How to tell if your solar panels are actually underperforming

By Olivier Beauchemin · 2026-08-18 · v1

You've been checking your solar app for weeks, trying to work out whether those numbers are good or bad, and it won't tell you. That isn't you missing something obvious. Your app reports what your system produced. It was never built to tell you what your system should have produced, and only that second number turns a suspicion into something you can act on.

So let's work it out, in the order that rules the most causes out fastest.

First, the comparisons that won't settle it

Comparing to last month

Your production swings by a factor of three or more across the year in most of the United States. Put June next to November and you've learned that the earth is tilted, not anything about your system.

Comparing to your neighbor

Your panel count, your panel model, your roof's orientation and pitch, your shading, your inverter type: every one of them differs from theirs. Two systems on the same street can honestly differ by thirty percent, so their good year tells you nothing about your bad one.

Comparing to the proposal number

This is the one almost everyone reaches for, and it gets closer without ever quite getting you there. Your proposal modeled a typical year. You lived through a specific year. If your twelve months were cloudier than typical, you'll land below the proposal with nothing at all wrong. If they were sunnier, you can beat the proposal while a whole string sits dead on your roof.

The test that works

Work out what your specific system should have produced under the weather that actually happened where you live, then put that next to what it did produce, across a full twelve months.

That expected figure isn't guesswork. The National Renewable Energy Laboratory publishes an open-source model built for this, and it's the same modeling family your installer used to write your proposal. What changes is the inputs. A claim-grade expected number uses:

• Your actual panel count and panel rating
• The real direction and angle of every plane on your roof, each one modeled separately
• Measured shading, including whatever has grown since your install
• Historical weather for your location for the specific months you're asking about, not a long-run average weather year
• Realistic system losses, including inverter efficiency and soiling

Run that and you've got a number you can defend. The gap between it and your metered production is the only figure worth arguing about. The full calculation is documented in how OwlWatt measures your system and in more technical depth on our methodology page.

What the gap is actually worth to you

Before you go further, do the arithmetic on your own numbers. Say your system is expected to make 10,000 kWh in a year. That's a round figure I'm using as an illustration, so put your own estimated annual production in its place, and take your own rate off your electric bill instead of borrowing my $0.30 per kWh.

OwlWatt costs $9.99 a month, or $87.89 a year. Even the top row, a 5 percent shortfall you'd never spot by eye, is $150 a year you can't see. At 10 percent you're looking at roughly three and a half times what the subscription costs.

Now the honest part, before you spend anything. OwlWatt doesn't get that money back for you. It measures your gap and documents it so you have something to hand over. Whether a claim succeeds depends on what your contract says and, if it goes that far, on an attorney. So read those figures as what's at stake for you and what you currently can't see, never as money you're going to receive. Your honest result may also be that there's no meaningful gap at all, and knowing that is worth something too.

A worked example from a public forum

A homeowner posted a thread on the solarpaneltalk forum with a very clean version of the problem you're in. Their words:

"I have been producing power for 1 year now and by looking at my Enphase Envoy I have produced a total of 7,798.73 Kwh, but what I was guaranteed was 9,996 Kwh annually."

That's a 22 percent shortfall against a stated guarantee, on hardware that reports panel by panel. It looks open and shut. It wasn't, and the reason should interest you.

The forum's experienced posters asked for things the homeowner didn't have. Was the array all one orientation? It wasn't. Panels faced east, west, and south, and a proposal modeled as though the whole array faced south will overstate output substantially. Had the shading been assessed, including trees that were smaller when the contract was signed? What did the guarantee clause actually set as the threshold? Was the monitoring hardware itself reporting correctly, or was a branch of the system quietly missing from the total?

The thread never resolved. The honest conclusion was that a 22 percent gap might be a legitimate claim, or might be an accurate picture of a multi-direction array with shade the sales model ignored. The homeowner couldn't tell, and neither could volunteers working from forum posts.

That uncertainty is what's actually wrong here. It isn't the gap that traps you, it's that you can't tell whether the gap is something your contract makes collectible.

The four things behind most real gaps

1. Your orientation and tilt were modeled wrong

Your roof probably has more than one plane, and fast sales modeling flattens them into a single average. East and west planes make meaningfully less energy over a year than south-facing ones do. If your proposal assumed the good case, your shortfall was designed in before anyone set a panel on your roof.

2. Shade, especially shade that grew

The shade study done when you signed describes your trees as they were that day. Five years of growth is a real effect, and often a large one. Partial shade on a string-inverter system costs you far more than the shaded fraction suggests, because one shaded panel can drag its whole string down with it.

3. Soiling and snow

Dust and pollen build up on your glass, and so do agricultural and wildfire particulates. Most models assume a modest soiling loss and count on rain to wash it off. If you're somewhere dry or dusty, your real losses can run several points above that assumption.

4. Equipment faults that never set off an alarm

This is the one that most rewards checking. A failed microinverter, a tripped breaker on one subpanel, a communications gateway that quietly stopped uploading part of your array, a firmware setting nobody enabled. Your system can run that way for months. In one Massachusetts case reported in the consumer press, a technician eventually found that panels had simply never been wired in.

None of those four has to turn your monitoring app red. Green means the parts that are reporting are reporting. It doesn't mean everything that should be on your roof is there and working.

What evidence your claim actually needs

If your gap survives the modeling, you'll have to hand someone a package. At minimum:

• Your metered production, month by month, for the full guarantee year, exported from your monitoring system rather than screenshotted
• An independently computed expected-output figure for the same months, with the inputs and the method disclosed so the other side can check it
• Your guarantee clause itself, with the threshold and measurement window it uses, plus any exclusions it carves out
• Your system as actually built: your panel count and model, plus the orientation and tilt of every plane on your roof
• A shading assessment current to your claim period
• A dated record of every time you raised this and what you got back

If your system reports panel by panel, as Enphase microinverter systems do, export your per-panel history alongside your system total. One panel sitting at a fraction of its neighbours across the same hours is the clearest single piece of evidence you can hand over, and it survives a disagreement about modeling assumptions.

So where does this leave you

You can gather five of the six items on that list yourself. The sixth, an independently computed expected-output figure with disclosed inputs, decides the outcome and no homeowner can produce it alone. Without it you've got a complaint. With it you've got a claim.

That number is what OwlWatt produces. It reads your existing Enphase data, models what your specific array should have produced under the weather that actually happened, and reports your gap every month with every input auditable. We're independent of your installer by design, which is what makes the figure worth anything when you're in a dispute.

Check your own system free, or read the methodology first if you'd rather see the arithmetic before you trust it.

We can tell you how easy this is to get wrong, because we got it wrong ourselves. Our own expected-output model used to treat a multi-plane roof as a single orientation. On an array where most of the panels faced east rather than south, that one assumption overstated what the system should have produced by roughly 10%, which made a perfectly ordinary system look considerably worse than it was. We found it, measured it and fixed it. If a model built specifically to check this can get it wrong, a sales estimate written to close a deal can certainly get it wrong about your roof.

If your installer has stopped answering you, the escalation path and what to collect first is covered in what to do when your solar company stops responding. If your bill is the symptom rather than your production number, start with why your electric bill is still high after going solar.