Mathify Accuracy and Corrections: What We Do and How to Report an Error

Short answer: Mathify calculators are built on small tested functions, they reject invalid input, and many show or return the formula behind the result. That lowers the chance of errors but does not remove it. If you find a wrong result, send the calculator name, the exact inputs, the answer you got and the answer you expected through the contact page. We do not promise a response time.

What backs a Mathify result?

Four practices sit behind a result. None is a guarantee on its own.

• Separated logic. The formula is a plain function, kept apart from the page. The build and testing article explains this in detail.
• Input checks. Functions refuse values that make no sense, such as a negative price or a compounding frequency of zero, and say so.
• Known-answer tests. The repository has automated tests (run with vitest) that compare function outputs with values worked out independently. Some test files are named for fixes and regressions, such as ones covering automotive, freelance-rate and self-employment corrections. That naming points to a habit of adding a test when something was found wrong, so the same mistake is caught if it returns.
• Visible inputs. Results returned through the programmatic tools list the inputs the calculator actually used. Many also include the formula and short worked steps.

What is not promised?

Being clear about this matters more than any claim of perfection:

• There is no promise that a result is free of error.
• There is no service-level agreement and no committed response time for reports.
• A calculator is not a substitute for a lender's disclosure, a tax professional, a doctor or a lawyer.
• Rules that change, such as tax thresholds or benefit limits, may be out of date. Check the current figures from the official source before relying on any result that depends on them.

Why do two correct calculators give different answers?

Often neither is wrong. They use different conventions. Here are cases you can verify yourself.

Day counts. Mathify's simple interest formula divides months by 12 and days by 365. For 1,000 at 5 percent a year over 90 days, that gives 1,000 times 0.05 times 90 divided by 365, which is 12.33. A tool using a 360-day year gets 12.50. Both follow their stated rule. Compare using the Simple Interest Calculator.

Compounding frequency. 5,000 at 6 percent for 10 years grows to about 9,096.98 with monthly compounding and about 8,954.24 with annual compounding. A difference of roughly 142.75 appears without any error at all.

Percent change from a negative number. Mathify's percent change formula divides by the absolute value of the starting number. Going from minus 50 to minus 25 is a change of plus 50 percent, because the loss became smaller. A formula without the absolute value would show minus 50 percent. If you know which convention your source uses, you can decide which result matches.

Direction of percent change. Going from 80 to 100 is a 25 percent increase. Going back from 100 to 80 is a 20 percent decrease. These are not mirror images, and the Percentage Calculator reflects that.

How can I check a result myself?

1. Confirm the inputs. Read each field on screen. Check units, such as percent versus decimal, monthly versus annual, years versus months.
2. Estimate first. A rough mental estimate catches factor-of-ten slips. 15 percent of 850 should land near 130, and the calculator says 127.5.
3. Use a second method. Work the formula by hand or in a spreadsheet. For compound growth, multiply the balance by one plus the periodic rate for each period.
4. Read the stated assumptions. Look for the compounding rule, the day count or the rounding.
5. Compare with the source of truth. For a loan or tax figure, the lender's or authority's own calculation controls.

How do I report an error?

Use the contact page. The following details let the problem be reproduced quickly:

1. The calculator name and page address.
2. The exact inputs you entered, copied rather than summarised. A prefilled link helps, as the guide to sharing prefilled links explains.
3. The result you saw.
4. The result you expected and how you worked it out, or the source you compared against.
5. Your browser and device, if the page looked broken rather than wrong.

Reports with a worked comparison are easier to confirm than reports that say a result "looks off". If the difference turns out to be a convention rather than a mistake, the fix may be clearer labelling instead of a changed number.

What happens after a report?

We can describe the usual practice but not the timing. The aim is to correct the calculation function and add a test with the corrected value so the case stays covered, though we cannot promise that for every report. We do not commit to a deadline or to replying individually to every message.

How does this apply to AI assistants using Mathify?

When an assistant calls Mathify's tools, the result carries the inputs used, a canonical page address and a citation line. The explain tool returns documented formulas, examples and common mistakes, and it leaves out fields that are not documented rather than making them up. If an assistant gives you a number with no inputs and no link, ask it for both. See the AI assistant connection guide for how the tools work.

Common mistakes when judging accuracy

• Comparing against a rounded figure. Compare at the same number of decimals.
• Mixing conventions. Match the day count, compounding and rounding before concluding an error.
• Trusting one tool, including this one. Cross-check anything important.
• Reporting without inputs. A report without inputs usually cannot be reproduced.

Key takeaways

• Results rest on tested functions, input checks and visible inputs. This reduces errors but does not eliminate them.
• Most disagreements between calculators come from different conventions, not mistakes.
• Check inputs, estimate, use a second method and read the stated assumptions.
• Report errors with the exact inputs, your result and your expected result.

FAQ

Is every Mathify calculator tested?

The repository contains a large set of automated tests across many areas, but we do not claim that every calculator and every input combination is covered. Treat test coverage as a risk reducer.

Will you tell me when a calculator is corrected?

We have not promised individual notifications. If your report is confirmed, re-run your inputs on the calculator later and compare.

Can I rely on a result for a tax return or loan decision?

Use it as an estimate. Rely on the official form, the lender's disclosure or a qualified professional for the final figure.

Why does the explain tool sometimes return less detail?

It returns only what is documented for that calculator. Missing sections mean no documented content, not hidden content.

Does the AI calculator have the same guarantees?

No. The AI calculator is a separate feature that interprets natural-language questions with an AI service, so its answers need the same cross-checking as any AI output. Compare them with a standard calculator page.

This article is for general information. It is not financial, tax or professional advice.