Two operators can sign contracts with the same scope of work, the same visit frequency and the same price, and get very different outcomes. The difference is almost never in the scope list. It is in the definitions: what counts as a response, what counts as available, and what the exclusions cover.
Those definitions are where an operations contract is won or lost, and they are usually the sections negotiated last. This guide is about what to pin down before signing, written for the owner or asset manager rather than for the service provider.
Why the contract matters more than the scope list
A scope list describes what gets done. It does not describe what happens when something goes wrong, and everything that goes wrong lands in the gaps between the clauses. Two contracts with identical scopes can differ enormously in what you actually receive during an outage.
The pattern to watch for is a scope list that is detailed and specific, attached to definitions that are vague. The detail is where the provider competes; the vagueness is where they are protected.
Response time and resolution time are different promises
This is the single most common gap, and it survives review because both numbers look like service commitments.
- Response time — how quickly the provider acknowledges and begins to act. Someone picks up the phone, or an engineer is dispatched.
- Resolution time — how quickly the fault is actually cleared and the plant is producing again.
A contract that commits only to response time has committed to starting, not to finishing. On a commercial plant the cost of an outage accrues from the moment generation stops until it restarts, so the number that matters to you is the second one.
Ask specifically: is there a committed time to restore generation, is it measured from the alarm or from the acknowledgement, and what happens if it is missed. A service credit that is smaller than the value of the lost generation is not an incentive; it is a fee for late performance.
What "availability" means, and how downtime is counted
Availability figures appear in most operations proposals. Whether the number means anything depends on definitions that are rarely written down:
- The denominator. Availability against all hours in the year, or against daylight hours? The two give different percentages from identical plant performance.
- What counts as downtime. A grid outage, a curtailment instruction, or a shutdown for planned maintenance is not the operator’s fault — but unless the contract says so explicitly, it may still be counted against their availability figure.
- Who classifies the cause. If the operator determines whether an outage was equipment-related or external, the availability figure is self-reported. Ask for the classification to be tied to the monitoring data.
- Excluded events. Weather beyond the design basis, third-party damage and utility-side events should be named and excluded rather than left to interpretation.
An availability guarantee without these four is a statement of intent. With them, it is measurable — and measurability is the whole point.
Spare parts: ownership, location and replenishment
Three questions that are usually settled by accident rather than by design:
- Who owns the spares held on site? A consignment stock owned by the provider can be withdrawn; stock you own cannot.
- Who pays for replenishment? A fuse or surge cartridge consumed clearing a fault is a consumable. Whether it falls inside the contract value or is recharged changes the economics of every visit.
- What happens to replaced parts? Ownership of a removed component matters when a warranty claim depends on returning or inspecting it.
Our guide to plant operations and maintenance covers which items are worth holding on site; the contract is what says who owns them once they are there.
Warranty and O&M are two different promises
They are frequently treated as one, and the gap between them is a real cost:
- A warranty is a product promise. The manufacturer undertakes that the equipment will perform, and remedies a defect in the product.
- An O&M contract is a service promise. Someone attends, diagnoses, and carries out work.
The gap is labour. A warranty that replaces a failed inverter does not necessarily pay for the technician who removes it, ships it, and reinstalls the replacement. On a large plant that labour can exceed the value of the part. Ask, in writing, whether warranty replacements include the labour to execute them — and if not, whose contract covers it. The claims process is set out in our module warranty claims guide.
Reporting, and who owns the data
Two things to settle, and the second is the one people regret:
- Format and frequency. Monthly is typical. What matters is that the report ties generation to the availability definitions above, so that a performance discussion is about numbers both sides accept.
- Data ownership and access. If the monitoring platform is the provider’s, you are reading your own plant through their account. Ask who holds the administrator credentials, whether you can export the full history, and what happens to the data when the contract ends.
Losing years of generation history at contract end is a real outcome, and it removes exactly the baseline you would need to diagnose a performance problem later. Our guide to diagnosing underperformance depends on having that history.
What happens at the end of the term
The clause most often missing entirely. If the contract ends — or the provider changes — you need:
- As-built documentation and any modifications made during the term.
- The full monitoring history, exportable, in a usable format.
- Credentials and administrator access transferred, not just shared.
- Site keys, access arrangements and any lender or insurer requirements documented.
- A condition report at handover, so the incoming provider does not inherit someone else’s deferred maintenance as an argument.
Without a defined handover, the transition is negotiated at exactly the moment you have the least leverage.
What to define before signing
Reference table
Swipe horizontally to view all columns →
| Item | What to pin down | What happens if it is left open |
|---|---|---|
| Response vs resolution | Both, measured from a stated trigger | Commitment to start, not to restore generation |
| Availability definition | Denominator, excluded events, cause classification | A figure that cannot be argued either way |
| Exclusions | Named events: weather beyond design, grid, third party | Decided per incident, by the paying side’s counterparty |
| Spares | Ownership, location, replenishment cost, removed parts | Consumables recharged on every visit |
| Warranty interface | Whether warranty work includes labour | Labour cost exceeds the value of the replaced part |
| Reporting | Frequency, format, and data ownership | Performance discussions without agreed numbers |
| Data at exit | Export rights and credential transfer | History lost at the worst possible moment |
| Handover condition | Documentation and condition report at term end | Transition negotiated with no leverage |
How this fits your procurement
Operations cost belongs in the same model as the capital cost — an availability gap or an uncovered labour line changes the return on the asset. If you are building that model, our commercial solar payback model is where the two meet. And the baseline that makes any of it measurable is taken at commissioning, not at handover; the sequence is in our commissioning checklist.
We supply the equipment and the documentation that goes with it, and we do not provide operations services — so none of the clauses above are ours to sell you. That is also why the list is worth having: it is written from the asset owner’s side of the table rather than the provider’s.
BUYER CHECKLIST
Buyer considerations for this scope
Ask for a resolution commitment, not just a response one
Outage cost accrues from the moment generation stops until it restarts. A contract that commits only to acknowledging and dispatching has committed to starting, not to finishing. Ask what happens if the restore time is missed, and check the credit is large enough to act as an incentive.
Make the availability figure measurable
An availability number without a stated denominator, a named exclusions list and an agreed cause classification cannot be argued in either direction. Tie the classification to monitoring data rather than to the operator's own reporting.
Settle data ownership and exit before you sign
If the monitoring platform is the provider's, ask who holds the administrator credentials and whether the full history is exportable. Losing years of generation data at contract end removes exactly the baseline needed to diagnose a performance problem later.
PRACTICAL ANSWERS
Questions buyers ask about this
What is the difference between response time and resolution time?
Response time is how quickly the provider acknowledges and begins to act; resolution time is how quickly generation is actually restored. A contract committing only to response has committed to starting, not to finishing. Since outage cost accrues from the moment generation stops, the resolution commitment is the one that matters, along with what happens if it is missed.
How should availability be defined in an O&M contract?
Four things have to be written down: the denominator (all hours or daylight hours), which events are excluded, who classifies the cause of an outage, and how that classification is evidenced. Without them the figure is a statement of intent rather than a measurable commitment, and grid outages or planned maintenance may be counted against the operator.
Does a warranty cover the labour to replace a failed part?
Frequently not. A warranty is a product promise; an O&M contract is a service promise. The gap is the labour to remove, ship and reinstall the replacement, which on a large plant can exceed the value of the part itself. Ask in writing whether warranty replacements include the labour to execute them.
What should happen to the monitoring data when the contract ends?
Settle it before signing. If the monitoring platform belongs to the provider, ask who holds the administrator credentials, whether the full history can be exported, and what happens to the data at term end. Losing years of generation history removes the baseline you would need to diagnose a later performance problem.
EXPLORE FURTHER