People analytics is the use of data about people at work to make better decisions about them. It takes information an organisation already holds, such as who it hires, how they perform, what they are paid and when they leave, and uses statistics to answer a business question: which hiring channels produce people who stay, whether pay is fair across groups, which managers lose their teams, or what a reorganisation will cost. The term is often used interchangeably with HR analytics, workforce analytics and talent analytics.
The definition is short. Doing it well is not. Most people analytics that goes wrong does so for the same few reasons: the question was vague, the data measured something else, or a pattern was read as a cause. This guide covers what people analytics is, where it came from, the metrics that matter, how teams mature, and then works through a real example on the human-capital disclosures of every company listed in Japan, including the mistake that a naive analysis of that data makes.
What people analytics is, and what it is not
People analytics sits on top of HR reporting but is not the same thing.
| HR reporting | People analytics | |
|---|---|---|
| Question | What happened? | Why did it happen, and what should we do? |
| Example | Turnover was 14% last year. | Turnover among engineers hired through agencies is twice that of referrals, and most leave in their first year under three managers. |
| Output | A dashboard or monthly pack | A decision: change the channel mix, coach or move three managers |
| Methods | Counts, ratios, trends | Comparison across groups, regression, survival analysis, experiments, text analysis |
Three things separate analytics from reporting. It starts from a decision someone has to make, not from the data that happens to be available. It compares groups, so that a number means something (14% turnover is high or low only against something). And it tries to separate cause from coincidence, which is where most of the difficulty lies.
The names overlap. HR analytics usually means analytics run by and for the HR function. People analytics became the more common term in the 2010s, partly because Google named its team that way, and signals that the questions belong to the whole business, not to HR. Workforce analytics leans towards headcount, cost and planning. In practice the work is the same.
A short history
People analytics is newer as a name than as an idea.
| Year | Milestone |
|---|---|
| 1911 | Frederick Winslow Taylor publishes The Principles of Scientific Management, arguing that work can be measured and improved by study rather than rule of thumb. |
| 1924–1932 | Researchers at Western Electric's Hawthorne Works near Chicago study how lighting, breaks and supervision affect output. The lasting lesson was that being observed, and the social setting of work, changed results. |
| 2003 | Michael Lewis's Moneyball shows a baseball team beating richer rivals by valuing players on data rather than scouts' judgment, and gives business a story for analytics about people. |
| 2009–2012 | Google's people analytics team studies what makes a good manager (Project Oxygen, from 2009) and then what makes an effective team (Project Aristotle, from 2012), and publishes the results. |
| 2010 | Thomas Davenport, Jeanne Harris and Jeremy Shapiro publish "Competing on Talent Analytics" in Harvard Business Review. |
| 2014 | Wharton holds its first People Analytics Conference, a sign the term had arrived. |
| 2018 | The International Organization for Standardization publishes ISO 30414, guidelines for internal and external human-capital reporting. Japan's People Analytics & HR Technology Association (ピープルアナリティクス＆HRテクノロジー協会) is founded the same year. |
| 2020 | The US Securities and Exchange Commission requires listed companies to describe their human capital in annual reports, to the extent it is material. |
| 2023 | Japan requires listed companies to report women in management, fathers' childcare-leave uptake and the gender pay gap in their annual securities reports. |
| 2025 | A new edition, ISO 30414:2025, turns the human-capital reporting guidelines into requirements that can be audited. |
Disclosure rules changed the field. For most of its history people analytics was an internal tool. Since 2020 in the US and 2023 in Japan, some people metrics are public, audited alongside the accounts, and comparable across companies. That makes it possible to analyse a whole market, not only one employer, which is what the worked example below does. The evolution of HR timeline puts these dates alongside the models HR adopted in the same years.
The questions people analytics answers
Almost every useful project answers one of six kinds of question.
| Area | Typical question | Data | Common method |
|---|---|---|---|
| Hiring | Which sources, assessments and interviewers predict who succeeds and stays? | Applicant tracking, assessments, performance, tenure | Comparison of cohorts, regression |
| Retention | Who is likely to leave, when, and why? | Exits, tenure, pay, manager, engagement survey | Survival analysis, classification models |
| Pay and equity | Is pay fair across gender, age and other groups once role and level are accounted for? | Payroll, grades, job architecture | Regression on pay with controls |
| Performance and managers | What do our best managers and teams do differently? | Ratings, engagement, 360 feedback, team results | Comparison, text analysis of comments |
| Workforce planning | How many people, with which skills, will we need, and at what cost? | Headcount, skills, attrition, business plan | Scenario models |
| Wellbeing and engagement | Where is burnout building, and does what we do about it help? | Surveys, overtime, leave, absence | Trend analysis, before-and-after comparisons |
The metrics that matter
There are hundreds of HR metrics. A short list covers most decisions, and most of them appear in ISO 30414's recommended set or in Japan's disclosure rules.
| Metric | What it tells you | Watch out for |
|---|---|---|
| Turnover rate (voluntary and total) | How many people leave, and whether the ones you want to keep are among them | Annual rates hide first-year exits; split by tenure and performance |
| Time to fill / time to hire | How long roles stay open | A fast hire who leaves in six months was not fast |
| Quality of hire | Whether new hires perform and stay | Needs a definition agreed before you measure it |
| Cost per hire | Recruiting spend per hire | Cheapest channel is rarely the best |
| Engagement | How people feel about their work, from surveys | Response rates and who answers matter more than the score |
| Gender pay gap | Average pay of women against men | The raw gap mostly reflects who holds senior roles; analyse both raw and adjusted |
| Women in management | Share of managers who are women | Definitions of "manager" vary by company and country |
| Fathers' leave uptake | Share of eligible fathers who took childcare leave | Small denominators swing between 0% and 100% |
| Span of control | People per manager | Too wide and coaching stops; too narrow and cost rises |
| Revenue or profit per employee | Output per head | Driven by business model and outsourcing as much as by people |
| Absence and overtime | Load on the workforce | Under-reporting is common where long hours are the norm |
How a people analytics project works
- Start with the decision. "Should we stop using agency X?" beats "Let's look at turnover." Write down what you would do differently for each possible answer before you look at data.
- Find the data, and check what it measures. A performance rating measures a manager's judgment as much as performance. Exit-interview reasons measure what people are willing to say.
- Compare like with like. Every finding is a comparison. Decide which groups to compare and what to hold constant (role, level, location, tenure) before running anything.
- Separate cause from coincidence. Ask what else differs between the groups. The strongest evidence comes from experiments or from changes that hit one group and not another.
- Make the decision, and measure it. Agree in advance what result would mean the change worked, and check it.
How teams mature
Teams typically move through four stages. The version below follows the four-level model Josh Bersin's research firm published in 2012, which ends in predictive analytics; the stage names vary between versions.
| Stage | What the team does | Typical output |
|---|---|---|
| 1. Operational reporting | Counts and compliance numbers, often by hand | Headcount and turnover reports |
| 2. Advanced reporting | Regular, automated reporting with benchmarks and trends | Dashboards for leaders |
| 3. Strategic analytics | Statistical analysis that explains causes and segments the workforce | Answers to specific business questions |
| 4. Predictive analytics | Models that forecast outcomes and test interventions | Risk scores, scenario plans, experiments |
Most organisations stay at stages one and two, and that is not a failure. A team that reliably answers three important questions a year adds more than a predictive model nobody trusts.
A worked example: what Japan's disclosures show
Since the fiscal year ending March 2023, every company that files an annual securities report (有価証券報告書) in Japan has had to report three people figures where it is required to publish them under the women's advancement and childcare leave laws: the share of managers who are women, the share of eligible fathers who took childcare leave, and the gender pay gap. Together with headcount, average age, tenure and salary, which were already required, that is a public people-analytics dataset covering a whole stock market. The Hardest Market holds the latest figures for all 3,789 listed filers; 2,539 report all three of the new measures.
The headline numbers first. The median listed company reports that 9.1% of its managers are women (3,035 companies report the figure; 111 report none, and 277 report 30% or more). Women's average pay is 70.2% of men's at the median company. The median share of eligible fathers taking childcare leave is 82%, and 881 companies report 100% or more (the legal formula can exceed 100%, because leave taken this year can relate to children born last year).
A question worth asking: do companies where more fathers take childcare leave also have more women in management? The idea is plausible. If men share care, women's careers are interrupted less.
Ask it naively, and the answer looks dramatic:
The same question, all companies mixed together
All listed companies that report both figures (n=2,625). Small companies, where one or two fathers make 0% or 100%, crowd the top group.
Source: annual securities reports of 3,789 listed companies (latest filing to 2026-10-14), parent-company figures. Medians. The Hardest Market.
The top group stands out at 15.1%, more than twice the others. But it is mostly an artefact. In a company with two eligible fathers, one taking leave makes 50% and two makes 100%. Small companies crowd the extremes of the leave measure, and small companies also report more women managers for reasons that have nothing to do with leave:
Smaller companies report more women managers
Listed companies grouped by parent-company headcount (n=3,009). Median share of managers who are women.
Source: annual securities reports of 3,789 listed companies (latest filing to 2026-10-14), parent-company figures. Medians. The Hardest Market.
So compare like with like. Among companies with 1,000 or more employees, where the percentages rest on enough people to mean something, the pattern is smaller but steady:
Where more fathers take leave, more women manage
Companies with 1,000+ employees (n=709), grouped by fathers' childcare-leave uptake. Median share of managers who are women.
Source: annual securities reports of 3,789 listed companies (latest filing to 2026-10-14), parent-company figures. Medians. The Hardest Market.
Women's share of management rises from 4.9% in the quarter of large companies with the lowest fathers' leave uptake to 8.9% in the quarter with the highest. That is a real association, and it holds within a group of comparable companies. It is still not proof that leave causes promotion. Both may follow from a third thing, such as an industry that hires more women or a management team that pushes on both. Industry, in particular, swamps almost everything else:
Industry matters more than almost anything else
Median share of managers who are women: the four highest and four lowest of the TSE 33 industries (industries with 20+ reporting companies).
Source: annual securities reports of 3,789 listed companies (latest filing to 2026-10-14), parent-company figures. Medians. The Hardest Market.
The median services company reports 19.1% women managers; the median iron and steel company, 2.8%. Any comparison across companies that ignores industry is mostly measuring industry.
Finally, a result that runs against intuition. More women in management barely moves the company-wide pay gap:
More women managers barely moves the pay gap
Companies grouped by share of women managers (n=2,807). Median women's average pay as a % of men's, all employees.
Source: annual securities reports of 3,789 listed companies (latest filing to 2026-10-14), parent-company figures. Medians. The Hardest Market.
Women's average pay rises only from 67.3% to 72.0% of men's between the companies with the fewest and the most women managers. Japan's gap is driven less by managers than by who works in which roles, on which contracts, and for how many years. A company that wants to close it has to look at job grades and career tracks, not only at the management count.
What the example teaches. Every step of the method above mattered. The question was specific. The data needed checking (a 100% leave rate at a company with one eligible father means little). The comparison had to hold size and then industry constant. And the honest conclusion is an association worth investigating inside a company, not a cause to announce. The company pages on this site show these figures for each listed company, with industry comparisons.
People analytics in Japan
Japan has come to people analytics through disclosure rather than through HR technology. The Ministry of Economy, Trade and Industry's Ito Report on human capital (人材版伊藤レポート, September 2020, updated May 2022) and the Cabinet Secretariat's guidelines on visualising human capital (人的資本可視化指針, August 2022, revised March 2026) pushed companies to treat people as an investment and to report on it. The 2023 disclosure rules then made three measures mandatory for listed companies.
Three features shape the work in Japan:
- Membership-type employment. Many large employers still hire graduates as members of the company and rotate them across roles, rather than hiring for a job (job-based vs membership-type employment). Job titles, grades and pay often do not map to a role, which makes comparisons across jobs harder than in job-based systems.
- Personal information law. Employee data is personal information under the Act on the Protection of Personal Information (個人情報保護法). Health information is "special care-required personal information" (要配慮個人情報), which generally needs the person's consent to collect. Tell employees what you collect and why.
- Small groups. Many Japanese subsidiaries of foreign companies employ a few hundred people. Splitting them by gender, age and grade quickly leaves groups too small to analyse or to report without identifying individuals. Aggregate or wait for more data.
Pitfalls
- Small numbers. Percentages on a handful of people swing wildly, as fathers' leave shows. Report counts next to rates.
- Confusing correlation with cause. The worked example is the standard trap. Ask what else differs between the groups.
- Measuring what is easy. Time to fill is easy to measure and easy to game. Quality of hire is hard and matters more.
- Biased history. A model trained on past decisions learns past bias. In October 2018 Reuters reported that Amazon had scrapped an experimental recruiting tool after it learned to mark down CVs that mentioned women's organisations.
- Surveillance. Analysis that employees experience as monitoring destroys the trust it depends on. Analyse groups, not individuals, unless there is a clear and disclosed reason.
- Metrics that become targets. Once a number is tied to a bonus, people manage the number. Keep some measures for learning only.
How to start
- Pick one decision that matters this year, such as where to hire, how to reduce first-year exits, or whether pay is fair, and answer it well.
- Fix the basics of the data: one record per person, consistent job and grade codes, clean start and leave dates. Most of the effort in people analytics is here.
- Benchmark externally. Compare your turnover, pay gap and women in management with companies of the same size and industry. In Japan, the annual securities reports make this possible for any listed peer.
- Put an analyst next to the business. The best people analysts know the decisions, not only the data.
- Publish the method, not only the number. Leaders trust analysis they can follow.
Global HR, Japan and talent acquisition side by side on one timeline, with the primary source for every date. We will email it to you.
Sent. Check your inbox in a few minutes.
Questions
What is people analytics in simple terms?
People analytics is using data about employees, such as hiring, pay, performance and turnover, to make better decisions about them. It goes beyond reporting what happened to explaining why and recommending what to do.
What is the difference between people analytics and HR analytics?
In practice they mean the same work. HR analytics is the older term and often refers to analytics run within the HR function; people analytics became common in the 2010s and signals that the questions belong to the whole business. Workforce analytics leans towards headcount, cost and planning.
What are examples of people analytics?
Common examples are predicting which employees are likely to leave, testing which hiring sources produce people who stay and perform, checking whether pay is fair across groups once role and level are accounted for, finding what the best managers do differently, and planning how many people with which skills a business will need.
What skills does a people analyst need?
Statistics and data handling (SQL, spreadsheets, a language such as R or Python), an understanding of how HR processes create the data, and the ability to turn a business question into an analysis and the result back into a decision. Survey design and research ethics help.
What is ISO 30414?
ISO 30414 is the International Organization for Standardization's standard for human-capital reporting. The first edition (2018) was a set of guidelines recommending metrics in areas such as costs, productivity, recruitment, turnover, diversity, leadership, health and safety, and skills. The 2025 edition, Requirements and recommendations for human capital reporting and disclosure, adds requirements that can be audited.
What people data do Japanese listed companies have to disclose?
Since fiscal years ending on or after 31 March 2023, the annual securities report must include the share of managers who are women, fathers' childcare-leave uptake and the gender pay gap (where the company is required to publish them under the relevant laws), together with policies on human-capital development. Headcount, average age, average tenure and average salary were already required.
Is people analytics legal in Japan?
Yes, within the Act on the Protection of Personal Information. Employers may use employee data for purposes they have specified, and must take extra care with sensitive data such as health information, which generally requires consent to collect. Analysing groups rather than individuals, and telling employees what is collected and why, keeps the work on safe ground.
Related guides
By Yan Sen Lu, Managing Director of Makana Partners and Editor of The Hardest Market.
Spotted an error, or want your organisation featured? Send a correction · Ask to be featured · Contact the editor