Why a quote says $3.50 and a marketplace says $2.62 per watt: four benchmarks, one reconciliation

Four benchmarks and what each one counts

The table sets each benchmark against the questions that decide whether it can be compared with your quote. A row that says no or not stated means the source does not say, not that the answer is no.

Why $2.62 and $3.2 to $5.5 can both be right

Berkeley Lab says outright that its national median is higher than a number of other common benchmarks and that those benchmarks generally align more closely with the 20th percentile of its data. It gives the reasons: different methods, data sources and definitions, and for loan-financed systems, reported prices that likely include dealer fees, which add anywhere from 5 to 50 percent to the total up-front price paid and are typically not included in other benchmarks. A separate slide in the full deck lists price versus cost, quotes versus actuals, mark-ups, system design and the scope of costs included as the sources of divergence.

Five differences are candidates for the gap, and the sources quantify only the first. The first is financing: a price paid on a loan can carry a dealer fee that a cash quote does not. The second is scope: a price paid can include a roof repair, an electrical panel upgrade or a change order added after signing, which a quote written before the site survey may leave out. The third is selection, which neither source measures: marketplace shoppers compare several quotes, and the median of competitive quotes need not match the median of what people paid. The fourth is time, since Berkeley Lab's prices are for 2023 and the marketplace median is for the first half of 2026, and the fifth is the system, because a median across all installations includes small systems, which cost more per watt, and battery-paired installs.

None of this makes any number wrong. It means the useful question is which one describes the transaction you are in, and that comes down to two answers: how you are paying, and what is in the scope.

What the lab says each of the other benchmarks counts

Berkeley Lab's full report puts the main benchmarks on one slide, and its notes are the most compact statement of what each number measures. Its own data are the median and the 20th and 80th percentile values among projects installed in 2023. The company figures for SunPower and Sunrun are average costs from their 2023 quarterly shareholder reports, covering installation, sales and general and administrative costs, averaged across quarters. The EnergySage figure is the median of price quotes issued in 2023, for either cash-purchased or loan-financed stand-alone systems, calculated by the lab from data EnergySage provided. NREL's figure is the modeled market price for a roughly 8 kW residential system. And the Wood Mackenzie figure is a modeled turnkey price averaged across quarters.

The lab's conclusion from that slide is specific. On the residential side, its national median installed price is similar to the average costs the large installers report, while the other benchmarks align more closely with the 20th percentile of its own data. It lists the usual reasons the figures diverge: price versus cost, quotes versus actuals, mark-ups, system design, and the scope of costs included. So a $2.62 solar-only median from a marketplace and a $2.95 modeled benchmark are not in conflict with a $3.2 to $5.5 band; they sit just below its lower edge, near the 20th percentile the lab says other benchmarks track, and the upper part of the band is where dealer fees, ancillary work and small systems live.

One more distinction from the same material affects batteries. The lab reports that prices for paired solar-plus-storage systems have generally declined over time even though reported battery costs have been flat or rising in real terms, and notes that these data come mostly from California's self-generation incentive program, so they are not a national guide to what a battery adds to a quote. That is why the adder used in the worked example is the regression effect, $1.4 per watt of PV, and not a battery price quoted from a vendor.

A 12 kW typical system against a 7.4 kW median one

The sources do not describe the same size of house. EnergySage's cost page says the typical home on its marketplace needs about 12 kW of solar and prices that system at an average of $31,135 before incentives, with a low end of $27,526 and a high end of $33,869 for the same size. Dividing, those are $2.59, $2.29 and $2.82 per watt. Berkeley Lab's median residential system installed in 2023 was 7.4 kW. A 12 kW system is about 62 percent larger than that, and the lab's regression says moving from the 20th to the 80th percentile of system size lowers the price by $0.7 per watt, with the smallest residential systems costing about $1.0 per watt more than the largest.

Size is a real reason for a lower per-watt price, because several costs, such as permitting, the inspection, the monitoring and the design work, do not scale with the number of panels. It also means a household that needs only 5 kW cannot expect the marketplace median. EnergySage's page says solar panels themselves are just 12 percent of the total cost of an installation; the rest is the installer's labor, equipment other than panels, and soft costs, which are the part that does not shrink when the panel count does; what happens to those lines when the owner does the installing is priced against the DOE model in the DIY solar rules and cost split.

The practical rule is to compare like with like on size first. If your roof supports a 6 kW system, the marketplace median is the wrong comparison and the lab's note that the smallest systems cost about $1.0 per watt more is the better guide. If your quote is for 14 kW, a median from 7.4 kW systems overstates the price you should expect.

The dealer-fee multiplier on the marketplace median

Berkeley Lab's 5 to 50 percent range lets the marketplace median be restated as what a loan-financed buyer might see in the quote's total. The multiplier is the one the lab gives for the price paid on loan-financed systems, applied here to the $2.62 solar-only median as an illustration. The dealer fee mechanics, and the way to compute the true cost of credit, are in the dealer-fee guide.

The 25 percent row lands just above the bottom of Berkeley Lab's 20th-to-80th percentile band, and a fee adding about 13 percent is all it takes to lift the marketplace median to the DOE line of $2.95 (2.95 divided by 2.62 is 1.13). A loan with even a modest fee turns a marketplace-median cash price into one that looks like a benchmark, so a quote that matches the DOE line on a loan may be, at the marketplace median, a cash price with a fee inside it.

Two ways to say a dealer fee, and why they differ

Fees are described two ways, and mixing them up shifts a comparison by several points. Berkeley Lab says a dealer fee adds 5 to 50 percent to the total up-front price. The CFPB's 2024 solar financing spotlight, cited in the cash against loan guide, reports hidden fees typically running 10 to 30 percent of the cash price. A lender, by contrast, usually states the fee as a share of the amount financed, because the fee comes out of the loan the installer receives. If the fee is f of the loan, the installer nets the cash price C from a loan of C divided by (1 minus f), so the amount financed is above the cash price by more than f.

The table converts one into the other for a $23,600 cash price, the DOE line for 8 kW. A fee of 20 percent of the loan puts the amount financed 25 percent above cash. The numbers matter when a quote prints only the financed amount: divide by the right factor to recover the cash-equivalent price, and the per-watt comparison above becomes honest.

Read the last column against the first. A quote that says $3.93 per watt and a quote that says $2.95 for the same equipment can be the same price, with a 25 percent fee between them, and only the loan paperwork says which. The request that settles it is simple: ask each installer, in writing, for the cash price and the financed price of the same system, the DC size in watts, and the dollar fee. If the installer will not state the cash price, that refusal is itself the information.

What moves a quote's price per watt

Berkeley Lab's regression on host-owned residential systems installed in 2023 puts a dollar value on features of the quote, holding the others constant. Battery storage added $1.4 per watt, new construction subtracted $0.7, moving from the 20th to the 80th percentile of system size subtracted $0.7, and DC optimizers and ground mounting each added $0.4. The market and installer variables it tested were each under $0.2 per watt and mostly not statistically significant. The lab also notes a $2 per watt range across state fixed effects, which the model could not explain with its other variables, suggesting local pricing drivers like cost of living, retail rates, incentives, solar resource and permitting.

The simple descriptive spreads agree. The difference in median price between the smallest and largest residential systems was about $1.0 per watt, state-level medians for residential ran from $3.2 to $5.2 per watt, and among the 100 largest residential installers, installer-level medians ran from $2.6 to $5.9 per watt, with about a third below $4, a third between $4 and $5, and the rest above $5. The median residential system installed in 2023 was 7.4 kW.

Normalizing three quotes before you compare them

The method is three steps. Divide the total price by the system size in watts DC, not AC. Remove the features that move the price, using the adders above. Then place the result against the benchmark that matches your payment method. The table works three invented quotes, with every input printed, so the arithmetic can be repeated on a real one.

Quote A is an 8.2 kW roof system at $29,520, solar only, paid in cash. Quote B is a 10.4 kW system with a 13.5 kWh battery at $58,240. Quote C is a 6 kW system at $24,000, solar only, financed with a loan that the paperwork shows has a 25 percent fee inside the amount financed.

Reading across, quote C looks more expensive than quote A as quoted and is the cheapest once the fee is separated, which is the point of normalizing. Quote B looks absurd at $5.60 until the battery is removed, at which point it sits mid-band. Quote A is a middle-of-the-road cash price that the marketplace median and the DOE line would both call high, 37 and 22 percent above them. Whether $3.60 is fair for A depends on its state, since state medians run from $3.2 to $5.2, and on its installer, which is why the next step is more quotes, not a conclusion.

Which line to hold which quote to

A cash quote for solar only is best compared with the $2.62 solar-only marketplace median first, since both are quote-side numbers written before change orders, and then with the DOE line as a modeled cross-check. A quote that will be financed should be compared by its cash-equivalent price, which means removing the dealer fee, as in quote C, because Berkeley Lab's price-paid data and the DOE line do not treat the fee the same way. A quote with a battery needs the battery adder removed before any comparison, and the battery itself deserves its own test against the DOE gap in the battery guide.

A lease or PPA quote has no purchase price, so none of these lines applies. EnergySage reports third-party-owned options in dollars per kilowatt-hour and says the share of installers quoting them rose from about 14 to 41 percent in the first half of 2026, a gain it partly attributes to the newness of the product on its platform. A kilowatt-hour price cannot be set against a per-watt benchmark without a production estimate, and the escalator and term matter more than the starting rate, as described in lease, buy or PPA.

The direction each number is moving

A benchmark has a date, and the dates here are years apart. Berkeley Lab found that from the start of 2021, when inflation began to rise, nominal installed prices for residential systems rose by $0.2 to $0.3 per watt while real, inflation-adjusted prices fell by roughly $0.5 per watt; from 2022 to 2023 the real fall was about $0.1. EnergySage's blended median rose about 3 percent between the second half of 2025 and the first half of 2026 to $2.57, but the rise came from solar-plus-storage quotes (up 5 percent): solar-only pricing held flat at $2.62, and its highest average quote fell from $3.07 to $2.94 per watt. DOE's modeled benchmark sits between the two in time: $2.95 per watt without storage for Q1 2025 transactions, in 2024 dollars, and that edition was still the latest DOE listed when this page was checked on October 30, 2026. DOE's definition leaves out tax credits and treats tariffs as a temporary distortion of the modeled market price, and the Q1 2025 edition predates the import price floors that start on December 4, 2026. What those floors add to an 8 kW system is worked separately.

These signs do not all point the same way, so a single benchmark cannot be projected forward. What the combination suggests is that the middle of the market is moving slowly, in both directions, while the spread between installers stays wide: the lab's top-100 installers had median prices from $2.6 to $5.9 per watt in 2023. A quote that is $0.30 off a benchmark is inside the noise of the spread. One that is $1.50 off needs an explanation, and the explanation will be one of the items already listed: scope, financing, size, a battery or the installer.

What a gap tells you

A quote 40 percent above the marketplace median is not evidence of overcharging, and one below it is not evidence of a bargain. Roof complexity, a main panel upgrade, a tile roof or a long trench to a ground mount each add cost, and the adders for several are in the panel upgrade guide, the tile roof guide and the adders guide. What a gap can do is tell you which line item to ask about. If the normalized price is well above the band, ask for the scope; if it is below the marketplace median, ask what has been left out.