Methodology
Every tool here follows the same few rules. This page says what they are, how each tool is tested before it goes live, and where the tools stop.
Two numbers, never one
Most HR figures come in a flattering version and an honest one, and both are usually true. A training report says 95% complete, counted per course; counted per person, a third of the workforce is missing at least one. A manager of four with three top ratings is at 75%; whether that means anything depends on how often chance does the same with four people. Where a figure has two readings, the tools show both, side by side, and say why they differ. The distance between them is usually the finding.
Every assumption is written down
Each result ends with a list of what it assumed: the date it was measured as of, how a month was counted, which rows were left out and why, what a setting was set to. None of these is hidden in the arithmetic. If the list says something you disagree with, change the setting and the answer changes with it.
Problems are named, not quietly fixed
Findings come in three kinds: a problem, something worth checking, and a note, each with a count and examples from your own file. When a tool has to decide something to read a file at all, such as which way round a date is written or whether 00123 and 123 are the same person, it says what it decided. A repair made silently is how a spreadsheet produces a confident wrong answer.
Small groups are held to the standard they deserve
Percentages flatter small groups at both ends, so a list sorted by percentage mostly sorts by size. Where a tool compares groups, such as managers, departments or recruiting sources, it tests each one with an exact method suited to small counts, corrects for having looked at every group at once, and prints how many findings chance alone would produce beside how many it found. Where a tool reports on people by demographic group, groups too small to report safely are withheld, along with any figure that would let them be worked out by subtraction.
Dates are where spreadsheets lie
A date like 3/4/2026 means March in one country and April in another, and a file often cannot say which. When it cannot, the tool asks rather than guesses. Months are counted by month and clamped to the end of a short one, so 31 January plus one month is 28 February, not 3 March. Dates in a workbook are read as the calendar date typed into the cell, never shifted by your time zone. And “today” is always a date you can set, so a result can be reproduced next week.
The tools assist; people decide
No tool here decides anything with consequences for a person: who is paid what, who is eligible for what, or whether anybody has complied with the law. The tools do not know employment law, tax rules or your policies. Where a rule matters, such as how often training renews or how long a benefits waiting period is, you set it, and the tool applies exactly what you set and lists it in the result. There are deliberately no tools for legal determinations such as overtime classification, leave eligibility or final pay deadlines. Those belong with people who are accountable for them.
How each tool is tested
Every tool’s arithmetic is checked by an automated suite of worked examples, where the right answer was worked out by hand before the tool was run against it. There are more than 4,500 of these checks across the 47 tools and the file reader they share, and all of them run before any change goes live.
Every tool is also loaded with the files that break software: an empty file, a single row, the wrong file entirely, rows with values missing or extra, cells holding Excel errors, formulas and thousands of characters, text saved in older encodings, workbooks with a cover sheet first or a password, and files of tens of thousands of rows. Each has to end in either a result or a plain sentence saying what is wrong with the file, never a crash.
The date at the foot of each tool page is when its method and tests were last checked. Every tool was checked on 12 September 2026.
The tests show that each tool does what it says. They cannot show that what it says is what your organization needs, which is why the assumptions are always on the page. No outside body has reviewed or certified these tools.
Your file stays with you
Every tool runs inside your browser. Your file is read there, worked on there, and never sent anywhere, and every page tells your browser to refuse any connection to another website, so it could not be sent even by mistake. The privacy page explains how that works and how to check it yourself in ten seconds.