Every number that was wrong here is still on this page – with the date, the old figure, and how it came to light. A correction you can no longer see is not one.
What was wrong: The article reversed negative equity, suggested putting APR into a nominal-interest model and described Total of Payments as the full cost of the car. Its year-four claim was not established.
What is right now: Negative equity means owing more than the vehicle is worth. APR and nominal interest are distinguished; scheduled payments, separate cash and ownership costs are separated. A synthetic example replaces a comparison assembled from different market averages.
How it came to light: A new local runtime review reproduced the three statements. Official CFPB and FTC sources, real HTTP regression cases and a separate offer-comparison task now cover the distinctions.
In the site itself
What was wrong: The debt calculator said “The balance grows every month” even at 0% interest and zero payment, or when payment exactly covered interest. Reaching the 1,200-month calculation limit was described as “never get out” even while principal was being repaid.
What is right now: Growing balances, unchanged balances and a remaining balance at the calculation horizon are distinct results. Beyond the horizon, total payoff time and total interest remain unknown. The extra-payment comparison has its own state. Monthly calculations are explicitly a model, not a lender statement; actual daily card billing is explained with a CFPB source.
How it came to light: Public calculations used a $1,000 balance, 0% interest with zero or $0.01 payments, and 24% APR with a $20 payment. Live HTTP regression checks then covered the result, FAQ and additional-payment comparison.
What was wrong: The embedded calculator said: “Anything above this does not belong in a savings account any more.” It also claimed that an accessible account pays little interest.
What is right now: The entered target is a planning choice. Holding additional accessible savings can fit upcoming expenses and personal circumstances. Ignoring interest is a calculation assumption, not a claim about an account’s actual interest or access time.
How it came to light: During a full browser review and inspection of calculator result states while building the rent affordability tool.
In the site itself
What was wrong: “It is on the rating plate, or on a plug-in meter.” The power calculator listed the two readings without distinguishing them.
What is right now: The calculator now distinguishes rated or maximum power from measured average power and explains how to avoid counting idle consumption twice.
How it came to light: During a live review and preparation of the measurement guide, checked against the explanation of varying power consumption.
What was wrong: The article priced a kilowatt-hour at 18 cents and justified it with: “Eighteen is the round number the calculator below works in.” The title read “Unplugging everything saves you $1.58 a year.”
What is right now: Since September 28, 2026 the calculator reads fractions of a cent, so the sentence was no longer true, and 18 cents was 2.4 percent below the figure its own source gives for May 2026. The article now uses 18.44 cents, the U.S. residential average in May 2026 (EIA, Electric Power Monthly, table 5.3), like every other price on the page. One watt of standby comes to $1.62 a year, and the title says so; the 2025 average of 17.30 cents is calculated alongside instead of typed in.
How it came to light: During a review of the sentences that describe one of our own calculators. This one described a limitation that had been removed the day before, so it went stale without a single number in it changing.
In the site itself
What was wrong: The source line under the standby field of the Appliance Power Calculator, also shown in the calculator built into the article on unplugging everything, read: “EU ecodesign rules cap standby at 0.5 watts for most new devices; smart speakers, set-top boxes and games consoles are exempt.”
What is right now: Commission Regulation (EU) 2023/826 exempts none of them: Annex II lists audio speakers, set-top boxes and games consoles expressly as covered. The 0.50-watt cap applies to standby with only a reactivation function (0.80 watts with a display); networked standby has a higher cap, 2.00 watts, and 8.00 watts for routers, modems and similar network equipment (Annex III). The line now says so and cites the regulation itself on EUR-Lex.
How it came to light: On September 28, 2026, while attaching an address and a retrieval date to every calculator default. Reading the regulation in order to cite it showed that the exemption is not in it. The line was corrected the same day, but this entry was not written until September 29, 2026.
What was wrong: “The usual answer is ‘three months of salary’.”
What is right now: That is the German rule of thumb, and it had traveled across with the translation. The American sources – SEC, FDIC, Federal Reserve – measure in EXPENSES without exception, not in salary. The article now gives the sourced range and says plainly that narrowing it to fixed costs is our own judgment.
How it came to light: While looking for an American source for the rule of thumb. There was none for that one – and three for a different one.
What was wrong: “Go twice a week and you pay under $5 a visit.”
What is right now: $588 a year divided by 104 visits is $5.65. The figure was 13 percent too low – and too low in favor of the gym the article argues against.
How it came to light: While rechecking every number in this article by hand against the calculator it uses.
What was wrong: “It is close to a full paycheque.” And the FAQ took 260 working days minus 10 public holidays minus a fortnight of leave and arrived at 250.
What is right now: The German version says “roughly half a monthly net salary”; the claim had grown in translation. And the arithmetic did not work: 260 minus 10 minus 14 is 240. The leave clause is gone – there is no statutory entitlement in this market anyway. The paragraph about the 7 percent contradicted itself as well and has been rewritten.
How it came to light: While rechecking the FAQ arithmetic against the days the calculator actually uses.
In the site itself
What was wrong: Nine English articles shipped empty headings: <h2 id="kapitel-1"></h2>, and entries without a name in the table of contents.
What is right now: The sections carried their text in the wrong field. Twenty headings are repaired, and the loader no longer lets a subheading through without a title.
How it came to light: While rendering an article against the real server. No test could see it: the pages answered with 200, and the loader checked the KIND of a section, never its field.