Payjoey.

8 corrections

What we have got wrong

A wrong figure matters to whoever relied on it, so nothing here is quietly edited. Each entry says what the site said before, what it says now, and how the error surfaced — that last part being the one worth reading.

This log was not empty on the day it was published, which is deliberate. Every entry below is a real defect found by the checks described on the methodology page, and several of them were found by a check that now runs on every build so the same mistake cannot ship twice.

Moving a dollar out of super was said to avoid Division 293 when it usually does not

· /division-293-calculator/

It said
The FAQ said that taking a dollar out of super and putting it back in your pay avoids 15c of Division 293 and 15c of fund tax, leaving you about 17c worse for every dollar moved.
It says now
The test adds your contributions to your income, so moving a dollar from one to the other leaves the total unchanged. Unless every one of your contributions is already caught, the Division 293 bill does not change at all, and the move costs about 32c a dollar at the top rate, not 17c. Only when all of them are caught does the dollar also escape Division 293. Both cases are now worked out from the same engine as the calculator, and a test pins each.
How it was caught
A sentence-by-sentence check of the day's new pages against the banked sources and the engine. The page's own default — $17,800 of $28,800 caught — was a case the sentence got wrong.

A quotation was given the wrong source

· /tax-tables/

It said
The FAQ quoted the ATO as saying the Schedule 1 formulas' results "may differ slightly from the sums of the amounts shown in the PAYG tax tables" because of rounding, and called the formulas "the method payroll software is written against". The same page said withholding and the year's bill are "close for steady income above $66,667".
It says now
The quoted words are the ATO's, but they are in Schedule 5, about combining PAYG and study loan amounts — not about the formulas this page uses. For those, the ATO's "Using a formula" page says results "may vary slightly" from the method in the tax tables' footnotes, only above $3,400 a week or $6,800 a fortnight, and that either method is accepted; the FAQ now says that. "The method payroll software is written against" is in no ATO source we have read, and is gone. And withholding runs ahead of the bill at every income where anything is withheld — above $66,667 by roughly the $1,000 standard deduction at your marginal rate, which the schedule does not include.
How it was caught
The same check of the day's pages: the quotation was not in the banked research, so the ATO page was re-read. The $66,667 sentence had been true until the standard deduction began in 2026-27; the test suite already knew the gap above it was $320 to $470, and the prose did not. This entry first said the sentence was not the ATO's at all, because only the Schedule 1 page had been searched; reading Schedule 5 for the bonus tax calculator on 9 October found it there, and the entry was corrected.

Two rules stated as if they had no exceptions

· /second-job-tax-calculator/ and the /tax-brackets/ year pages

It said
The second job page said that if you claim the tax-free threshold from no job, "More is withheld than you owe". The 2025-26 and 2026-27 pages said study loan repayments are charged "only on" repayment income above the first threshold.
It says now
Not claiming anywhere usually over-withholds, but two well-paid jobs can still fall short, because neither employer's scale reaches the brackets the two jobs reach together — two equal jobs from about $85,500 each. And from 2025-26 the study loan repayment becomes 10% of the WHOLE repayment income at the top ($186,051 in 2026-27), not a rate on the part above a threshold.
How it was caught
The same check of the day's pages, which asked of every "always", "only" and "never" whether the source or the engine said it.

Division 293 was tested on salary, not on taxable income

· /salary-packaging-calculator/ and / (the pay calculator)

It said
Both pages added super contributions to your SALARY to decide whether Division 293 applied. The law adds them to taxable income, which from 2026-27 is $1,000 lower because of the standard deduction. On a $251,500 salary sacrificing $2,000, the salary packaging page said the sacrifice caused $225 of extra Division 293; and the pay calculator warned that Division 293 was likely on salaries between about $223,200 and $224,100, where none is owed. The error was at most $150 of tax and only near the threshold, but it was wrong.
It says now
Both pages now use taxable income, through the same function as the new Division 293 calculator, so the three cannot disagree. The $251,500 example is $75, and the warning appears only once taxable income plus contributions passes $250,000.
How it was caught
Building the Division 293 calculator from section 293-20 of the Act, which names the income it tests. The standard deduction correction of 16 September had fixed the tax on every page but missed this second use of income, and all 404 unit tests passed on it: none of them sat within $1,000 of the threshold.

Every take-home figure was missing the $1,000 standard deduction

· / (the pay calculator), and the novated lease and salary packaging pages

It said
The site worked out your tax on every dollar you earned. On a $100,000 salary it showed income tax of $20,520 and a Medicare levy of $2,000, leaving $77,480. That understated take-home by about $320 a year at a 30c marginal rate, and by about $470 at the top rate. It was wrong from the day the site launched.
It says now
Tax is worked out on $99,000 — $20,220 of income tax and $1,980 of Medicare, leaving $77,800. From the 2026-27 income year the ATO applies a standard deduction of up to $1,000 for work-related expenses automatically, without you claiming it or keeping a receipt. It reduces the income you are taxed on rather than the tax itself, so it is worth your marginal rate and not its face value, and it is capped at what you earned. It also widened the gap on 'Why your payslip takes more than this': your employer withholds as though the deduction did not exist, so at $100,000 about $368 now comes back at assessment where the page used to call that figure rounding.
How it was caught
An audit of the live competitors, of all things. One of them offered a '$1,000 standard work deduction' toggle, which was recorded as something to verify rather than copy — and reading the ATO's own page found it was real, published on 8 July 2026 and in force for the year this site computes. All 337 unit tests and 164 browser tests passed throughout, because every one of them had been written against the same wrong assumption. A test suite cannot tell you about a rule you have never heard of.

The study loan comparison printed a shortfall as a refund

· / (the pay calculator)

It said
Ticking the study loan box compared two figures that were measuring different things — one side included the compulsory repayment and the other did not. At $100,000 the page said 'the difference of $4,523 is money you get back at assessment'. It was a $4,523 shortfall the reader would owe. At $180,000 it claimed $17,522 back.
It says now
The comparison runs on income tax plus Medicare on both sides, so the two figures are comparable. Schedule 1 of the withholding tables does not withhold for study loans at all — that is Schedule 8, a separate instrument.
How it was caught
A review of the previous day's work, reading the two sides of the comparison against each other. All 293 tests passed throughout, because nothing compared the two states.

A correct figure cited to a page that did not contain it

· Low income tax offset, used across the site

It said
The low income tax offset parameters were sourced to the ATO's tax rates page, which carries no LITO at all. The numbers were right; the citation did not support them.
It says now
Every LITO parameter is cited to the ATO's own LITO page and was re-read there in full. Sources now carry a field recording which claims each one actually covers, so a citation cannot drift onto a page that does not make the claim.
How it was caught
Working through the verify list and reading the cited page rather than trusting the reference.

The ETP cap was applied in full to someone who had already used part of it

· /redundancy-calculator/

It said
The calculator gave the excluded and non-excluded components the full ETP cap each. That is correct for one termination paid once, and too generous for anyone who has already received an employment termination payment in the same income year.
It says now
The page discloses the limit beside the result. Modelling it properly needs earlier payments the page does not ask for, so it is stated rather than estimated.
How it was caught
An engine review reading the calculator against its own data file, which recorded the reduction the engine was not applying.

Found something wrong?

Tell us and it gets fixed and logged. Every calculator names the source it was checked against and the date it was checked, which should make an error easy to point at — start from contact. What we already know we do not model is listed on the methodology page rather than left for you to discover.