Four checks answer this, in order of how fast they are: the inverter status light, this month against the same month a year ago, the utility meter on a sunny noon, and your annual total against the installer's year-one estimate. The first takes ten seconds. The last one catches the failures the other three miss.
Panels don't announce their own failure. A single dead string keeps the inverter online, the app still shows a graph, and the only symptom is a bill that crept up.
Check One: What Is the Inverter Telling You?
The inverter is the only part of the system that reports on itself. Find it, usually in the garage or on an outside wall, and look at the status indicator.
| What you see | What it means | What to do |
|---|---|---|
| Solid green, or a normal-state icon | Producing and exporting | Nothing |
| Flashing green or amber | Starting up, or in a reduced state | Recheck in an hour of daylight |
| Red, or an error code | Fault, grid problem or shutdown | Note the code, then call |
| Nothing at all, at midday | No DC input, or the inverter is dead | Call the installer |
A blank screen at noon on a clear day is the one unambiguous failure signal on this list. Everything else needs a second data point before you can call it.
Worth knowing: your inverter shuts itself down during a grid outage, by design and by code, unless you have specific backup hardware. That's an anti-islanding rule protecting line workers, not a fault. Our guide on reading solar inverter data unpacks what the rest of the numbers on that screen mean.
Should You Compare Against Last Year, Not Last Month?
This is where most people get it wrong. December output can be less than half of June output on the same roof, so a month-over-month drop tells you about the calendar, not about your panels. In snow country a covered array adds its own dip; here is how big it gets.
Open your monitoring app and pull up the same calendar month from a previous year. Sun angle, day length and typical weather all roughly hold. Now the comparison means something:
- Within about 5 percent: normal, including a year of expected degradation
- Down 10 to 15 percent: soiling, new shading, or one underperforming string
- Down 25 percent or more: a string or an MPPT channel is likely offline
Panels themselves degrade slowly, around 0.5 percent a year by the NREL fleet median, so genuine ageing never explains a double-digit drop. Something else does. Tree growth is the most common culprit nobody suspects, because it happens over years and nobody photographs their roofline in spring.
Why Watch the Meter at Midday?
No monitoring app? The utility meter still works as a crude instrument. Pick a clear day near noon, switch off the big loads, and watch the display. A bidirectional meter should show export, or a net figure that stops climbing.
That test proves AC is reaching the panel and flowing the right way, which is more than the inverter light proves on its own. It won't tell you whether you're producing 6 kW or 3 kW, so treat it as a pass or fail, not a measurement.
What Does the Number on Your Contract Say?
Every reputable installer supplies a year-one production estimate, in kWh, before you sign. That's the number your system is actually accountable to, because it already accounts for your pitch, orientation and shading. Most people never look at it again.
For a rough sanity check without the paperwork: US mid-range yield runs about 1,400 kWh per installed kW per year, so an 8 kW array should land near 11,200 kWh. Cloudy states run closer to 900 per kW, the desert southwest closer to 1,750.
What's the shortfall actually costing? At the 18.16 cents per kWh US households averaged over the first half of 2026 (EIA), a 15 percent gap on an 11,200 kWh system is about $305 a year. Quietly, every year, until someone checks.
I'd put an annual reminder in the calendar for this one. It's fifteen minutes in January, and it's the only check that catches slow decline rather than sudden failure.
Which Pattern Is Your Production Curve Showing?
When the checks say something is off, the shape of one clear day tells you what kind of off. Open the daily graph in your app, pick a sunny day, and set it beside the same date a year ago. Six shapes cover most of what you'll see.
| What the curve does | Likely cause | A fault? |
|---|---|---|
| Flat plateau at midday, same ceiling every bright summer day | Inverter clipping | No, a design trait |
| A notch at the same clock time on clear days, drifting a few minutes each week | Shade from a chimney, vent pipe or tree | Only if it keeps growing |
| Whole curve a similar share lower all day, worse over dry weeks, better after rain | Soiling | Clean, or wait for rain |
| Normal shape, but about a third lower from one day onward | One string or MPPT input offline | Yes, call the installer |
| Dip on the hottest afternoons, recovering by evening | Inverter heat derating | Check ventilation |
| Everything a little lower than last year, no sudden change | Ordinary ageing | No |
Clipping scares people for no reason, so take it first. Picture an imaginary 8 kW array on a 6 kW inverter, a ratio of 1.33. On a bright June noon the panels could deliver 6.5 kW or more, the inverter converts 6, and the graph goes flat-topped at 6.0 kW. It looks like a fault. It isn't. Designers accept it because the clipped sliver is small over a year, and our optimization guide puts inverter clipping at 0.5 to 3 percent of annual energy. Worry only when the ceiling sits well below the inverter's rated AC output. That points to heat or a voltage-window problem instead.
Shade makes a vivid notch and a modest annual loss. A chimney shadow crosses a few panels for a few hours, and in a series string the shaded panel drags its neighbours down while it lasts. The same guide covers the ZHAW study: under light-to-medium chimney shade a plain string inverter still returned about 90.7 percent of unshaded energy. Photograph the roof at the notch time each season. Tree growth is the one that keeps getting worse.
Soiling and shade both have a yardstick. NREL's PVWatts model, the free calculator many installers quote from, assumes 2 percent lost to soiling and 3 percent to shading inside a default 14.08 percent total system loss (PVWatts manual). A few percent from either is the model working as designed. A double-digit gap isn't.
The step change is the one pattern that's nearly always a fault. If a system with three equal strings loses a third of its output overnight and holds there, an input has dropped out. Note the date, check the inverter log for the string or MPPT, and hand both to the installer.
I'd save a screenshot of one clear day each season. It takes a minute, and it turns "it seems lower" into a comparison someone can act on.
What If Your Panels Aren't Working?
If two or more checks fail, stop diagnosing and call the installer while the workmanship warranty still applies. Note the inverter error code, the month-on-month-last-year figures, and the date you first noticed. That turns a vague complaint into a service ticket someone can act on.
If the panels are visibly damaged, cracked glass, browning, or a scorched junction box, our damage identification guide covers what's a warranty claim and what's an insurance one.
Summary
Working and working properly are different questions. The inverter light answers the first in ten seconds. Only the year-on-year comparison and the contract estimate answer the second, and those are the ones that catch a dead string quietly costing you a few hundred dollars a year. Check the light whenever you walk past it, and check the numbers every January.