I have been in HR for 21 years. I used HR analytics at Alfa-Bank, at the STB TV channel, at Wargaming, Preply and iDeals, and it helped me in every one of those jobs. Because I was good with data, I got harder tasks, more responsibility and new roles.
Below: what HR analytics is and how it differs from a metrics report, the three kinds of questions it answers, the six levels of analysis, what data you already have, how to get management on board, and a free business question builder for your first project.
What is HR analytics (people analytics)?
HR analytics, also called people analytics, is the application of evidence to management decisions about people. Hiring, pay, benefits, promotion, training, retention, succession, dismissal: every one of these is a decision, and every decision about people is a decision about money.
Here is a simple illustration I use in class. An employee earns $2,000 a month, which is $24,000 a year. If people stay with the company for five years on average, a hiring decision about one ordinary employee is a $120,000 decision. Yet many companies spend more time choosing a coffee machine for the kitchen that costs $1,000 a year.
When I worked at Alfa-Bank in 2006, I saw HR job ads asking for a psychologist who understands people. Today many HR vacancies ask for a data-driven approach. HR is still about understanding people, but also about deciding on data, not on intuition.
HR analytics vs HR metrics reporting: what is the difference?
Most people expect HR analytics to start with metrics. I do not start there. Metrics such as turnover or time to fill are raw material. Today you can ask an AI assistant for a list of metrics for any situation in a minute. If you need the formulas, they are in my article HR Metrics: 27 Formulas With Examples.
Analytics starts where a metric meets a question and a decision. A turnover report says how many people left. Analytics asks why, in which segment, what happens next and what to change.
What questions does HR analytics answer?
Like every kind of business analysis, people analytics answers three kinds of questions:
- Questions that give new knowledge. There are things you know you know, things you know you don't know, things you don't know you know, and the worst kind for a business: things you don't even know you don't know. That blind spot can cost millions a year.
- Questions that solve a problem. Something is not working in the business, and analytics finds where and why.
- Questions that evaluate decisions already made, so the next decision is better.
The survey question behind this case is simple enough to copy. For each factor of the job (content of the work, autonomy, recognition, feedback, communication processes, workplace and so on), ask: "In your opinion, in our company this factor is worse than in other companies, the same, or better?" Then look at the share of "worse" answers and read the open comments.
A problem-solving example. At the STB TV channel I headed recruiting and onboarding, and I noticed that people were leaving unevenly. I asked HR records for an export and calculated turnover by department: one department had 141% against a company average of 26 to 28%, and took about 20% of our recruiters' time. Exit interviews and calls to the people who had left showed the causes, problems with management and the difficulty of the work, and we started fixing them.
An evaluation example: correlate interview scores with performance ratings a year later to see which assessment methods predict well, or review the talent pool to see who grew, who left and why.
How to frame a business question for HR analytics
Every analytics project I run follows one chain: pain → question → data → decision. One pain, one question. The data is chosen for the question, never the other way round.
- Pain. What hurts the business, in people or in money. The four pains that managers raise most often: competitiveness in hiring and retention, voluntary turnover, employee engagement and the likelihood of people leaving.
- Question. One question that data can answer, ending with a question mark.
- Data. The metrics and sources that answer exactly this question.
- Decision. What the company will do differently once it has the answer. Start it with a verb.
The self-check: which decision will this number change? If you cannot answer, rewrite the question.
To turn a vague complaint into a sharp question, ask the manager who owns it: how do you recognize the problem, what measurements show it, what is causing it, where and when does it show up, and who is affected most?
Three questions you can copy, one of each kind (fictional examples):
New knowledge: Which factors of the job do our employees rate worse than at other companies?
Solve a problem: Why do new hires leave in their first three months, and under which managers does it happen more often?
Evaluate a decision: Do the people we rated highest at interview perform better after a year than those we rated lowest?
What data do you need to start?
The usual objection: we have no data, or no system. In practice there is more data than it seems: the HR system (HRIS), the applicant tracking system (ATS), payroll, employee surveys, exit interviews, performance reviews, timesheets and financial reports. Beyond exports, your tools are surveys, interviews, focus groups and correlation analysis.
I kept finding sources I had not thought of. At STB it was the access control system at the office entrance: a full year of who was at work. I matched it with vacation and sick-leave records and found people absent for more than three days with no vacation registered.
The six levels of HR analytics
Analytics has levels of maturity. The higher you go, the harder the analysis, and the more useful it is to the business. Companies usually start at the bottom and climb:
- Regular reports — what happened?
- Reports on request — how much and how often?
- Statistical analysis — why did it happen?
- Forecasting — what happens if the trend continues?
- Predictive modeling — what is most likely to happen?
- Optimization — what could happen in the best case if we change something?
Business cares more about the future than the past. Turnover is useful to know, but it is far more useful to know who of the people still here is most at risk of leaving. Last year's turnover cannot be changed; with a person at risk you can hold a stay interview and find out what would keep them.
How to sell HR analytics to management
When you need management's support for an analyst, time or a system, do not pitch "we will know our turnover and hiring metrics". Link analytics to financial results with one chain: employee results → process results → customer results → financial results. This is the logic of Kaplan and Norton's balanced scorecard: what we learn and change about people affects processes, customers and money.
The second way to show it is the ABC model: antecedents, behaviors, consequences. Something precedes people's behavior, and the behavior creates consequences. For example, people get low pay; because of that they are unhappy at work, speak badly of the company and update their CVs; the result is that they leave and company plans fall apart. If we don't like the consequence, analytics shows what to change in the antecedents to get different behavior and different results. Then we look at the new consequences and adjust the antecedents again.
In the conversation itself:
- Don't sell the tool. Find out what worries the manager, then show how analytics helps with that.
- Start with their pain. Hiring and competition for candidates, turnover, engagement and the likelihood of people leaving get almost any manager's attention.
- Ask for buy-in, not a sign-off. If you plan an employee survey, don't open with "let's run a survey". Open with the turnover, the weaker hiring and the declined offers that worry them, then describe the survey and ask them to back it, so that when employees ask "what is this survey?", the manager says "it's fine, HR is running it, fill it in".
Your first HR analytics project in five steps
- Write down why your company needs HR analytics. Name a decision that is made on gut feeling today and what will change when there are numbers.
- List your top five metrics: the ones you already calculate, and five you would like to.
- Write the questions you want analytics to answer. They depend on your role.
- Pick one question with a clear owner and a decision attached, and run it at the lowest level that answers it.
- Show the result to the decision maker in a meeting, not in an email, and agree what changes.
Common mistakes when starting with HR analytics
- Starting with a dashboard. Fifty metrics and no question. Nobody knows what to do with the numbers.
- Trusting the company average. An average of 26 to 28% hid a department with 141%. Always split by department, manager and tenure.
- Jumping to the obvious fix. "Raise salaries" looked obvious at Wargaming. A survey showed the cause was communication.
- Expecting AI to replace the basics. AI makes analysis faster, but setting the task, judging the result and correcting it still requires you to understand HR analytics.
One question for this week: which decision about people in your company costs the most, and what data was it based on?
I teach the whole path, from the first business question through segmentation, employee lifetime value, recruiting, turnover, surveys and regression to forecasts and experiments, in the HR Analytics Course. It is also part of the Pritula Academy subscription: $1,000 a year, 175+ courses, access for one year.
