Solar / Guides

What old solar calculators get wrong

On 6 September 2026 we read 14 public solar calculators. Six applied a federal tax credit to a present-day purchase that a buyer can no longer claim; five said so correctly; three said nothing about it at all. Only two stated what they assume your exported electricity is worth — the input that separates one state from another by years.

Before any of that: we published the same class of error ourselves, on Illinois, and it cost a reader $8,344 of imagined savings. That is where this page starts, because a site that documents its own mistakes has standing to describe anyone else's, and a site that does not is running an advertisement.

Start with ours

On a $200-a-month bill, this site published Illinois at 12 yr payback and $39,854 over 25 years. The correct figures are 8.5 yr and $31,510. Two separate mistakes, made at the same time:

We credited Illinois exports at the full retail rate

The Illinois Power Agency had already closed retail-rate net metering to new customers after 31 December 2024 under Public Act 102-0662; new customers receive only supply credits for what they export. We had 100% on file where the answer was 49%. On its own, that error alone overstated the 25-year return by $22,075.Illinois Power Agency, read 2026-09-05. Correction: 2026-09-05.

We recorded the Illinois Shines REC contract and then ignored it

It was on file, sourced, and displayed on the page as "not modelled". It is worth $13,734 to this household, half of it paid at switch-on. Leaving it out understated the return by $13,731 and pushed payback out to 16.9 yr.Illinois Shines, read 2026-09-05. Correction: 2026-09-05.

How we found it, and why it survived so long. The two errors pointed in opposite directions. Crediting exports at full retail made Illinois look better; ignoring the REC contract made it look worse. The published payback, 12 yr, sat in a plausible middle and never looked absurd enough to question. It surfaced only when we stopped spot-checking outputs and went back to the primary sources for every state's export rule — at which point the first error appeared, and correcting it made Illinois look so poor that the missing incentive became obvious in turn.

The lesson we took is the one this page is built on: errors that partly cancel are the dangerous kind, because the result stays believable. You cannot find them by looking at answers. You find them by checking inputs against sources — which is something a reader can only do if the tool tells them what its inputs are and where they came from.

What 14 public calculators state

So we asked how many tools do tell you. On 6 September 2026 we ran 3 searches, took every result offering a calculator or a state payback estimate in the order returned, and read each one's public page against four questions. 14 could be read; 6 refused the request or timed out and are excluded from every count rather than assumed either way.

6 of 14 is a real number and a small one, and it is the number we found rather than the number that would make a better headline. The more telling column is the last: 1 of 14 stated all four things a reader would need to check the answer. Most of these tools may compute perfectly well. On the evidence of the page a reader actually sees, there is no way to know.

How to re-run this survey

Everything needed to repeat this is published. The searches were:solar panel savings calculator estimate payback, how much can I save with solar calculator 2026, solar cost calculator by state payback period — run on 6 September 2026 through a search tool that returns United States results, with no narrower location set.

Two limits worth stating plainly. The pages were read, not driven: a tool may handle a case correctly without saying so on the page, and would be scored 'not stated' here. That is a real limit of this method and it cuts against the tools, so read the counts as 'what a reader can find out before trusting it', not as 'what the arithmetic does'. And a sample is a snapshot: one engine, one day, one country; pages change without notice, and each record carries the date it was read. The record — every URL, every quote, every date — is a data file in this site's repository, so the survey can be audited and re-run. We intend to re-run it in 2027-03 and publish whatever it says then.

Four error classes, with the size of each

Each of these is a mechanism, not a company, and each is quantified on this site's own engine so you can see what it is worth. We have committed two of the four ourselves.

1. Applying an expired credit to a present-day purchase

25D unavailable for property placed in service after 2025-12-31 (One Big Beautiful Bill Act). A 2026 cash or loan buyer receives 0%. A tool that still applies 30% to that buyer is not slightly optimistic; on the reference house it overstates the credit by $6,380, overstates the 25-year return by the same, and reports payback of 8.1 yr where the answer is 12 yr — 3.9 years out.IRC 25D (expired) — IRS, read 2026-09-05. What ended and what survives.

The subtle version is worse than the obvious one: the credit did not vanish from the industry, it moved. A lease or PPA provider still claims 30% and prices it into the offer. A tool that drops the credit entirely, for every way of paying, has replaced one error with another.

2. Assuming full retail net metering where it has closed

The largest error by value, and the one we made. Crediting exports at the full retail rate in a state that no longer does:

Each export rule is read from the regulation, tariff sheet or agency publication that governs it, linked with its date on the state page. These rules change: Illinois closed retail netting to new customers after 31 December 2024, and California moved in April 2023. A tool built before those dates is not badly made — it is simply out of date, which is the same thing to the reader relying on it.

3. One price basis presented as "the price"

Installed cost has two defensible answers. Competitive marketplace quotes and the prices buyers actually reported paying differ by up to 1.8× in the states we publish, and payback moves with them:

A calculator showing one of these without saying which is not exactly wrong. It is unfalsifiable, which is worse: a reader quoted the other number has no way to tell whether the tool was mistaken or simply meant something else. Both bases, by state.

4. Ignoring an incentive that is on file

Our second Illinois error, and the mirror image of the first. A performance incentive pays for electricity produced rather than up front, so it is easy to record and forget:

Understating a return is the same failure as overstating one, and it is the one nobody complains about. This site now refuses to publish a state carrying an incentive the engine cannot pay out — the check runs in both directions.

How to check any solar calculator in sixty seconds

Six questions. Put them to any tool, this one included. None requires you to know anything about solar; they ask only whether the tool has told you enough to catch it being wrong.

1. Does it name the year its policy assumptions belong to?
   A federal credit is a fact about a year, not about solar. A tool that shows a percentage without a year cannot be checked, and will be wrong the moment the law moves.
   This site: Yes. Federal policy is a versioned data file keyed by year, and nothing in the calculation code knows the number 30.
2. Does it say where its electricity rate comes from, and for which month?
   Rates move several percent a year and the rate is the largest single input to any saving. "The average US rate" without a date is unfalsifiable.
   This site: Yes: U.S. EIA API v2, electricity/retail-sales, residential sector, period 2026-07, shown on every page that uses it.
3. Does it treat buying and leasing differently?
   In 2026 they diverge completely: a buyer gets nothing federally, while the owner of a leased system still claims a credit and prices it in. A single "savings" number that covers both describes neither.
   This site: Yes, three ways to pay are computed separately, and a state credit reaches a lease customer only where the source says it does.
4. Does it tell you what it assumes your exports are worth?
   It is the input that separates one state from another, and the one most often left silent. Full retail and a quarter of retail are both common, and the difference is years.
   This site: Yes, an export ratio per state with the tariff, statute or regulator it was read from and the date.
5. Can you see the formula?
   Not the code necessarily — the arithmetic. If you cannot see how the number was reached, you cannot find the error, and neither can anyone else.
   This site: Yes, every formula is on the methodology page, and the same engine runs the pre-rendered pages and the calculator, so the two cannot disagree.
6. Does it publish corrections?
   The strongest signal available, because it costs something. A tool that has never published a correction has either never been wrong or never looked.
   This site: Yes — 3 entries so far, including the two on this page.

A tool that answers all six can be audited by a stranger. That is a different and more useful property than being accurate, because accuracy is something you have to take on trust and auditability is not.

Where this site is still weak

We would rather you found these here than found them yourself, and they are the places to point a challenge first:

Texas exports are a modelling choice, not a published rule

Texas has no statewide net metering mandate, so no source states a single export number. We use 70% of retail as a stand-in for a typical energy-only buyback. Every Texas figure on the site carries a warning saying so, and it is the one export ratio here that a reader should replace with their own provider's plan.

The lease model rests on two literature defaults

The price a lease charges (70% of retail, rising 2.9% a year) and the share of output a house consumes itself (40% without a battery) are typical published values, not measurements. Under a poor export rule they decide whether a lease saves or costs money, so any negative lease result here is directional until checked against real quotes.

Eight states' year-end surplus rules are unread

In-year export credits are sourced for all eleven states we publish. What happens to a surplus left at the end of the year is read from a primary source for only three, so the site does not publish figures for a system deliberately larger than the household uses.

One installed price at every system size

Small systems really cost more per watt, because permitting and a day of labour do not shrink with the array. Our figures for unusually small systems are therefore optimistic.

Every one of those is on the record elsewhere on the site — in the flags on the state pages, in what is not modelled, and in the corrections log. If you find a figure here that is wrong, that log is where it will end up, and the address is in the footer of every page.

Questions

Do most solar calculators still apply the expired federal tax credit?

In a survey of 14 public tools read on 2026-09-06, 6 applied a federal credit to a present-day purchase without qualification, 5 correctly stated that a 2026 buyer gets none, and 3 said nothing about a federal credit at all. Six more tools in the sample could not be read and are excluded from those counts.

How much difference does that one error make?

On the reference house this site uses — a 9.58 kW system in Austin, Texas, on a $200-a-month bill — applying the expired 30% credit overstates the credit by $6,380, overstates the 25-year return by the same amount, and shortens payback from 12 yr to 8.1 yr — an error of 3.9 years.

What is the biggest error, in dollars?

Assuming full-retail net metering in a state that has ended it. California exports are credited at 20% of retail; crediting them at 100% overstates the 25-year return by $34,634 on a $200 bill. We made this exact error on Illinois and published it.

Which calculator should I use?

That is the wrong question, and this page deliberately does not answer it. Use the checklist: ask any tool, including this one, whether it names its policy year, cites its rate with a date, separates buying from leasing, states what your exports are worth, shows its formulas, and publishes corrections. A tool that answers all six can be checked; one that answers none cannot, however good its arithmetic may be.

Is this site accurate?

We do not claim that, because we cannot prove it. What we claim is that we are checkable: every input names its source and the date it was read, every formula is published, the calculator runs the same engine as the pages, and every figure we have got wrong is written up with what changed. 3 corrections are on the record, two of them described on this page. If you find another, the address is on every page.

Why does this page not name the calculators it surveyed?

Because the error classes are the useful part and a league table is not. The tools were read once, on one day, on one search engine; a page changes without notice and a company can fix a number the day after we read it. The survey record in the site repository carries every URL and quote with the date observed, so the survey can be audited and re-run — but the page reports counts, not names.

This is general information computed from published data and statute, each cited with the date it was read. It is a screening estimate, not a quote, and not financial or tax advice: confirm the figures against real quotes and with a professional before you rely on them.

Published 6 September 2026. Data last verified 14 October 2026. About the author.