Your Solar Is Producing Less Than the Estimate. Why?

If your solar production is coming in lower than the estimate, work through the causes in order of likelihood before assuming the panels are faulty: you are probably comparing the wrong two numbers, or comparing an incomplete year, or comparing a New York winter against an annual average. Genuine equipment faults are real and worth finding, but they sit fourth or fifth on the list, not first.

The estimate you were given is almost always a modelled annual AC output for a typical weather year. Almost every disappointed comparison turns out to be against something else — a month, a DC figure, a bill, or a real year that was cloudier than typical. Here is how to check each one, in the order that finds the answer fastest.

First: make sure you are comparing like with like

Dig out the proposal and find four things: whether the number is annual or monthly, whether it is AC or DC, what percentage of system losses was assumed, and what shading loss was modelled. Then check your comparison against these three traps.

• A month against one twelfth of a year. This is the most common error by a wide margin. In New York, output is heavily seasonal — the summer months carry the year and December is the annual trough by a long way. A December that comes in far below one twelfth of the annual estimate is not a fault; it is what the model predicted, just not in a form the proposal displayed.
• DC against AC. Panels are rated in DC watts; your meter records AC kilowatt-hours after inverter and wiring losses. If someone quoted you a number derived from nameplate DC, you will never meet it.
• Production against bill savings. These are different quantities. Your bill is production, consumption, rate structure and net metering credit rules interacting. If you have added a heat pump, an EV or a new tenant, your bill can rise in a year when production was exactly as modelled — see net metering in New York for how the credit side actually behaves.

Second: are you looking at a full year that has actually happened?

The clock does not start at installation. It starts at Permission to Operate, when the utility authorises the system to run. A system energised in September has its first twelve months dominated by autumn and winter, and judging it in March is judging the worst half of the year in isolation. If the PTO wait pushed you past the strong months, that alone can account for the whole apparent shortfall — the mechanics of that wait are covered in Permission to Operate: the wait between install day and switch-on.

Wait for twelve consecutive months after PTO before you draw a conclusion. Anything shorter is not evidence.

Third: the weather year, which nobody warns you about

Solar production models run on a typical meteorological year — a synthesised long-term average dataset, not a forecast of the year you are about to have. Real years vary around it in both directions. A persistently grey spring in the Hudson Valley, or a wetter than usual summer across the Finger Lakes, will put a system below its modelled figure while every component works perfectly.

You can sanity-check this. Run your own system’s specifications through PVWatts, NREL’s free public modelling tool, using your actual array size, tilt, azimuth and location. It is the same class of model most proposals are built on, it is free, and it gives you an independent monthly expectation to compare against rather than a single annual headline. If PVWatts broadly agrees with your proposal and your production is still short, the problem is in your specific installation. If PVWatts is well below the proposal, the proposal was optimistic and that is a different conversation.

Fourth: the New York physical causes

Snow

A snow-covered array produces essentially nothing until it clears. In the lake-effect belts off Erie and Ontario — Buffalo, Syracuse, Oswego, Watertown — and in the higher parts of the Catskills and Adirondacks, that can be a meaningful number of days across a winter. Steeper arrays shed sooner; low-tilt arrays on flat roofs hold snow longest. This is normal, largely unavoidable, and worth understanding before you conclude something is broken: snow on solar panels in New York covers what is and is not worth doing about it.

Shading, especially shading that did not exist at survey

The shade analysis in your proposal was a snapshot. Trees grow, and a maple that cleared the array in year one will not in year six. Neighbouring construction, a new addition, even a new rooftop mechanical unit can put a shadow across a string at exactly the hour that matters. Walk the roof line at mid-morning and mid-afternoon in different seasons — winter shadows are far longer than the summer ones you probably remember from the site survey.

How much a given shadow costs depends on the electronics. With module-level power electronics, a shaded module drags down mainly itself. On a straight string inverter, a shaded module can pull down the whole string. Know which you have before estimating the damage.

Soiling

Pollen in spring, dust, bird droppings, and in dense urban areas particulate deposition all reduce output. In most of New York rain does the cleaning adequately on a tilted array, so soiling is rarely the main story — but low-tilt arrays drain and self-clean poorly, and a persistent stripe of debris along the bottom edge of the modules is worth removing. Localised, permanent-looking marks are a different matter and should be photographed for the installer.

Fifth: actual faults, and how to find them

Now look at the equipment. The signature of a fault is different from the signature of weather: weather affects every module and every day together, while a fault usually shows as a step change on a specific date, or one part of the array behaving differently from the rest.

• An offline inverter or a dead string. Check the inverter’s own display and fault log, not only the app. A tripped AC disconnect or a tripped breaker at the panel is a common and trivially fixable cause of a sudden drop to zero.
• A single failed microinverter or optimiser. If you have module-level monitoring, sort the modules by output and look for the outlier. If you do not, this failure can hide for a long time.
• Monitoring dropouts. A lost Wi-Fi connection produces zeros in the app while the system is producing normally. Before panicking, read the production meter itself, or compare with the utility’s recorded export.
• Inverter clipping. Systems are frequently designed with more DC than the inverter’s AC rating, and some clipping on peak summer days is intentional and was priced in. It looks like a flat top on the daily curve. Persistent, heavy clipping across many hours is worth questioning; a brief midday plateau in June is not a defect.
• Degradation. Modules lose output slowly over decades under their performance warranty. It is a real effect and an irrelevant explanation for a year-one or year-three shortfall — see how long solar panels last in New York.

What to do with the answer

If the shortfall is real and unexplained by weather, snow, shading or a partial year, go back to your contract. Look for a production guarantee or performance guarantee clause, what it promises, over what period, and what remedy it specifies. Many proposals contain an estimate with an explicit disclaimer and no guarantee at all — which is legal, and which is exactly why the distinction matters now.

Then open a documented service request with the installer in writing, attaching monthly production figures, the original proposal, and screenshots of any fault codes. If the company has stopped responding and it was a NYSERDA-associated installation, the relevant complaint routes are through NYSERDA and, for consumer issues generally, the New York State Division of Consumer Protection and the Attorney General’s office. Keeping the paper trail from the beginning is what makes any of those routes usable — a point worth carrying into your next project, and one of the reasons how to vet a solar installer in New York puts so much weight on written commitments.

Finally, if the real complaint is the bill rather than the meter, read the bill properly before blaming the array. On Con Edison in particular the interaction between supply, delivery and net metering credits confuses a great many people whose systems are performing exactly as designed — reading a Con Edison bill after solar unpicks it line by line.

What to do next

• Find the original proposal and confirm whether the estimate is annual or monthly, AC or DC, and what shading and system losses were assumed.
• Identify your PTO date and count only full months after it. Fewer than twelve is not yet a case.
• Run your array through PVWatts and compare month by month, not against an annual headline.
• Read the production meter directly to rule out a monitoring dropout, and check the inverter’s fault log.
• If you have module-level monitoring, look for a single underperforming module or string rather than a general shortfall.
• Walk the shade line in winter as well as summer, and note any tree growth or new construction since the survey.
• Check the contract for a production guarantee, then raise a written, evidenced service request with the installer.