How to read a solar production estimate

Start with what the number is, and where it sits in the contract

The headline figure is estimated first-year production in kilowatt-hours of AC electricity, the energy your home and the grid actually receive. The U.S. Department of Energy's contract guide for homeowners, now available as an archived copy, noted that a proposal's system overview may state the value of the system's annual solar production, and that the annual output in kilowatt-hours you can expect will likely appear in the financial summary alongside the savings it produces. Find the figure in both places if you can; a production number that appears in the sales pages but nowhere in the contract documents is marketing, not a term.

Hold two facts alongside it from the first reading. The estimate describes year one, the system's best year, not its average year. And it is a model result, built from a weather file and roof inputs, so it carries the model's assumptions inside it. Everything that follows in this guide is a way of opening the number up and reading those assumptions directly.

Read the monthly pattern, not just the year

Ask for production by month, not only per year. The panel sizing guide on this site makes the same request, because the shape of production across the year decides how much of it your home can actually use. A system whose output peaks with your summer cooling load offsets expensive consumed energy; the same annual total peaking when your home barely uses power exports the surplus at whatever your tariff pays for exports, which may be far less. Two estimates with identical annual figures can be worth quite different amounts once their monthly shapes meet your bills.

Lay the estimate's twelve monthly figures beside your twelve monthly bills, which the Department of Energy's contract guide noted your installer will ask for, a full year of them, when sizing the system. Months where estimated production roughly matches usage are the safe core of the savings claim. Months where production far exceeds usage are where the export rate, not the panel, decides what you earn, and those months deserve the hardest look.

Here is what that shape looks like for one modelled system: 8 kW in Boston, scaled from the 1 kW PVWatts runs in the direction section below. Facing south at 20 degrees it produces about 471 kWh in December and 1,142 in July, about 2.4 times as much. Facing west, December falls to about 300 kWh while July stays near 1,101, so about 73 percent of the west plane's yearly shortfall, 1,217 of 1,674 kWh, falls between October and March. If your quote prints twelve monthly figures, this is the comparison to make: does its winter-to-summer ratio look like a run for your roof's direction, or like a south roof's?

Table: 8 kW DC in Boston, PVWatts version 8 runs of 1 kW multiplied by 8, fixed roof mount at 20 degrees, 14 percent losses, typical-year weather. Monthly figures are rounded separately, so they may not sum exactly to the year.

Find the inputs: size, tilt, azimuth, and shade

An estimate is only its inputs restated. The inputs that move it a great deal are the system size in DC kilowatts, the roof tilt, the azimuth (which direction the array faces), and the shade assumption. Ask for each one in writing and check it against your roof: an estimate modelled on a south-facing, unshaded plane when your array will sit on a shaded east roof is not a rounding difference, it is a different system. Installers may model with NREL's PVWatts or with their own design software, and this site's sizing guidance is explicit that your roof's estimate should come from a site-specific model with your tilt, azimuth, and shade, not from a regional average.

One ratio hidden in the inputs deserves its own question: the DC to AC ratio, the panel total divided by the inverter rating. Designers may pair them above one, so an 8,000 watt DC array on a 7,600 watt AC inverter is a ratio of 1.05, and brief midday output above the inverter limit is clipped in exchange for stronger production at other hours. Clipping is normal engineering, but it should be visible in the estimate's assumptions, not discovered in your first year of monitoring data.

Direction and tilt: what the same kilowatt does on different planes

The inputs that move an estimate most are the ones a quote can quietly get wrong: which way the panels face and how steeply. To show the size of the effect, the same 1 kW system was run through PVWatts version 8 on different roof planes in two cities, with everything else held constant: standard module, 14 percent losses, DC to AC ratio 1.2, 96 percent inverter efficiency, fixed roof mount at a 20 degree tilt, and the weather file PVWatts chose for each city. A Seattle ground mount tilted to 48 degrees, about the city's latitude, was added for contrast.

Two readings matter for a quote. First, a west or east plane is not a disaster: at these inputs a west roof gave 84 percent of the south figure in Boston and 87 percent in Seattle, and an east roof 84 percent in Seattle. A north plane at 20 degrees fell to 68 percent in Seattle, which is why designers avoid it. Second, those gaps are exactly the size of error you get if a quote models a south plane and the panels go on the west one. On an 8 kW Boston system, south comes to about 10,415 kWh a year and west about 8,741, a gap of about 1,674 kWh, or 16.1 percent. The ground mount guide covers when a separate array is worth what it costs.

Rerun the estimate in PVWatts, input by input

PVWatts asks for the same handful of inputs a quote's design page should already state, which makes it the natural audit tool. The version 8 programming interface documents each one: system size as nameplate capacity in kilowatts, a module type (standard, premium or thin film), an array type (fixed open rack, fixed roof mounted, or one of the tracking options), system losses as a percentage, tilt in degrees from 0 to 90, and azimuth in degrees, the compass direction the panels face. Two more inputs have defaults in the interface documentation: a DC to AC ratio of 1.2 and an inverter efficiency of 96 percent at rated power. If the quote names different values, use the quote's, because the point is to reproduce its model, not to replace it with yours.

1. Enter the street address. PVWatts picks the nearest weather file and reports which one it used; note it.
2. Copy the DC system size from the quote's design page: panel count times panel watts, divided by 1,000.
3. Choose the module type and the array type. A flush rooftop array is fixed, roof mounted; a ground array is fixed, open rack.
4. Enter the losses the quote used if it states them. If it does not, run 14 percent and ask the installer for theirs.
5. Enter the tilt (roof pitch in degrees) and the azimuth: 180 is due south, 90 east, 270 west, 0 north.
6. Set the DC to AC ratio from the quote: DC size divided by the inverter's AC rating.
7. Run it, then compare the annual total and, more usefully, the twelve monthly totals with the quote.

If the array sits on two or more roof planes, run each plane separately and add the results; a single run at an averaged direction can hide a large difference. If the shade report shows losses, raise the shading loss for that plane rather than leaving the 3 percent far-horizon default in place, and price the loss with the shade guide. Save the printed inputs with your run. A number you can reproduce is the only kind worth arguing about with an installer.

When your run and the quote disagree, work through the gap in a fixed order rather than arguing about the total. First confirm you entered the same DC size, tilt and azimuth the quote used; a wrong plane explains more than anything else. Then compare losses, because a quote at 10 percent and a run at 14 percent differ before any weather enters. Then compare the weather source each model names, since the two may not be using the same file. Whatever gap is left after those three is the part the installer has to explain, and asking for that explanation in writing, against your printed inputs, gives both sides something specific to settle.

The losses line: where modelled energy goes missing

Between the sunlight on the panel and the AC energy at your meter, a model subtracts losses: inverter conversion, wiring, soiling, mismatch, temperature, and availability. Production models itemize these, and the total is one of the most quietly adjustable figures in the whole estimate. An estimate built on unusually low loss assumptions produces an unusually flattering year-one figure without changing a single panel. Ask what total system losses the model used and whether each loss category matches the tool's standard values or was adjusted, and by whom.

This is also the fastest cross-check between quotes. If two installers model the same roof at the same size and one estimate is far above the other, compare their loss totals and shade assumptions before you compare their prices. The higher estimate may well be explained right there, in inputs you can question, rather than in better equipment.

The 14 percent losses line, item by item

The single losses percentage in a PVWatts run is a bundle. The PVWatts Version 5 manual, the national laboratory's technical reference for the calculator, lists the ten loss mechanisms it covers and a default value for each, and explains that the total is not their sum: each loss reduces what is left after the others, so the reductions multiply. The ten defaults add up to 15 percent, but multiplied out they remove 14.08 percent, and the manual states the default total as 14 percent. It is also the losses figure behind the eight-city sizing runs on this site and every run on this page.

Three lines deserve your attention because they are site facts, not equipment facts. Shading: the manual says the 3 percent default represents blocking of the horizon by faraway features such as large buildings or mountains, and that it should be increased for a system shaded by nearby trees or structures, using shading software or an on-site survey. A quote that keeps 3 percent on a roof under a tall oak has not modelled your roof. Snow: the default is zero, which flatters a run for a snowy climate. Soiling: 2 percent assumes rain does the cleaning; dusty, dry sites lose more between rains.

Read the total in the other direction too. An estimate built on total losses well under 14 percent, with no reason given, produces a bigger year-one number from the same panels. There can be good reasons, such as a shade-free site with a measured survey, but they should be written down. Ask for the losses the model used, item by item if the software reports them, and for who changed any default.

Degradation: year one is the best year

Panel output declines slowly with age, and a careful estimate states the annual degradation rate it assumes. The arithmetic compounds quietly. On an illustrative estimate of 9,000 first-year kilowatt-hours with a stated rate of 0.5 percent a year, year ten production is 9,000 times 0.995 to the ninth power, about 8,603 kilowatt-hours, roughly 400 kilowatt-hours a year less than the headline figure the savings were sold on. The rate used here is an illustration of the math, not a claim about any panel; the rate that matters is the one printed in your quote and warranted by your panel manufacturer.

So read any savings or payback figure in a quote with one question attached: was it computed on year-one production held flat, or on production declining at the stated rate? A payback period built on flat year-one output overstates every later year. The honest version is slightly longer, and it is the only version worth financing.

One year is not the estimate: weather moves the total

A PVWatts estimate runs on a typical meteorological year, and an installer's model may too: a weather file assembled to represent normal conditions at a location, not any real year. The PVWatts runs on this page each reported their weather source as an NSRDB typical-year file. Real years scatter around that typical value. The PVWatts web calculator shows this directly: next to its result it gives an expected range which, the calculator says, "is based on 30 years of actual weather data at the given location and is intended to provide an indication of the variation you might see." The calculator notes that range is computed for a fixed open-rack system, so treat it as a guide to the spread, not an exact band for a roof mount.

That changes how to judge the first year of monitoring data. A system that delivers less than its estimate in a dull year may be working perfectly, and one that beats it in a sunny year may be hiding a fault. Before calling a shortfall, check whether the shortfall is spread across the year (a weather or soiling pattern) or concentrated in a few months or hours (a shade, inverter or outage pattern). The second kind is the one worth a service call. The first kind is worth checking against the next year before anyone argues about it.

Estimate or guarantee: read the exact words

The Department of Energy drew the line plainly in its contract guide (archived copy, December 2024 version): a quote's bill savings may be presented as estimated or guaranteed, or both, and the two are different legal objects. A production estimate, however carefully modelled, promises nothing. A production guarantee is a contract term, and like every term it has edges: what is guaranteed (annual kilowatt-hours, or savings in dollars), over what period, measured by whom and with what equipment, and what happens in a shortfall year, typically a payment, a repair obligation, or nothing enforceable at all. Read those edges before you weigh the guarantee in a comparison, because a guarantee with a weak remedy is an estimate wearing a costume.

The quote red flags guide covers the sales patterns that gather around this number, including the guaranteed-savings pitch built on an estimate nobody will warrant. If a proposal's language blurs the two, ask the installer to point to the contract clause that guarantees production, in the contract, before you sign anything.

Reading a production guarantee clause, line by line

If a contract does guarantee production, the clause has six working parts, and each one decides whether the guarantee is worth anything. None of them appears in the headline estimate, which is why they are worth listing before you sign.

A labelled illustration shows why the guaranteed level matters most. Take the 9,400 kWh estimate from the example on this page, a guarantee at 90 percent of it, a payment of 15 cents per kilowatt-hour short, and a year that delivers 8,200 kWh. The guaranteed floor is 8,460 kWh, so the payment covers 260 kWh, or $39. The full gap from the estimate is 1,200 kWh, worth $180 at the same rate. The guarantee paid about a fifth of the gap. That is not dishonest; it is what a 90 percent guarantee means. It is also why the guarantee should not be read as a promise that the estimate will happen.

Two further checks belong with the clause. Find out whether the guarantee transfers to a buyer if you sell the house, and whether it survives the installer: a payment obligation from a company that has closed is worth what its assets are, as the guide to installers that go out of business explains. And compare the guaranteed level with your own PVWatts rerun. A guarantee set well below a careful independent model costs the installer little to offer.

Cross-check it yourself before you sign

You do not need the installer's software to test the number. Run your own roof through NREL PVWatts with your tilt, azimuth, size, and shade, and compare its annual figure with the quote's. A quote within a reasonable distance of your own model run is a quote whose inputs you can discuss; a quote far above it has explaining to do. Then reduce every surviving quote to cost per first-year kilowatt-hour, the unit the cost worksheet in the companion guide builds step by step, and line them up in the Quote Comparison tool. Production you cannot verify should never be the reason one quote beats another.

The quickest comparison across quotes is specific yield: first-year kilowatt-hours divided by DC kilowatts. In the example on this page, Quote A's 10,200 kWh on 8 kW is 1,275 kWh per kW, and Quote B's 9,400 kWh is 1,175. Two installers modelling the same roof should land close to each other; a 100 kWh per kW gap on identical size is the gap their inputs have to explain. For scale, the sizing guide prints PVWatts factors for a south-facing 20 degree roof in eight cities, from 1,088 kWh per kW in Seattle to 1,755 in Phoenix. A quote claiming more than a south roof's figure for your city on a west roof needs a reason in writing.