Performance Guarantees: What They Actually Cover

Owners tend to read a performance guarantee as "they promised it would make this much power". Contractors read it as "we accepted responsibility for a specific, weather-corrected technical ratio during a defined window". Those are different documents, and the gap between them is where the argument happens.

Performance ratio, briefly

PR compares what the plant produced against what it theoretically should have produced given the sunlight that actually fell on it. It normalises out the weather, which is the entire point.

A modern, well-built commercial rooftop system lands somewhere around 78-84% PR in temperate climates. Higher in cool, clear conditions; lower in hot climates where module temperature drags efficiency down. If someone guarantees 90% PR on a rooftop in a warm region, they are either using a non-standard definition or they have not thought it through.

What the guarantee excludes, by design

Standard exclusions – and most of these are legitimate:

• Grid outages and curtailment. If the utility disconnects you or caps export, that is not the contractor's doing.
• Force majeure and extreme weather. Hail damage, storms, flooding.
• Shading that appeared after design. The neighbour's extension, your own new plant on the roof, a tree nobody trimmed.
• Soiling beyond an agreed baseline, where cleaning is your obligation under the O&M split.
• Owner-caused downtime. Denied access, delayed approvals for remedial work.

The exclusion to negotiate hard is soiling. If cleaning is your responsibility and the guarantee excludes soiling losses, the contractor has an incentive to attribute any shortfall to dirt. Agree the cleaning frequency, the measurement method, and who does the soiling assessment, in the contract.

The measurement problem

PR is only as trustworthy as the irradiance measurement underneath it. Which means the pyranometer becomes a contractual instrument, and you should treat it that way:

• Class A (secondary standard) sensor per ISO 9060 for anything where real money rides on the result.
• Calibration certificate at handover, and a recalibration schedule – typically every two years.
• Mounted in the array plane, kept clean, and not shaded by anything at any hour.
• Agreed procedure for what happens when the sensor fails mid-test.

I have seen a PR dispute resolved entirely by the discovery that the reference cell had not been cleaned in eight months. The plant was fine. The instrument was reading low, which made the plant look better than it was, and the owner had been happily accepting the number.

Availability guarantees are the other half

PR tells you how well the plant converts sunlight when it is running. Availability tells you how often it is running. You want both, because a plant can hold excellent PR while being offline 6% of the year, and the PR figure will not reveal it.

Typical commitment is 97-99% availability, with grid outages and force majeure excluded from the calculation. Define whether availability is measured by time or by energy – energy-weighted is more meaningful, since an outage at noon in June costs far more than one at dusk in December.

What to do when the plant misses

Before you invoke anything, do the boring diagnostic work, because the answer is usually mundane:

1. Check the reference sensor calibration and cleanliness first. Always first.
2. Pull string-level data and look for a pattern – one inverter, one orientation, one string, or uniform across the plant?
3. Uniform underperformance points at soiling, model error or a systemic design issue. Localised points at a fault.
4. Compare against a neighbouring plant or a satellite irradiance dataset if you have access.

A guarantee is a mechanism for allocating a risk you have already quantified. It is not a substitute for knowing how your plant is behaving. The owners who get value from these clauses are the ones who were reading their monitoring data long before the dispute.