Workforce Demographics Profile
A profile of your workforce by any category, built so it is safe to share, with representation at the top beside representation overall.
Reads .xlsx and .csv inside your browser. Nothing is uploaded, and no individual row is ever part of a download.
The workforce
Drop a spreadsheet here
One row per person with the categories you want to profile: gender, ethnicity, department, level, location. A date of birth becomes age bands and a hire date becomes tenure bands; neither is shown.
A small number is a person
A table of gender by department with a 1 in it does not describe a population. It names somebody. So categories too small to show are combined, and any count still under the minimum is hidden. That is the easy half.
Hiding one number is not hiding it
If a row has one hidden number and a total, subtraction gives it straight back. So a second number is hidden beside the first, then every row and every column is checked again, and the totals of the rows and columns too, until no single sum can recover anything hidden. Statistical offices have done this by hand for decades. Here it is done for every table you build, including the ones you download, and it is tested against hundreds of random tables every time the site is built. A hidden figure always says “hidden”, never a blank, so nobody reads protection as a zero.
An overall share hides where people are
“Women are 45% of the workforce” and “women are a small fraction of the top two levels” are both true of the same file. With a level column whose labels carry numbers, each group’s share of the most senior levels sits beside its share of every level below, and the difference is tested against chance with all groups corrected together. The top levels are taken together until they hold at least ten people, and a share there built on fewer than the minimum is shown as “fewer than” rather than a number.
A missing answer changes what a share means
If a fifth of people did not answer a question, every share of the answers describes four fifths of the workforce, and the people who did not answer are rarely a random slice of the rest. Each profile shows the share of everybody and the share of those who answered, and says how many gave no answer. Answers such as “prefer not to say” keep your file’s own wording and are counted as no answer.
Ages and tenure, without the dates
A date of birth becomes an age band and a hire date a tenure band on the date you choose. Neither date is shown or downloaded, and an age under 14 or over 90 is treated as the data error it almost always is.
What this is not
It is not a statutory report. It uses the categories in your file and follows no government format or category rules, so it is a way to understand your data before any filing, not the filing. For comparing outcomes rather than headcounts, the Promotion and Pay Action Analyzer tests promotion and pay by group, and the Survey Cut Analyzer applies the same protection to survey results.