Changelog
Data refreshes and model changes, with the effect each had on published numbers. Only changes that actually happened are listed.
2026.08.6-r1
Two published heat-pump claims withdrawn after audit; one replaced with an exact result
Before publishing a citable research page, every headline claim was re-derived from the engine and two failed. First, an internal list of 'the 16 states with the most aggressive heat-pump incentives' had no published ranking behind it and no definition of 'aggressive', so the claim built on it was dropped rather than defended. It is replaced by a dated, named, verifiable fact: nine states and the District of Columbia signed the February 2024 NESCAUM multi-state MOU on zero-emission residential buildings, and 9 of those 10 currently show higher heat-pump operating cost than gas. Second, a proposed 'electricity-to-gas price ratio below 3.0 favours heat pumps' threshold was tested and misclassifies three jurisdictions (Virginia, Kentucky, Nevada); winners span ratios 1.84 to 3.25 and losers 3.09 to 6.26, so the ranges overlap and no fixed cut can separate them. It is replaced by the exact switching condition, price ratio < COP/AFUE, which is algebra rather than a fitted threshold and which the build now checks in all 50 jurisdictions with a heating season. A claimed contrast against 'industry 4-7 year paybacks' was also dropped: the traceable sources for that range describe geothermal, new construction, or heat pump plus solar, so the comparison was not like-for-like.
Headline counts restated precisely; no calculation changed
'Cheaper to run in 17 of 51' was imprecise: Hawaii has no meaningful heating season and is not a comparison the model makes, so the correct statement is 17 of the 50 jurisdictions with a heating season, out of 51 covered. '33 never pay back' was verified as exact rather than rhetorical — payback is null in precisely the 33 jurisdictions with no annual operating saving, and with no saving there is nothing to accumulate at any horizon. '2 of 51 recover the premium within 15 years' (Arkansas 12.4, Georgia 14.7) was confirmed, and the premium is now everywhere described as a $4,000 incremental cost over replacing a furnace and air conditioner, never as the price of a heat pump. No formula or input changed, so every calculator result, ranking and verdict is unchanged and the model version stays 2026.08.6.
2026.08.6
Target price added: what a purchase would have to cost to make sense
State pages now show the installed price at which a system would pay back inside 10 years, solved from the same engine that produces the payback figure. It is a direct inversion of the cash-flow model rather than an approximation: because system size is set by consumption and not by price, the annual savings stream is independent of what you pay, so the cumulative-savings curve can be read backwards exactly. The solved price is rounded down to the dollar so the stated payback is never later than promised, and every state is checked at four target periods by feeding the answer back through the calculator. Ten years is described on the site as a reader-facing benchmark, not a rule of finance, and it sits inside the existing 'Reasonable' verdict band so the commercial framing can never be looser than the editorial one.
Commercial layer built, switched off, and walled off from the maths
Payback Math will eventually earn money by referring readers to installers when the maths supports it. That machinery now exists and is disabled: no approved partner, no affiliate link, no tracking script, no cookie, no analytics provider, and no personal data collected. Nothing about it can reach a calculation — an automated check runs all four calculators across all 51 jurisdictions with the commercial layer active and inactive and fails the build if any financial output moves. The visible consequence is that the site declines to sell on a bad result: 33 of 51 heat-pump pages state that we would not recommend switching and carry no quote prompt, and home batteries carry no commercial prompt at all because they pay back within warranty in 0 of 51 states.
2026.08.5
Battery migrated to the shared export-compensation policy
The battery calculator carried its own copy of the old 100/50/25 percent export buckets, so after solar moved to the jurisdiction-level export policy the two disagreed about what an exported kilowatt-hour is worth in 22 states. California was the visible case: solar priced an export at 6 cents while battery still priced the same kilowatt-hour at 16.4. Both calculators now read one record per jurisdiction. They are pricing the same transaction — a residential rooftop export under the state's rules — once as income to a solar owner and once as the income a battery owner forgoes by storing it. Mismatched states went from 22 to 0. The duplicated bucket maps in both models and the redundant per-state exportCredit column were deleted, so there is exactly one place export compensation is defined.
Batteries look slightly better, not worse — the opportunity cost runs the other way
Correcting export compensation downward makes a battery MORE valuable, because storing a kilowatt-hour costs you less foregone export income. Median annual bill saving moved from $203 to $213 and the mean from $247 to $265; mean payback among states that ever pay back shortened from 73.2 to 72.0 years. Three more states now do better storing solar than arbitraging a time-of-use spread. Twenty-one ranks moved — Michigan rose from 12th to 5th as its export value fell, while Oklahoma dropped from 11th to 25th as its export value rose. Time-of-use spreads, battery cost, capacity, round-trip efficiency, cycle count and warranty were all left untouched so the effect isolates the policy correction.
Home batteries still pay back within warranty in 0 of 51 states
Re-checked rather than assumed. The closest is Hawaii at 10 years 9 months against a 10-year warranty, still about $1,020 short when the warranty expires, on the highest bill saving in the country at $1,383 a year. California follows at 15 years 7 months. Five states have no viable value stream at all. Every state's verdict remains 'does not pay back' on bill savings alone, which is the same conclusion as before and now rests on consistent arithmetic. Solar, EV and heat-pump findings were verified unchanged.
2026.08.4
Export compensation rebuilt: policy structure separated from what exports are worth
Solar export value was a single field with three buckets — full, partial, none — mapped to 100%, 50% and 25% of the retail rate. Two things were wrong. It conflated whether a state HAS net metering with what an exported kilowatt-hour actually EARNS, which are different facts that DSIRE publishes as two separate maps. And expressing avoided-cost compensation as a share of retail is a mathematical error rather than a rounding: avoided cost tracks wholesale generation value and does not rise with a state's retail rate. Arizona's published export tariffs are about 5.7 to 6.2 cents against a 15.6 cent retail rate, while the same order of magnitude against California's 32.8 cent retail is closer to 18%. The old model therefore overstated Arizona export value by about a third and California's by roughly 170%, and the overstatement grew with the retail rate — largest exactly where solar already looked best. Export compensation is now recorded per jurisdiction with an explicit unit, and non-retail compensation is modelled in cents per kilowatt-hour.
Seven jurisdictions verified; the other 44 openly flagged as unverified
California, Arizona, Texas, Idaho, Vermont, New York and Oregon were checked against a primary regulator or DSIRE's published policy notes and now carry a source and a verification date. California in particular is modelled on the Net Billing Tariff that applies to a new 2026 installation, not the legacy NEM terms that grandfathered customers keep. The remaining 44 jurisdictions carry the site's earlier assumption, translated into the new taxonomy and labelled unverified rather than promoted: DSIRE's per-state records sit behind a licensed API and its summary maps encode states as unlabelled shapes, so reading 51 classifications out of map colours would have been inferring data from a picture. Policy classification is now 7 sourced and 44 unverified; the numerical export value is modelled in 48 of 51 jurisdictions and sourced only where verified retail-rate net metering sets the price directly.
Solar looks materially worse in high-rate net-billing states
Twenty-two of 51 export values changed. National median payback moved from 12 yr 10 mo to 13 yr 6 mo, mean first-year saving from $1,517 to $1,483, and mean lifetime gain from $27,398 to $26,312. Seventeen states shifted by more than a year. California fell from 5th to 19th and from a Strong verdict to Reasonable — the largest single correction, and a direct consequence of pairing the country's highest retail rate with avoided-cost exports. Alabama, Indiana and Oklahoma improved, because the old 'none' bucket had been penalising them at 25% of retail when the label had only ever meant no statewide mandate, not no payment. Seven verdicts changed in total and 46 of 51 ranks moved. Nothing else was adjusted to soften this: consumption, cost per watt, yield, self-consumption, degradation, escalation and the federal credit are all untouched.
2026.08.3
Federal residential clean-energy credit verified against current law
The solar model applies no federal tax credit. That was previously an unverified assumption; it is now a checked legal position. IRC §25D was terminated by Public Law 119-21, enacted 4 July 2025, for expenditures made after 31 December 2025, confirmed against the enacted statute, IRS Fact Sheet FS-2025-05 and IRS guidance current as of July 2026. The rate is unchanged at 0%, so no payback period, ranking, verdict or headline figure moved — this is a provenance correction, not a recalculation. A sourced zero and an unexplained zero are not the same thing.
Corrected the federal credit cutoff from 'purchases' to completed installation
Three places on the site said the credit ended for systems 'purchased' after 31 December 2025. That was wrong in a way that mattered: §25D(e)(8)(A) treats an expenditure as made when the original installation is completed, so a system paid for in 2025 but commissioned in 2026 receives nothing. Anyone who put down a deposit in December 2025 would have drawn the wrong conclusion from the old wording. The federal-credit treatment now lives in one policy record keyed by model year, and every public statement about it — calculator notice, state page FAQ and cost table, research footnote, methodology and data sources — is generated from that record rather than restated. The record also carries a last-verified date, and the provenance audit reports how long a legally sensitive input has gone unreviewed.
2026.08.2
Residential electricity use replaced with EIA Table 5.A observations
The per-state monthly electricity use that sizes every solar system was a bucketed internal estimate — 51 states sharing just 17 distinct values — and was flagged on this site as having unresolved provenance. It now comes from EIA's Electric Sales, Revenue, and Average Price, Table 5.A: average monthly consumption per residential customer, 2024 (released 7 October 2025), covering all 50 states and the District of Columbia. This replaces modeled estimates with authoritative state-level observations; it is not a claim that any state average describes a particular home. Note the metric is per residential customer account, not per household, and the site's wording was changed to match. All 51 values changed, by 4.8% at the median and up to 14% (Wisconsin 750 to 645 kWh/month). Electricity prices remain 2026 year-to-date — each field uses the strongest source available for it, so data vintages differ by design.
Solar system sizes fell; payback barely moved
Because a system is sized to cover a year of use, lower consumption cuts both the up-front cost and the annual saving, and the ratio between them is what payback measures. National median modelled system size fell from 7.2 kW to 6.9 kW and mean up-front cost from about $21,900 to $21,100, but the national median payback moved only from 12 yr 9 mo to 12 yr 10 mo. No state moved by more than three months, no state changed financial verdict, and the top ten was unchanged. Eight states shifted rank, none by more than two places. The residual sensitivity comes entirely from installed cost per watt rising on smaller systems. Nothing was tuned to keep the old results: the only input that changed was the consumption dataset.
2026.08.1
Annual vehicle mileage replaced with FHWA data
Per-state annual miles now come from FHWA Highway Statistics 2023: state vehicle-miles travelled (Table VM-2) divided by licensed drivers (Table DL-22), rounded to the nearest 100. The national result, 13,662 miles per licensed driver, lands within 1.3% of the commonly cited figure. This replaced internal estimates and changed 49 of 51 states, by 11.2% on average and up to 30% (Missouri 14,500 to 18,800; DC fell 8,500 to 6,700). Median annual EV fuel saving moved from $1,081 to $1,157; the highest-saving state stayed Wyoming but rose from $1,574 to $2,023, and the lowest is now DC at $470 rather than Rhode Island at $592. EVs remain cheaper to fuel in all 51 states.
Provenance classification added for every material input
Every input and output is now classified in one place as sourced data, a modeled assumption, a derived value, or an unresolved source. Previously the site said 'estimates' without distinguishing a value we measured from a value we chose. Five material inputs are now openly flagged as having unresolved provenance rather than being presented as fact: household electricity use, net metering status, solar export credit regime, outage duration and the federal solar tax credit rate. No calculation changed as a result of this reclassification.
Calculators now separate the published default from your own value
Every adjustable input shows where its starting value came from — a named dataset, a Payback Math assumption, or an input whose source is unresolved. Change a control and the label switches to 'Your value' alongside the published default and a reset link, so a personalised result can no longer be mistaken for the figure we publish in the rankings. The ranking tables, state pages and structured data continue to use the published baseline regardless of what you enter. One unused constant, a national 13,500-mile fallback in the EV model, was deleted; every state now carries an FHWA figure, so nothing read it.
Installed cost per watt and time-of-use spread kept as assumptions
Both were reviewed for promotion to sourced data and deliberately left as modeled assumptions. Installed solar price varies more by installer, system size and financing than by state, and the authoritative dataset lags roughly two years; a statewide TOU spread is meaningless when many households are not on a time-of-use tariff at all. Publishing either as a state fact would be false precision.
2026.08
Solar production now uses modelled AC yield instead of a flat derate
Production was calculated as peak sun hours x 365 x a fixed 0.86 loss factor. Checked against PVGIS's own modelled AC output, that ran about 9% high across all 51 states, because a flat derate cannot capture temperature and angle-of-incidence losses separately. The model now uses per-state kWh per kW installed from PVGIS. Payback periods lengthened by roughly a year; Texas moved from 13 yr 7 mo to 14 yr 5 mo.
Solar resource data sourced from PVGIS
Peak sun hours and annual yield per kW for all 51 states retrieved from PVGIS v5.3 (PVGIS-ERA5, 2005-2023) at each state's largest metro. Previously these were internal estimates.
2026.08-data
Electricity, gas, gasoline and climate data replaced with primary sources
Residential electricity prices from EIA Electric Power Monthly Table 5.6.B (year-to-date through May 2026). Residential natural gas prices from EIA, latest published year per state. Gasoline from AAA state averages. Heating degree days from NOAA CPC population-weighted state normals. All 51 states, no gap-filling. These replaced internal estimates; the largest corrections were to gasoline, which had been understated by up to 34% in some states.
Headline findings re-verified against sourced data
Heat pumps cheaper to run than gas moved from 18 states to 17. EV cheaper than gasoline stayed at 51 of 51. Home battery payback within warranty stayed at 0 of 51.
What gets logged
Anything that changes a published number: a dataset refresh, a formula change, a corrected assumption. Where a change moves results, we state the direction and rough size rather than only that something changed.
Cosmetic and copy edits are not logged. The reasoning behind what we do and do not record is in our editorial policy.