How our tools work

The rules every tool follows

1. No score means anything it cannot explain. Where a tool produces a number, that number is a weighted sum of specific checks listed on this page. It is not a prediction, and it is not calibrated against hiring outcomes because no such calibration data exists.
2. Show the working. The tools display the intermediate values, not just the answer — the terms that matched, the hours that were counted, the arithmetic that produced the figure. A result you cannot check is a result you should not act on.
3. Every published figure is stamped and editable. Tax rates, mileage rates and contribution limits appear on the page with their year and a link to the primary source, and every one can be overridden with a current value.
4. Fail visibly. If an input is missing or nonsensical, the tool says so. It never quietly substitutes a default and presents the result as if you had supplied it.

Resume checker: every check it runs

The resume checker runs a fixed set of structural checks on pasted text. It is not an applicant tracking system simulator, it does not model any employer's software, and its score has no relationship to your chances. Here is the entire check list:

• Length. Word count, flagged when it falls outside a typical range for a resume.
• Contact details. Whether an email address, a phone number, and a portfolio or profile link are present.
• Section headings. Whether recognisable experience, education, skills and summary sections can be found.
• Bullet points. Whether experience is broken into bullets rather than paragraphs.
• Quantified results. How many genuine achievement metrics appear — and the checker shows you the exact snippets it counted.
• Weak phrasing. Matches against a list of filler phrases such as "responsible for" and "team player".
• Run-on lines. Very long single lines, which usually indicate a bullet doing too much.

A correction worth recording here. The metrics check used to match any digit sequence. That meant the digits in a phone number and the years in an education section were counted as career achievements, so a resume with no measurable results at all could be told it made good use of numbers — the exact opposite of the advice the tool exists to give. Separately, the phone-number check failed on (555) 123-4567, the most common US format, so people with a correctly formatted number were told to add one. Both were fixed on 8 August 2026 and are recorded in the changelog. The metrics check now excludes contact blocks and bare calendar years, and requires a genuine quantity signal.

Job description matcher: how terms are scored

The matcher compares a posting against your resume. It does not produce an "ATS score" and cannot tell you whether you will be shortlisted. What it does:

• Segments the posting. Terms under a requirements or qualifications heading are weighted well above terms under "preferred" or "nice to have". Trailing boilerplate — equal-opportunity statements, benefits blurbs, "about us" — is weighted to zero.
• Removes noise. Common English stopwords and an authored list of job-posting boilerplate phrases are discarded before anything is scored.
• Collapses word forms. Light stemming so manage, managed and management are treated as one term.
• Weights by specificity. A common English word scores lower than a distinctive one, so a named tool, system or credential outranks "communication".
• Handles abbreviations. A term present only as an acronym is reported as such, with a prompt to spell it out once, rather than being scored as a match or a miss.
• Flags over-use. A term repeated unusually often in your resume is flagged, because keyword stuffing reads badly to the human who eventually opens the file.

For why this is worth doing at all — and why the widely repeated claim about automatic rejection is not — see what an applicant tracking system actually does.

The writing tools and their validators

The bullet generator, the summary generator, the cover letter generator and the STAR answer builder all work the same way: they arrange material you supply into a structure, using authored sentence templates. They invent nothing. If you give them no number, they cannot produce a bullet containing one.

Generated candidates are checked before they are shown, and failures are discarded rather than displayed. The bullet generator requires a past-tense verb in first position, at least one figure, a length under thirty words, no first-person pronouns, no filler phrase from an authored list, and no repeated leading verb across the set it returns. The STAR builder checks that the result names an outcome, that the action is written in the first person singular — flagging every "we" — and that the whole answer fits the ninety-second spoken band.

A second correction. The bullet generator previously assembled some variants by placing a verb phrase into a slot that expected a noun, producing ungrammatical output such as "Increased results by the company social media accounts". It was rebuilt around the X-Y-Z formula with the validator described above, on 8 August 2026.

Example figures shown on profession pages are illustrative and are marked as placeholders to replace with your own. A number you cannot defend when an interviewer asks how you arrived at it is worse than no number.

Application tracker: what it derives

The application tracker stores nothing but what you type, in this browser only. It hardcodes no external reference data — no benchmark response rates, no "typical" conversion figures — because we have not measured any and the numbers circulated online are almost never attributable. What it does build in are four conventions, and they are conventions rather than findings: a follow-up prompt at 14 days after applying, adjustable on the page between 1 and 365; a 12-week window for the applications-per-week chart; weeks that run Monday to Sunday; and the four funnel stages Applied → Phone screen → Onsite → Offer.

Everything beyond the table is derived from your own rows at render time, and always from the whole dataset — the search box and the status filter change which rows the table shows, never the panels below it. Every day count is taken at UTC midnight on both ends, so a figure cannot gain or lose a day because of the hour at which you opened the page.

• Funnel counts per stage, where each row is counted at the furthest stage it ever reached rather than at its current status — an onsite that ended in a rejection is still an onsite. Each stage after Applied is divided by the stage immediately before it, never by your total; the Applied figure is shown as a share of every row tracked, saved leads included. Bars are scaled against the applied count. Rates below 10% keep one decimal, because the difference between 2% and 2.5% matters at the top of a funnel. At small sample sizes these percentages are close to meaningless — three applications and one screen is not a 33% screen rate in any useful sense — so read the counts, not the percentages, until you have real volume.
• Median days to first response — a median rather than a mean, because a single recruiter who answers four months later would drag an average somewhere useless. It is computed only from applications that actually received a reply. Rows still waiting are excluded and reported separately, which means the figure is optimistic by construction: it describes how fast replies arrive when they arrive, not how likely a reply is. Rows whose reply date falls before their applied date are excluded too, and counted back to you so you can correct them. Below five data points it is labelled a hint rather than a number.
• Next actions due or overdue, from the date you set on each row that is not yet rejected or ghosted.
• The follow-up queue, listing applied rows with no logged reply after your chosen threshold of days.
• Stale leads — roles you saved but never applied to.

Export is available as both JSON and CSV, and import accepts the JSON form. This is deliberate: a tool that holds months of your job search in browser storage and offers no way out of it is a trap, not a convenience. Clearing your browser data deletes everything, and we have no copy.

The pay calculators and their formulas

Every pay calculator states its formula on its own page. All of them work in gross figures. None models your full tax position, and none is tax advice. The core arithmetic:

Two modelling decisions worth naming. The offer comparison tool values public-company restricted stock at the current price but treats private-company equity and stock options as speculative and shows them separately, because a private valuation is not a sale price and options can end up worth nothing. And the freelance and contract calculators deliberately omit the qualified business income deduction, which makes their estimates conservative — the safe direction to be wrong in.

Published figures we rely on

Where a calculator needs an externally published figure, that figure appears in the page text with its year and a link to the source that publishes it, and it is editable. The sources used are:

• Social Security wage base — Social Security Administration.
• FICA and self-employment tax rates — IRS Topic 751.
• Retirement contribution limits — IRS retirement topics.
• Standard mileage rate — IRS standard mileage rates.
• Supplemental wage withholding — IRS Publication 15.
• Federal minimum wage and overtime rules — US Department of Labor, Wage and Hour Division.
• Inflation (CPI-U) — Bureau of Labor Statistics.

These change with the tax year. They are reviewed annually and each lives in a single marked block in the source, so an update is one edit. If you find one out of date, that is a bug — tell us and it will be fixed and logged.

What we deliberately have not built

Some popular tools cannot be built honestly as a free web page, and their absence here is a decision rather than an oversight. There is no take-home pay or withholding calculator, no unemployment benefit estimator, no retirement growth projection, and no cost-of-living index. The reasoning for each is in the editorial policy.

If you think a check is wrong, a formula is misapplied, or a figure is stale, email hello@freecareertools.com. Corrections are prioritised over everything else and are logged publicly.