Why “Average” Response Time Is the Wrong O&M Metric
Every O&M SLA I have ever reviewed includes some version of this metric: “Average alarm-to-response time: under 4 hours.”
It sounds rigorous measured, tracked, and reported quarterly. Operators take pride in hitting it.
It is also a nearly useless indicator of actual operational performance. And the fact that it dominates O&M contracts is costing asset owners significant money while giving them the comfortable illusion of accountability.
Let me explain why, and what to measure instead.
The problem with averages
Averages obscure the information you actually need. That is true in statistics generally, and it is especially destructive in O&M performance measurement.
Consider two O&M operators, both hitting “average alarm-to-response under 4 hours.”
Operator A responds to 490 low-value alarms — sensor noise, transient faults, communication glitches — within 2 hours each. He responds to 10 high-value alarms — inverter failures, string faults, BESS anomalies — in 18 to 26 hours each. Average: 2.3 hours. SLA: hit.
Operator B triages all alarms by revenue impact. He responds to the 10 high-value alarms within 45 minutes each. He processes the 490 low-value alarms within 6 hours, most automatically. Average: 5.1 hours. SLA: missed.
Which operator is delivering better financial outcomes for the asset owner? Operator B, by a significant margin Operator A’s average response time looks better. Operator B’s revenue recovery is better.
The average metric rewards operators who respond quickly to noise and slowly to what matters. It punishes operators who correctly prioritize by revenue impact.
The revenue-weighted alternative
The metric that actually matters is revenue-weighted mean time to resolution, or RW-MTTR: the average time to resolve faults, weighted by the euros per hour revenue impact of each fault.
This metric has completely different properties than simple average response time:
- A fast response to a €3/hour communication fault contributes minimally to the weighted average.
- A slow response to a €180/hour inverter fault dominates the weighted average.
- Operators are incentivized to prioritize the most expensive faults, not the highest volume of fast responses.
The difference in practice is meaningful. A portfolio in RW-MTTR is actively managed will typically recover far more annual value than one where only simple response averages are tracked is because the faults that cost the most money get addressed first.
The second wrong metric: uptime percentage
“Site availability: 98.7 percent.” I see this everywhere. It is the second most misleading metric in O&M SLAs.
Availability percentage is calculated against total hours. A site that is offline for 113 hours over the year hits 98.7 percent availability regardless of whether those 113 hours happened at 3 AM in January during a grid curtailment event or at 2 PM in July during peak irradiance and peak energy prices.
The availability percentage is identical. The financial impact is radically different.
What matters is production-weighted availability: hours offline weighted by the expected generation energy price at the time of each outage. A one-hour outage at peak summer generation costs far more than a one-hour outage in December at 3 AM. Any SLA that treats these equally is measuring the wrong thing.
For a useful benchmark on availability language, the IEA PVPS Guidelines for Operation and Maintenance of Photovoltaic Power Plants in Different Climates are a good reference.osti+1
The third wrong metric: number of alarms resolved
Some O&M operators report “alarms resolved per week” as a productivity metric. This is perhaps the most misleading of all.
Resolving 500 alarms per week sounds like high productivity. But if 450 of those were automatic resolutions of sensor noise events, and the 50 genuine faults took an average of 3 days to reach resolution, the productivity number tells you nothing useful.
The question is not how many alarms were processed. The question is: of the alarms that represented real revenue loss, how quickly did the revenue loss stop?
What ClearSpot measures instead
ClearSpot’s AI Agents for Solar O&M track O&M performance using the metrics that actually reflect financial outcomes.clearspot
Revenue-weighted MTTR
Every fault is tagged with its euros per hour revenue impact at the time of detection. Resolution time is tracked and weighted. The portfolio dashboard shows RW-MTTR by site, by fault category, and trending over time. O&M managers see, in real time, whether high-value faults are resolved faster or than last quarter.
Production-weighted availability
Site downtime logged with the concurrent irradiance and energy price at each outage event. Availability reported in two forms: standard percentage and production-weighted percentage. That makes the difference between a low-value outage and a high-value outage visible immediately.
False positive dispatch rate
Every dispatch are tagged as confirmed fault or false positive rate is tracked and reported. An increasing false positive rate is a leading indicator of alarm classification problems, which typically precede genuine fault coverage gaps.
€/MWh O&M cost
Total O&M cost divided by total energy produced is the ultimate efficiency metric. It normalizes O&M spend against the production it is protecting and allows meaningful comparison across sites with different capacities, ages, and locations.
These are the metrics the AI solar performance monitoring stack puts in front of asset managers, because they are more honest about where money is being lost and recovered.clearspot
Why the drone loop matters
The agents and drones closed loop is what makes these metrics trustworthy. Drone inspection missions provide the physical confirmation that cleans the underlying data.clearspot
An RW-MTTR figure means little if its calculated against a mix of real faults and sensor noise. fault in ClearSpot’s system is visually confirmed before they enter the resolution tracking system, the numbers are clean.
That same logic is part of the broader Solar Asset Management Software layer, which helps normalize equipment-level performance across the portfolio.clearspot
The conversation to have with your O&M provider
Next time you review your O&M SLA, ask three questions:
- “Can you give me the revenue-weighted mean time to resolution, not the simple average, broken down by fault type?”
- “What was our production-weighted availability last quarter, not the standard percentage?”
- “What percentage of dispatches last quarter were false positives?”
If the answers are not readily available, that is the answer.
Closing thought
The right O&M metric is not the one that makes the report look neat. Its the one that tells you where money is lost and how fast its recovered.
If your SLA cannot separate high-value faults from low-value noise, it is measuring activity, not performance.
FAQs
What is RW-MTTR?
RW-MTTR means revenue-weighted mean time to resolution. It measures fault resolution time based on the revenue impact of each fault.
Why is average response time misleading?
Because it treats all alarms equally, even though some faults cost far more money than others.
What is production-weighted availability?
Its availability weighted by expected generation and energy price at the time of each outage, so peak-hour downtime counts more than low-value downtime.
How does ClearSpot improve O&M measurement?
ClearSpot’s AI Agents for Solar O&M and AI Solar Performance Monitoring track revenue-weighted MTTR, production-weighted availability, false positive dispatch rate, and €/MWh O&M cost.clearspot+1
Which external benchmark is useful for availability language?
The IEA PVPS O&M guidelines are a strong reference for availability and maintenance benchmarks.