Methodology

This page is the engine, written out. Every number on the site — the pre-rendered state pages and the interactive calculator alike — comes from one module, src/lib/engine/, that implements exactly what follows. If a formula changes, this page changes in the same commit.

Principles

1. Federal policy is data, not a constant. The credit percentage lives in a versioned file keyed by year. Nothing in the calculation code knows the number 30.
2. Every input carries a source and a verification date.
3. A value we have not verified is null. The engine falls back to a national figure and raises a flag that the page must show. We never type in a plausible number.
4. One engine feeds everything. The state pages are rendered at build time by the same function the calculator runs in your browser. They cannot disagree.

The federal credit in 2026

Two different statutes, two different recipients, and in 2026 they diverge completely.

25D unavailable for property placed in service after 2025-12-31 (One Big Beautiful Bill Act).

Why the owner gets 30% rather than the 6% base rate. One-megawatt exception: a facility with maximum net output under 1 MW AC is entitled to the increased credit rate without meeting prevailing-wage and apprenticeship requirements (26 CFR 1.45Y-3). Residential arrays are 5-15 kW, far under the threshold.26 CFR 1.45Y-3.

Risks a lease customer should ask about:

• Systems whose construction begins after 4 July 2026 fall under different termination rules.
• Systems placed in service after 31 December 2027 face restrictions.
• Equipment sourcing must satisfy the foreign-entity-of-concern (FEOC) rules.

1 · Your consumption

We start from the one number you know: your monthly bill. Part of it is a fixed charge solar cannot remove.

energyBill  = max(0, monthlyBill − fixedMonthlyFee) × 12
annualKwh   = energyBill ÷ rate

rate is your state's average residential price in $/kWh from the EIA. Overridable in the calculator.

2 · Sizing the system

The system is sized to produce a target share of your annual use, given how much a kilowatt of panels yields where you live.

sizeKw           = annualKwh × offsetTarget ÷ specificYield
panels           = ⌈ sizeKw × 1000 ÷ panelWatts ⌉
roofSqft         = panels × sqftPerPanel
year1Production  = sizeKw × specificYield

specificYield is kWh per kW DC per year from NREL PVWatts v8 for your city, or a statewide default, at 20° tilt, due south, 14% system losses, fixed open rack.

3 · What it costs

grossSolar      = sizeKw × 1000 × usdPerWatt
grossTotal      = grossSolar + (battery ? batteryCost : 0)

federalCredit   = grossTotal × policy[year].purchase.creditPct ÷ 100
stateIncentives = Σ over each verified incentive paid to the buyer:
                    tax_credit_pct   → min(grossTotal × pct ÷ 100, cap)
                    tax_credit_flat  → value
                    rebate_flat      → value
                    rebate_per_watt  → min(sizeKw × 1000 × value, cap)

netCost         = max(0, grossTotal − federalCredit − stateIncentives)

A performance incentive is not a discount on the price: it is paid for electricity
produced, on a contract term, and enters the cash flow in the years it arrives.
                    as_generated → productionᵧ × value, for term years
                    part_upfront → the term's expected output × value, of which
                                   upfrontShare arrives at switch-on and the rest
                                   is spread over the following years
An incentive on file that the engine cannot pay out raises an error and the page
does not publish: understating a return is the same failure as overstating one.

usdPerWatt is the installed cost before incentives on the modelled basis. The engine runs the identical formulas a second time on the reference basis where one exists, and every page shows both (see Two price bases). A state without its own cost record is not published.

4 · Twenty-five years of bills

For each year y from 0 to 24, we compute what you would have paid without solar and what you pay with it. The difference is your gross saving.

rateY        = rate × (1 + escalation)^y
productionY  = year1Production × (1 − degradation)^y

selfUsed     = min(productionY × selfConsumption, annualKwh)
exported     = max(0, productionY − selfUsed)
imported     = max(0, annualKwh − selfUsed)

solarCharges  = productionY × productionChargePerKwh + sizeKw × capacityChargePerKwMonth × 12

billNoSolar   = annualKwh × rateY + fixedFee × 12
billWithSolar = imported × rateY + fixedFee × 12 − exported × rateY × exportRatio + solarCharges

grossSavings  = billNoSolar − billWithSolar
netSavings    = grossSavings − om − (y + 1 = inverterYear ? inverterCost : 0)

exportRatio is the single number that makes California different from Massachusetts. Under full-retail net metering it is 1.0 and self-consumption does not matter. Under California's net billing tariff it is roughly 0.20, and a battery — which raises self-consumption from 40% to 75% — becomes decisive.

5 · Cash, loan, lease

Cash

cashFlows   = [ −netCost, netSavings₁, netSavings₂, … netSavings₂₅ ]
payback     = breakEvenYear(cashFlows)
netSavings25 = Σ netSavings − netCost
roi         = netSavings25 ÷ netCost
irr         = irr(cashFlows)

Loan

payment      = monthlyPayment(netCost − downPayment, apr, termYears)
loanFlows    = [ −downPayment, netSavingsᵧ − 12 × payment (while y ≤ term), then netSavingsᵧ ]
totalInterest = 12 × payment × termYears − (netCost − downPayment)

The federal credit, when there is one, is applied to the amount financed.

The curves: cumulative position

position[y] = Σ flows[0..y]      for each way of paying, y = 0 … 25
cash:   flows = [ −netCost, netSavings₁ … netSavings₂₅ ]
loan:   flows = [ −downPayment, netSavingsᵧ − 12 × payment … ]
lease:  flows = [ 0, grossSavingsᵧ − leasePaymentᵧ + stateCreditᵧ … ]

Every chart on the site is drawn from these series, at build time or in your browser, so a picture can never disagree with the table beside it. Year 0 is the day the system is switched on; the cash curve crosses zero at the payback year printed in the table.

Zero-upfront scenarios have no payback

Nothing was paid up front, so "payback" is meaningless for a zero-down loan or a lease. The engine reports a cash-flow verdict instead, and the pages print it in words:

threshold  = annualBill × immediateMinShareOfBill
immediate  if netSavings₁ ≥ threshold
neutral    if 0 ≤ netSavings₁ < threshold      → "cash-flow neutral from year 1 (+$x/yr)"
later      if netSavings₁ < 0 and Σ netSavings₁..ᵧ ≥ 0 for some y   → "positive from year y"
never      otherwise

A one-dollar margin must never read as "immediate". The 25-year net is the column to compare; it is printed first and in bold.

Lease / PPA

ppaRateY     = rate × ppaRateVsRetail × (1 + escalator)^y
leasePayment = productionY × ppaRateY
netSavingsY  = grossSavingsY − leasePayment
upfront      = 0
federalCreditToYou = 0

The lease is priced below your retail rate because the owner keeps the 30% credit. You see it only as a discount. The engine never credits it to you.

Finance functions

monthlyPayment(P, apr, years)
  r = apr ÷ 12,  n = years × 12
  = P × r ÷ (1 − (1 + r)^−n)            (P ÷ n when apr = 0)

breakEvenYear(flows)
  flows[0] is the outlay at t = 0; flows[i] arrives over year i (t = i−1 → i)
  find the first i where Σ flows[0..i] ≥ 0, then interpolate inside that year:
  payback = (i − 1) + (−Σ flows[0..i−1]) ÷ flows[i];  null if never
  (checked by a test: the cash curve crosses zero at the printed payback)

irr(flows)
  the r where Σ flows[i] ÷ (1 + r)^i = 0,
  found by bisection on [−90%, +150%], 200 iterations; null if no sign change

Two price bases

Installed cost is the one input with two defensible answers, and they differ by roughly 60%. A single number would be false either way — and a reader who is later quoted the other price would rightly stop trusting the site — so the engine computes the economics on both bases, through the same code path, and every state page and the calculator show the range.

The gap is methodology, not staleness. Which basis matches the quote in your hand decides your payback — which is exactly why both are shown.

Assumptions

Read straight from src/data/finance/defaults.json, last verified 2026-09-04. Each one is exposed as an adjustable control wherever it matters.

Data and sources

What is not modelled

Being clear about this is part of the method. This is a screening estimate, not a site survey.

• Two literature defaults, not measurements. Self-consumption (40% without a battery, 75% with) and the lease price (70% of retail) are typical published values, not figures we measured. Under California's NEM 3.0 they decide whether a lease saves or costs money, so treat any lease result there as directional until it has been checked against real quotes.
• A ceiling on export credits. The engine credits every exported kilowatt-hour at the state's export ratio, with no annual limit. That is right for a system sized to the household's own use — the default — where annual production is about equal to annual consumption and the year-end surplus is near zero. It is wrong for a system generating materially more than the home uses: Pennsylvania settles a year-end surplus at the price to compare, New Jersey at the wholesale value and Maryland at the commodity portion, all below the in-year credit, and New Jersey does not permit a system larger than the customer's annual needs at all. Each of those provisions is recorded in netmetering/state.json with its source; none is modelled, so the site does not publish figures for a deliberately oversized system.
• Price per watt by system size. One installed price is applied at every size. Real small systems cost more per watt, because permitting, design and a day of labour do not shrink with the array, so figures for unusually small systems are optimistic.
• Time-of-use rates. A single average $/kWh is used. Under TOU tariffs the value of exported and self-used energy differs by hour.
• Merchant SREC markets. Where certificates trade at a market price rather than a fixed contract rate — Pennsylvania, for instance — there is no single value to record, so nothing is paid out. Fixed-rate contracts, such as those in Illinois and New Jersey, are modelled, on their own term and payment schedule.
• Roof specifics. Orientation, shading and pitch beyond the PVWatts reference assumptions.
• Home value, insurance, and property tax effects.
• Loan fees and dealer fees, which in practice are often folded into the quoted price.
• Discounting. The 25-year net is a plain sum, not a present value. The IRR figure is there for readers who want the time value of money.
• Utility rate structure changes, including the grid-access and demand charges some utilities have introduced for solar customers.