Robot Fleet Service Credits: Availability Clauses and Excluded Downtime
At a glance: An availability promise of 99 percent sounds like a guarantee and behaves like an aspiration. The clause that decides whether a service credit ever gets paid is not the percentage, it is the list of exclusions underneath it. This guide sets out how availability should be defined to be auditable, which exclusions to accept and which to strike, where a credit stops being a remedy, and when liquidated damages are the better instrument.
Why a 99 Percent Promise Can Still Leave You Short
The first thing to understand about availability clauses is that the percentage is usually measured across the fleet, while the consequence is experienced per unit. Those two things are not the same, and the difference is where service credits quietly fail.
Take a four-unit fleet contracted at 99 percent availability over a month of 30 operating days. The aggregate promise is 0.99 multiplied by 4 units multiplied by 30 days, which is 118.8 unit-days of availability. A fleet that delivers 119 unit-days has met the promise.
Now distribute those unit-days the way a real fleet fails. One unit is down for eleven days after a drive fault and the other three run perfectly, each delivering 30 days. The total is 30 plus 30 plus 30 plus 19, which is 109 unit-days, so this fleet has clearly missed the promise at 90.8 percent and would trigger a credit.
But move the failure shape and the arithmetic reverses. If each of the four units loses three days spread evenly through the month, the total is 108 unit-days and the fleet again sits below target. The aggregate view is not insensitive, so far so good. The problem appears when the measurement window is monthly and the failure is short and severe: a unit down for six consecutive days during a peak trading week is a serious service failure, and if the other three units over-deliver slightly by running extra hours, the monthly aggregate can still land inside the threshold and no credit is due.
This is why per-unit availability, measured against each unit’s contracted operating hours, is the only definition that matches the operational reality. A site does not experience an average; it experiences the specific aisle that was not cleaned on the specific morning the regional manager visited.
The second distortion is the denominator. Availability computed over 24 hours a day is a different number from availability computed over contracted operating hours, and a vendor who calculates over the full calendar has an easier target by construction. A unit that is available for all 720 hours of a month but not for the six hours a night it is needed has a good monthly number and a failed service, so the denominator has to be the operating window, not the calendar.
Defining Availability You Can Audit
An availability definition needs five components to be auditable, and most contracts specify two of them.
The denominator. Contracted operating hours per unit per month, stated explicitly, including which days count. If the site runs six nights a week, the denominator is not 30 days.
The unit of account. Per unit, not per fleet, with a separate aggregate figure if the buyer also wants a fleet measure. Where both exist, the contract should say which one triggers the credit.
The failure threshold. The point at which a unit counts as unavailable. A robot that completes 60 percent of its duty is not available, and a contract that counts any unit that powers on as available is measuring the wrong thing. The threshold should be a proportion of contracted duty, typically 80 or 90 percent, with the measurement taken from fleet software rather than from a callback.
The start of downtime. When the clock starts is worth money. From the moment the fault is logged by the system, from the moment the operator reports it, or from the moment the provider acknowledges it are three very different clocks, and the third is the least favourable to the buyer.
The attribution rule. Who decides whether a given stop was a provider fault, a site fault or a scheduled event, and what happens when the parties disagree. Without an attribution rule, the exclusion list becomes the whole argument.
| Element | Buyer-favourable position | Vendor-favourable position |
|---|---|---|
| Denominator | Contracted operating hours | All calendar hours |
| Unit of account | Per unit | Fleet aggregate |
| Failure threshold | Duty below 90% of plan | Any reported fault |
| Downtime clock starts | Automatic fault log timestamp | Provider acknowledgement |
| Attribution | Independent evidence, fleet logs | Provider determination |
| Measurement source | Buyer-accessible telemetry | Provider monthly report |
The measurement source row is the one that determines whether any of the others can be enforced. If availability is reported by the provider from data the buyer cannot see, the definition is a statement about the provider’s reporting, not about the fleet. Standing read access to the operational dashboard, or a monthly export with the raw stop events, is what makes the rest auditable, and the same access question arises in telemetry and predictive maintenance.
Excluded Downtime: The Clause That Decides the Outcome
Excluded downtime is time when a unit is not working but does not count against availability. It is the most consequential list in the contract, and it is usually presented as boilerplate.
The categories below appear in most agreements. The question for a buyer is not whether they exist but how narrowly they are drawn.
| Exclusion | What vendors write | Recommended buyer position |
|---|---|---|
| Scheduled maintenance | Any planned service activity | Narrow to a published schedule with agreed windows; anything outside it counts as downtime |
| Customer-caused | Any fault attributable to the site | Accept, but require evidence and a written determination within a stated time |
| Network and connectivity | Loss of site Wi-Fi, internet or cloud services | Narrow to outages outside the provider’s own managed network; do not accept a blanket carve-out |
| Site access | Inability to access the site or the unit | Accept if the buyer controls access; strike if the provider schedules its own visits |
| Consumables | Running out of water, detergent or bags | Accept only where the buyer holds the consumables obligation in the contract |
| Force majeure | Broad list including supplier delay | Strike supplier and component delay; those are the provider’s commercial risk |
| Awaiting parts | Time spent waiting for spares | Accept for a defined maximum, typically 5 working days, then downtime resumes |
| Site conditions | Floor, temperature or obstruction issues | Accept only against the documented site survey baseline |
Two of these deserve expansion because they account for most of the value at stake.
The network carve-out is the most abused exclusion in the category. A robot that loses its cloud connection may stop accepting tasks, and a contract that excludes any downtime attributable to connectivity hands the provider a defence against the fleet’s most common operational failure. The narrower and more defensible position is that the provider is responsible for the connectivity it specifies and manages, and that only outages in infrastructure the buyer controls are excluded. A buyer who follows the provider’s own network recommendation should not carry the risk of that recommendation failing.
The awaiting parts exclusion is where a short repair becomes a long outage. Without a cap, a unit waiting for a component counts as excluded downtime indefinitely, and a parts shortage becomes the buyer’s problem in a contract where spare parts availability was supposed to be the provider’s. A cap of five working days, after which the clock restarts, converts an open-ended exclusion into a bounded one. The parts-side planning that makes the cap achievable is described in spare parts and consumables planning.
There is a principle running through all eight rows. An exclusion should cover a risk the buyer genuinely controls, and nothing else. Every exclusion that shifts a provider-controlled risk onto the buyer is worth a negotiation, and each one struck is usually worth more than a percentage point on the availability target.
How Service Credits Are Calculated, and What They Are Worth
A service credit is a percentage of the monthly service fee, scaled by how far availability fell below target. The structure is standard; the parameters are where the money is.
| Availability delivered | Typical credit | Note |
|---|---|---|
| At or above target | 0% | No credit |
| 1–2 points below | 5% of monthly fee | The common entry tier |
| 2–5 points below | 10% | Where most real failures land |
| 5–10 points below | 15–20% | Serious and sustained |
| Below 10 points | 25% capped at 15–30% of annual fee | Cap is the binding constraint |
Work an example. A four-unit fleet at 900 currency units per unit per month is 3,600 per month. A month at 94 percent availability against a 99 percent target sits in the 5 to 10 point band, so the credit is around 15 to 20 percent, or 540 to 720. That is the entire compensation for four units running eleven days short across the month.
The buyer has lost more than that. If each unit-day of missed service costs the site the labour it had to substitute, and a single night of manual cover for one unit costs 180 units, then eleven unit-days of lost availability is roughly 1,980 in substituted labour, before any quality impact. The credit covers perhaps a third of the direct cost and none of the management time.
This is not an argument against service credits. It is an argument for understanding what they are. A credit is a signal and a pressure mechanism, not compensation. Its function is to make poor performance visible in the provider’s revenue line and to create an internal conversation at the provider about the account. Buyers who expect a credit to make them whole are consistently disappointed, and buyers who treat it as a lever negotiate the surrounding clauses better.
The cap is worth attention because it binds more often than expected. A cap at 15 percent of annual fee means a catastrophic year is compensated at 15 percent, and a cap that applies across all credit types, including downtime and response-time breaches, is reached quickly. Where a cap is unavoidable, the useful counter is a termination right triggered by sustained underperformance, because the right to leave is worth more than any credit.
Service Credits and Liquidated Damages: Different Instruments
Service credits and liquidated damages are often used interchangeably in negotiation and they are legally and commercially distinct. Confusing them leaves a contract with a remedy that does not fit the loss.
| Dimension | Service credit | Liquidated damages |
|---|---|---|
| What it responds to | Measured shortfall against an SLA metric | A defined failure of a defined obligation |
| Trigger | Monthly calculation against target | Specific event or milestone |
| Amount | Percentage of fee, scaled by shortfall | Pre-agreed sum per day or per event |
| Typical use in robot contracts | Availability, response time, reporting | Missed go-live date, missed coverage obligation, failed acceptance |
| Legal character | Contractual price adjustment | Genuine pre-estimate of loss; must not be a penalty |
| Practical value | Small, recurring, a signal | Larger, one-off, tied to a real loss |
| Best used for | Ongoing performance pressure | Milestone and mobilisation risk |
The practical division is that service credits govern the steady state and liquidated damages govern the transition. A deployment that goes live four months late hurts the buyer far more than a fleet running at 97 percent for a quarter, and an availability credit schedule is the wrong instrument for it. A pre-agreed sum per week of delay, falling away once the fleet is accepted, addresses the actual loss.
Two cautions apply. First, liquidated damages must be a genuine pre-estimate of the loss, not a punitive figure, or they risk being unenforceable in many jurisdictions. Second, a contract that carries both instruments usually contains a single aggregate liability cap that swallows them, so the cap and its carve-outs matter as much as the schedules. Clause wording of this kind should be reviewed by counsel in the relevant jurisdiction; what a buyer can do without legal support is make sure the two instruments are present, separately identified and not collapsed into one capped pot.
The same drafting discipline applies to the warranty boundary, since a failed acceptance and a warranty claim can describe the same event. That overlap is examined in service contract versus warranty economics.
The Escalation Ladder: Credit, Replacement, Termination
Credits deal with noise. Escalation deals with failure. A contract without an escalation ladder leaves the buyer with a remedy that never escalates, and a provider whose revenue is unaffected by sustained underperformance.
A workable ladder has three rungs, each with a defined trigger and a defined response.
Rung one: the credit. Monthly availability below target, credit applied automatically without a claim from the buyer. Automatic application matters, because a credit that must be requested each month is a credit that is quietly forgotten, and the burden of claiming reverses the intended incentive.
Rung two: the service improvement plan and replacement unit. Triggered by two consecutive months below a defined floor, typically three points beneath target, or by a single unit below 80 percent for a defined period. The response is a written remediation plan with dates, plus a replacement unit at no charge while the original is out. Replacement units are the part buyers most often omit, and they are the part that keeps the service running. A provider with a spares pool can usually meet this obligation easily; a provider that cannot is telling the buyer something about its support model.
Rung three: the exit right. Triggered by three consecutive months below the floor, or by failure to deliver the remediation plan, or by the same fault recurring on the same unit a defined number of times. The remedy is termination for cause of the affected scope, without penalty, plus a defined transition period and an obligation to provide data and spares continuity during it. Without the transition obligation, the exit right is theoretical, because leaving a fleet with no maintenance and no data is not a real option.
The transition clause deserves its own line. The right to terminate is only as useful as the ability to move, so the exit right should be paired with an obligation to hand over fleet data in a usable format, to continue support for a defined notice period, and to make spare parts available at published prices for a period afterwards. The mechanics of that exit are set out in contract renewal, price escalation and exit.
Twelve Clauses to Strike Before Signature
Each of these appears routinely and each is worth removing or narrowing. The rewrite column is what to ask for, not what to accept as a compromise.
- “Availability measured on a fleet average basis.” Rewrite: per unit against contracted operating hours, with the aggregate as a secondary measure only.
- “Downtime calculated from provider notification.” Rewrite: from the timestamp of the fleet system’s own fault or stoppage event.
- “Excludes downtime due to network or connectivity.” Rewrite: excludes only outages in infrastructure under the buyer’s control.
- “Excludes time awaiting spare parts.” Rewrite: excludes up to five working days, after which availability resumes counting.
- “Excludes any maintenance activity.” Rewrite: excludes only maintenance performed within the published schedule and agreed windows.
- “Credits must be claimed within 30 days.” Rewrite: applied automatically in the next invoice cycle.
- “Aggregate liability limited to fees paid in the preceding three months.” Rewrite: retain a cap but carve out the escalation ladder, data return and indemnities.
- “Provider determines cause of downtime.” Rewrite: determination on documented evidence, with a dispute route and a default position if the provider does not respond in time.
- “Force majeure includes supplier or component delay.” Rewrite: strike, because supply of the service is the provider’s core obligation.
- “Credits are the sole and exclusive remedy for performance failure.” Rewrite: credits are the remedy for availability shortfall only, without prejudice to termination rights and to other specified remedies.
- “Response time measured from buyer agreement of a site access window.” Rewrite: measured from the fault event, with site access assumed reasonable and defined.
- “No obligation to replace a failed unit before repair.” Rewrite: replacement unit within a stated period after a unit falls below the floor.
Not every one will survive negotiation, and the order above is roughly the order of value. The first three usually account for most of the total exposure, and a buyer who wins only those has materially changed the contract.
Instrumenting Availability So You Can Claim It
None of the above works without evidence the buyer holds independently. The practical minimum is four artefacts, all of which should be a contractual obligation rather than a favour.
A monthly availability export per unit, with operating hours as the denominator, availability computed to two decimal places, and a list of every stop event with its start time, duration and attributed cause. A dashboard screenshot is not an export.
The exclusion log, showing which stop events were excluded and under which clause. This is the document that makes the exclusion list auditable, and a provider that cannot produce it is applying exclusions without evidence.
A credit statement, showing the calculation and the amount applied, so the buyer does not have to reconstruct the arithmetic.
The maintenance schedule, published in advance, so that scheduled downtime is bounded and unplanned activity cannot be retro-fitted to it.
The measurement design is not complicated, and it is worth agreeing before go-live rather than after the first dispute, because the definition of a stop event is much easier to negotiate while the fleet is not yet running. The operational side of that instrumentation, including the KPI set most fleets should hold, is in the guide to fleet utilisation benchmarks and the intervention and response tiers in incident response and escalation, which covers the emergency side that the availability clause does not.
If you are drafting or reviewing an availability schedule for a fleet, we can supply our standard credit schedule, the stop-event export format and the escalation ladder we work to, so the contract is measurable in practice rather than in principle. The cleaning and delivery ranges share the same reporting model, and the quote form reaches a technical contact directly.
