage Actually Commits To

The Percentage Is the Least Important Number in the Clause

Every service robot contract in the market carries an uptime figure. They range from 95% to 99.5%, and buyers read the number as a proxy for reliability. It is not. Availability is a ratio, and both the numerator and the denominator are defined in the contract, which means the same physical fleet can report 96% or 99% depending on definitions chosen at signature time.

The definition work is worth doing properly because the consequences are asymmetric. If your operation depends on the fleet and the fleet is down, the credit you receive under the SLA will rarely cover the cost of covering the work manually. The clause that matters is not the payout schedule; it is the set of conditions under which the clock stops.

Availability: Agree What Is Being Measured

Start by fixing the unit of measurement. There are three conventions in commercial contracts and they produce very different results.

Machine-hours versus task-hours

Machine-hours availability measures whether the robot was powered and ready. Task-hours availability measures whether the robot completed its assigned tasks. A robot that is powered on but stuck behind an uncollected pallet is 100% available under the first convention and 0% under the second. For cleaning and delivery fleets, task-hours is the honest measure, because a machine that cannot reach its route has not delivered the thing you bought.

Fleet-level versus per-unit

A five-unit fleet with one machine down is 80% available per-unit and frequently reportable as 100% available at fleet level if the remaining four absorbed the routes. Fleet-level availability is legitimate for operations that genuinely have redundancy, and it is a trap for operations that do not. Insist on per-unit reporting with a fleet aggregate, not a fleet aggregate alone.

The operating window

Availability must be measured across the hours the site is operable, not across the calendar. A machine available 100% of a two-hour nightly window is at 100%, not 8%. This is the one place where a narrower denominator genuinely favours the buyer, and it is also the most frequently disputed line in a post-implementation audit.

Convention	What it hides	Recommendation
Machine-hours, calendar base	Idle machines, blocked routes	Reject
Machine-hours, operating window	Task failures while powered	Acceptable floor
Task-hours, operating window	Little; the honest measure	Preferred
Fleet aggregate only	Single-unit chronic failure	Require per-unit too

Excluded Downtime, and Why It Decides Your Real Uptime

Standard contracts exclude a list of conditions from the availability calculation. Some exclusions are reasonable; others quietly move most real outages out of scope. The four that matter most:

• Customer-caused downtime. Legitimate in principle. The dispute is always about cause. A robot that faults because a corridor was blocked by a tenant delivery is arguably customer-caused; a robot that faults because it cannot navigate a corridor it was commissioned on is not. Require that any fault the robot logs as a navigation rather than environment event stays in scope.
• Force majeure and scheduled maintenance. Scheduled maintenance windows should be agreed as a defined allowance, not an open exclusion. Two planned visits per machine per year is typical and defensible; "routine maintenance" as an unbounded category is not.
• Site conditions. Floor damage, unplanned construction, network changes and power interruptions appear in most exclusion lists. Ask what evidence the vendor must produce to invoke the exclusion, and require that it be logged at the time of the fault rather than reconstructed at renewal.
• Parts availability. This is the exclusion with the largest real-world impact and the smallest print. If a repair is delayed because a component is on back-order, many contracts stop the clock. That transfers supply-chain risk to you while leaving the service fee unchanged. Strike it, or cap the days it can apply.

The service contract economics of these exclusions, including how they shift the effective premium, are worked through in our warranty versus service contract analysis.

Response Time and Restoration Time Are Not the Same Promise

Almost every contract commits to a response time. Far fewer commit to a restoration time, and the difference is the whole value of the agreement.

A four-hour response window means an engineer acknowledges and begins diagnosis within four hours. It says nothing about when the machine works again. A restoration commitment, by contrast, states that the unit returns to service within a defined period, or the vendor provides a substitute. Restoration terms are harder to obtain and considerably more valuable, and where a vendor resists them, the substitute-unit provision is the next-best instrument.

Commitment	Typical 2026 term	What it guarantees
Response time	4-8 business hours	Diagnosis begins
On-site attendance	Next business day	Physical presence
Restoration time	48-72 hours from attendance	Unit back in service
Substitute unit	After 72 hours down	Route coverage continues
Parts availability cap	Rarely offered	Back-order risk limited

If a vendor offers only the first two rows, the practical uptime guarantee you hold is considerably weaker than the 97% headline suggests. Ask for restoration plus substitute as a package; that combination converts an availability percentage into an operational commitment.

Service Credits, and Why They Rarely Pay

Availability SLAs enforce themselves through service credits, expressed as a percentage of the monthly service fee. Three structural features typically prevent those credits from ever being paid.

First, the threshold is set below normal performance. A 95% floor on a fleet that routinely delivers 97% produces a credit only in a genuinely bad month, which is the intent and is fine. The problem is when the floor is set at 90% on a fleet that normally runs at 92%, at which point the credit is a theoretical instrument.

Second, the credit is capped. A common cap is 10% of the monthly fee. On a $300 per machine monthly service fee, the maximum credit for a month in which the machine was unusable is $30. Compare that to the cost of covering the route manually, which for a 1,500 m² nightly route is several hundred dollars. The credit is a token, not insurance.

Third, credits often require the buyer to claim them, with a short window and evidence the buyer may not have collected. Vendors are not obliged to volunteer a credit the buyer never requests.

The practical response is not to chase a larger credit. It is to negotiate the remedy that actually protects the operation: a substitute unit, an extension of the service term for downtime days, or a defined reduction in the following month's fee. Our warranty and contract terms guide catalogues those substitute remedies and the clauses that deliver them.

The Four Clauses to Fix Before Signing

If you take one thing from this piece, take these four amendments. They are the difference between an SLA that describes your operation and one that describes the vendor's reporting system.

• Define availability as task-hours across the operating window, reported per unit with a fleet aggregate alongside it.
• Cap parts-availability exclusions at a stated number of days, after which the clock restarts and the substitute provision triggers.
• Attach a restoration commitment and a substitute unit provision to the response window, with the substitute threshold expressed in hours of downtime.
• Make credits automatic and payable in cash or term extension, with the claim window removed and the cap raised to a level that reflects the cost of manual coverage.

Auditing those commitments once the fleet is running requires the same measurement discipline as the SLA itself, and the target ranges to audit against are in our KPI benchmarks. The measurement pipeline that makes per-unit reporting possible is described in our telemetry and predictive maintenance breakdown.

AOMAN FUTURE provides per-unit task-hour availability reporting on the D1 and C2 Pro platforms and will attach a restoration commitment and substitute-unit provision to service agreements where the route geography supports it. If you are drafting an SLA specification, send us your route plan and we will return the availability definition we report against, written as contract language rather than marketing.