1. Home>
  2. Resources>
  3. Strategic & Sustainable HR>
Kiharatukkainen nainen hymyilee työpöydän ääressä

HR Analytics and People Data: A Practical Guide

HR analytics turns the data already sitting in your HR, payroll and time systems into evidence you can put in front of the board. Often called people analytics, it covers everything from counting heads accurately to understanding why one site loses twice as many people as another. The useful news is that most organisations get further by reporting a handful of measures reliably than by attempting prediction. This guide covers the metrics worth tracking and how to calculate them, how to get your data into a state you can trust, how to build a dashboard people actually read, what UK data protection law requires, and where to start.
 

At a glance

  • Start with the questions the business is already asking, then choose metrics to answer them. Choosing metrics first is how dashboards end up ignored. > Getting started
     
  • Five measures cover most needs: headcount and FTE, turnover, absence, time to hire and workforce cost. > Core HR metrics
     
  • Two teams can calculate different turnover rates from identical data, because the denominator and the period are rarely agreed. Definitions matter more than tools. > Core HR metrics
     
  • Poor data is the usual blocker, not missing software. Reconciling your HR headcount against payroll is the fastest way to find out where you stand. > Data foundations
     
  • A useful dashboard is built for one audience. The board, the HR team and line managers need different numbers, not the same screen at different sizes. > Dashboards
     
  • Diversity and health data are special category data, so you need a condition under Article 9 of the UK GDPR as well as a lawful basis, and group sizes large enough that nobody can be identified. > Data protection
     
  • Prediction comes after the basics. Most organisations get more from reliable monthly reporting than from a model they cannot explain. > Predictive analytics
     

    <h2 id="what-is-hr-analytics">What is HR analytics, and how is it different from people analytics and HR reporting?</h2>
    <p><strong>HR analytics is the practice of using data about your workforce to answer business questions.</strong> People analytics means the same thing: the terms are used interchangeably, with people analytics more common in larger organisations and in the academic literature, and HR analytics more common in everyday UK usage. Workforce analytics usually signals a wider scope that takes in contractors, shift patterns and operational data alongside employee records. The distinction matters far less than whether anyone acts on what you produce.</p>
    <p><strong>The real distinction is between reporting and analytics.</strong> Reporting tells you what happened: 14% turnover last year, 2.1% absence last month. Analytics tells you why it happened and what to do about it: turnover is 14% overall but 31% in one distribution centre, concentrated in the first six months of service, and the exit data points at shift patterns rather than pay.</p>
    <p>Most descriptions divide the work into four types.<br></p>
    <table>
     <thead>
       <tr>
         <th>Type</th>
         <th>What it does</th>
       </tr>
     </thead>
     <tbody>
       <tr>
         <td><strong>Descriptive</strong></td>
         <td><strong>What happened.</strong> Counts, rates and trends: headcount, turnover rate, absence rate. The foundation everything else sits on.</td>
       </tr>
       <tr>
         <td><strong>Diagnostic</strong></td>
         <td><strong>Why it happened.</strong> The same measures cut by site, manager, tenure, contract type or shift, so patterns become visible.</td>
       </tr>
       <tr>
         <td><strong>Predictive</strong></td>
         <td><strong>What is likely to happen.</strong> Models estimating future turnover, absence or demand. Needs clean history and real capability.</td>
       </tr>
       <tr>
         <td><strong>Prescriptive</strong></td>
         <td><strong>What to do about it.</strong> Recommended actions with a cost or benefit attached. Rare in practice.</td>
       </tr>
     </tbody>
    </table>
    <br>
    <p><strong>Most organisations sit somewhere in the first two, and that is a respectable place to be.</strong> The CIPD positions people analytics as part of evidence-based HR rather than a specialist technical discipline, and its <a href="https://www.cipd.org/uk/topics/people-analytics/" target="_blank">people analytics resources</a> are a good starting point. If you are ever asked to report workforce data externally, ISO 30414 is the international reference standard for human capital reporting.</p>

      The core HR metrics and how to calculate them

      You do not need thirty measures. You need five or six that everyone agrees on, calculated the same way every month.
       

      Workforce shape

      Start with the questions of who you employ and in what shape.
       

      Metric Definition and formula
      Headcount The number of people employed on a given date. Simple to state, easy to get wrong. Decide whether you count people on long-term absence, on parental leave, on fixed-term contracts, and whether contractors appear at all. Write the decision down.
      Full-time equivalent (FTE) Headcount adjusted for contracted hours. Contracted hours divided by your standard full-time week. Two people each working 17.5 hours against a 35-hour week are 2 headcount and 1.0 FTE. Use headcount for people questions and FTE for cost and capacity questions.
      Span of control Employees divided by managers. Useful for spotting layers that have crept in.


       

      In practice: the board number and the payroll number will disagree the first time you compare them. The reasons are usually mundane: a leaver recorded in one system and not the other, a starter added to payroll before the HR record exists.
       

      Turnover and retention

      Four measures that answer different questions. Do not use them interchangeably.
       

      Metric Definition and formula
      Turnover rate Leavers in the period divided by average headcount, times 100. Average headcount is best taken as the mean of monthly headcounts rather than the average of the opening and closing figures.
      Voluntary turnover The same calculation using only resignations. This is the number that tells you about retention. Total turnover mixes in redundancies, dismissals and retirements, which say something different.
      Retention rate Employees in post at the start who are still in post at the end, divided by the headcount at the start, times 100. Not the inverse of turnover: someone hired and lost within the period affects turnover but not retention.
      Stability index Employees with a year or more of service divided by the headcount a year ago, times 100. Shows whether you are keeping experience.
      First-year attrition Leavers with under 12 months' service as a share of all leavers. Points at recruitment, onboarding or a mismatch between the advert and the job.


       

      The denominator problem is the single most common source of argument. Take a business that employed 1,100 people in January, 1,300 in December and lost 180 people across the year. Divide by the average headcount of 1,200 and turnover is 15%. Divide by the opening headcount of 1,100 and it is 16.4%. Both are defensible. Only one can be your number, and it needs to be the same one next year.

      The same applies to the period. A rolling 12 months smooths seasonality; a calendar or financial year aligns with everything else the board sees. Pick one and make every system use it. Reducing attrition is a management topic in its own right, covered under employee retention.
       

      Absence

      Absence is where measurement and employment law meet most directly.
       

      Metric Definition and formula
      Absence rate Working time lost to absence divided by working time available, times 100. A team scheduled for 16,000 hours that loses 320 hours has a 2.0% absence rate. Decide up front whether you count calendar days, working days or hours, and whether long-term absence is included.
      Bradford Factor Spells squared, multiplied by total days lost. Weights frequent short absences more heavily than one long one. Four separate single days score 64. A single four-day absence scores 4.


       

      Important: the Bradford Factor is a market convention, not a legal requirement. Acas guidance on absence trigger points is clear that reaching a review point should start a conversation rather than automatically trigger disciplinary action, and that employers should consider adjusting review points or discounting absence related to disability or pregnancy. A scoring system that applies itself mechanically to a disabled employee who needs regular treatment is a discrimination risk, not an efficiency.

      For context, ONS puts the UK sickness absence rate at 2.0% of working hours in 2025, or 4.4 days per worker, and classes these as official statistics in development. Policy and process sit with absence management.
       

      Hiring

      Two of these are routinely confused, which matters when you are reporting improvement.
       

      Metric Definition and formula
      Time to hire Days from a candidate entering your process to accepting an offer. Measures how well you handle candidates once you have them.
      Time to fill Days from the vacancy being approved to the offer being accepted. Covers the whole pipeline, including time a requisition sat waiting for sign-off. Routinely confused with time to hire, and almost always the larger number.
      Cost per hire Internal and external recruiting costs divided by the number of hires. Internal costs, such as recruiter time and referral payments, are the part people forget.
      Offer acceptance rate Offers accepted divided by offers made, times 100. A falling rate is an early signal on pay position or candidate experience.


       

       

      Cost, engagement and development

      These are the measures most likely to land in a board pack.
       

      Metric Definition and formula
      Total employment cost Everything it costs to employ your people, not just salaries. Gross pay plus employer National Insurance, employer pension contributions, benefits and any apprenticeship levy. A figure that excludes employer on-costs will not survive contact with finance. Decide whether contractors are in or out, and label it accordingly.
      Employment cost as a share of revenue Total employment cost divided by revenue, times 100. The measure finance is most likely to recognise, and the easiest way to put a workforce conversation in business terms.
      Engagement score and eNPS Survey measures. eNPS is the percentage of promoters minus the percentage of detractors on a single recommendation question. Useful as a trend, weak as an absolute. Survey design belongs with employee engagement surveys.
      Training hours and internal mobility Training hours divided by average headcount; internal moves divided by average headcount, times 100. Internal mobility is a better signal of whether development is real than training spend is.
      Overtime as a share of hours Overtime hours divided by total hours worked, times 100. Often the first sign of a staffing gap, and it shows up in payroll before anywhere else.


       

      For the operational measures on the payroll side of the line, such as payslip accuracy and processing time, see payroll KPIs. If you are weighing the cost of running payroll internally, our guide to payroll outsourcing costs breaks down the comparison.
       

      Diversity, pay gaps and the equal pay distinction

      Diversity reporting covers workforce composition by characteristic, plus representation at each level and in each pay band. Two things are conflated constantly and worth separating.

      • Equal pay is the legal right under the Equality Act 2010 to the same pay for the same work, work rated as equivalent, or work of equal value. An equal pay problem is a legal problem.
      • A pay gap is the difference in average pay between two groups across the whole organisation, regardless of role. It is usually structural, about who holds which jobs, and you can have a substantial gap with no equal pay breach at all.

      Saying "we have no pay gap because we pay everyone in the same role the same" conflates the two, and tends to be picked apart quickly.

        Getting your people data in order

        Almost every analytics project that stalls, stalls on data rather than tools. The good news is that the fixes are unglamorous and achievable.
         

        Where the data lives

        Most organisations draw on six or seven sources: the HR system, payroll, time and attendance, applicant tracking, learning records, engagement surveys and finance. Each was bought to do a job, not to feed a dashboard, and each holds a slightly different version of the truth.

        In our experience, the most common blocker is not a missing tool. It is two systems disagreeing about who left and when. Payroll records a leaving date driven by the final pay run. HR records the last working day. Finance sees the cost stop in a third month. Until those agree, every turnover number you produce is arguable.
         

        The reconciliation worth doing first

        Pull your HR headcount and your payroll headcount for the same date and compare them person by person. The differences fall into a few categories: leavers processed in one system only, starters not yet set up, people on unpaid leave, contractors sitting in one and not the other. Resolving that list once is often worth more than any new reporting tool, and it gives you a defensible headcount to build on. Doing it continuously is one of the practical arguments for integrated HR and payroll systems.
         

        A data quality checklist

        • Duplicate records, usually from rehires or a system migration
        • Missing or inconsistent leaver dates, the single biggest distorter of turnover
        • Cost centres that do not match finance, so workforce cost cannot be allocated
        • Contract type and FTE gaps, including FTE held as a percentage in one system and a decimal in another
        • Free-text fields where a list should be, especially leaving reason, which makes exit analysis impossible after the fact

         

        Definitions and ownership

        Agree what counts as a leaver, how FTE is calculated and which date is the leaving date, and write it down in one place. Then give every number an owner. A measure nobody owns drifts within two quarters.
         

        If everything lives in spreadsheets

        That is a legitimate starting point, not a confession. A clean monthly extract with consistent columns will get you further than a reporting tool pointed at messy data. What spreadsheets do badly is version and access control, so keep a single source file, restrict who can edit it, and move on when the admin outweighs the benefit. Smaller organisations weighing up a first system can look at HR software for small businesses.

        Important: retention periods cut both ways. You cannot analyse records you have lawfully deleted, and you should not keep records longer than you need simply to enable analysis. ICO guidance on keeping employment records covers what to consider.

          Building an HR dashboard people actually use

          A dashboard is built for one audience, answering questions that audience actually asks. The most common failure is a single screen intended to serve everybody, which ends up serving nobody.
           

          Three audiences, three dashboards

          • The board or executive team. Five to eight measures, trends over at least four quarters, with commentary. Headcount and FTE, turnover, workforce cost and cost as a share of revenue, absence, and one or two measures tied to current priorities. They want direction of travel and the reason for it.
          • The HR team. More measures and more cuts: by site, function, manager, tenure and contract type. This is the diagnostic layer, where you answer the questions the board asks next.
          • Line managers. Their team only, three or four measures, with enough context to know whether their number is unusual. A manager who sees a figure with no comparison does nothing with it.

           

          What makes one useful

          Trends beat snapshots. A single month's absence rate tells you almost nothing; twelve months with a comparison tells you a lot. Context beats precision. A turnover figure next to last year and the organisation-wide average is interpretable; the same figure to two decimal places on its own is not. Drill-down beats detail on the face. Keep the front screen short and let people click through.

          What to leave off matters as much as what goes on. A dashboard with 20 measures gets looked at once. If a number would not change a decision or prompt a question, it belongs in a report.
           

          Tools

          Four categories: spreadsheets, the reporting built into your HR system, general business intelligence tools such as Power BI or Tableau, and specialist people analytics platforms. Most organisations should exhaust the reporting in the systems they already own before buying anything. The constraint is rarely the visualisation layer. It is whether the underlying data is joined up and trusted. Broader selection questions belong with HR software.

            Using people data to answer real questions

            The point of all of this is to answer a question someone has actually asked. Four worked examples.

            Example: turnover concentrated in one place. A 1,200-person logistics business is asked by its board why warehouse turnover has doubled. HR, payroll and time data sit in three systems. Joining them shows the overall rate hides the real pattern: the increase is in one site, among people with under six months' service, and almost entirely on the night shift. The action is a shift-pattern review and a change to onboarding at that site, not an organisation-wide pay response.

            Example: absence that follows a pattern. Absence sits at 3.1% against a company average of 2.2%. Cut by team and day of week, it clusters on the first shift after a weekend in two teams with the same manager. The action is a conversation about workload and management practice, with occupational health support where needed, rather than a tightening of trigger points.

            Example: a workforce cost forecast finance will accept. HR provides establishment, planned recruitment, attrition assumptions and the pay review position; finance adds employer on-costs and inflation. The value is not the spreadsheet. It is that both teams forecast from the same headcount and the same definitions, so the number stops being disputed.

            Example: a pay equity review you run for yourself. Separate from any statutory publication, you compare pay within job levels and roles, controlling for tenure and location, and look at where new starters sit in each band. Done properly this is analysis with legal implications, so involve HR, reward and legal from the start, and expect to act on what you find. Longer-term resourcing questions sit with workforce planning.

              What you are required to publish

              Analytics and statutory reporting are different jobs, but they draw on the same data, so it is worth knowing what the law already asks of you.

              Gender pay gap reporting. Employers with 250 or more employees on the snapshot date must report. The snapshot date is 5 April for private, voluntary and most other employers, and 31 March for most public authorities, with publication due within a year. Six measures are required: mean and median gender pay gaps in hourly pay, mean and median bonus gaps, the proportion of men and women receiving a bonus, and the proportion of men and women in each hourly pay quarter. GOV.UK holds the employer guidance and the reporting service. Calculating and publishing belongs with gender pay gap reporting.

              Action plans. Under the Employment Rights Act 2025, employers with 250 or more employees can publish an action plan alongside their gender pay gap data, setting out at least one action to reduce the gap and at least one to support employees experiencing the menopause. Publication has been voluntary since 6 April 2026, and GOV.UK's action plan guidance states that, subject to legislation, it becomes mandatory from spring 2027. An action plan needs evidence behind each action and a way to track progress, which is a data question now rather than later.

              Ethnicity and disability pay gap reporting. The Government confirmed in March 2026 that it intends to make this mandatory for employers with 250 or more employees, using the same measures, dates and service as gender pay gap reporting, plus a workforce breakdown and the proportion of employees who did not state their ethnicity or disability status. This is not yet law. It requires the Equality (Race and Disability) Bill, which has not been introduced to Parliament, followed by regulations. If you are configuring an HR system now, make sure you can capture and report the data rather than publishing anything you are not required to.

              Recording sex. GOV.UK updated its gender pay gap employer guidance in May 2026, replacing its section on recording employees' gender with one on recording employees' sex, and addressing how to treat an employee who holds a Gender Recognition Certificate. If your HR system records this field, check its configuration against the current guidance.

                Data protection, ethics and employee trust

                People analytics is processing personal data, and the rules are not optional. This is the area where UK guidance is most specific and most often skated over.
                 

                Lawful basis and the Article 9 condition

                Every use of personal data needs a lawful basis under Article 6 of the UK GDPR. Where the data is special category data, you need a condition under Article 9 as well. Special category data includes health, racial or ethnic origin, religious or philosophical beliefs, trade union membership, sex life and sexual orientation, genetic and biometric data. Absence data that reveals a health condition, and the diversity data behind pay gap reporting, both fall inside it. The ICO's special category data guidance sets out the conditions.

                For equality monitoring specifically, paragraph 8 of Schedule 1 to the Data Protection Act 2018 provides a condition covering racial or ethnic origin, religious or philosophical belief, health and sexual orientation, where processing is necessary to identify or keep under review the existence or absence of equality of opportunity or treatment. It comes with real limits: the processing must not be for measures or decisions about a particular individual, must not be likely to cause substantial damage or distress, and must stop if an employee objects in writing. Relying on it also requires an appropriate policy document under paragraph 5.

                Key takeaway: that condition is built for exactly the analysis you want to do, and it deliberately stops short of individual decisions. If you find yourself using equality monitoring data to make a decision about one person, you have left the condition behind.
                 

                Aggregation and re-identification

                Circulate aggregated figures, not individual records. Set a minimum group size below which you do not report a breakdown, and apply it consistently. If a department has four people in a category, a percentage effectively names them. Pseudonymisation helps during analysis but it is not anonymisation: if you can reverse it, the data is still personal data.
                 

                DPIAs

                A data protection impact assessment is required where processing is likely to result in high risk, which a people analytics programme using special category data, profiling or large-scale monitoring often will. Doing one early is also the fastest way to surface whether you need the data at all.
                 

                Automated decision-making

                The UK rules changed in 2026. Section 80 of the Data (Use and Access) Act 2025 replaced Article 22 of the UK GDPR with new Articles 22A to 22D, in force from 5 February 2026. The position is now more permissive for significant decisions taken solely by automated means where no special category data is involved, but the safeguards under Article 22C still apply: tell the person, let them make representations, give them human intervention and let them contest the outcome. Where special category data is involved, Article 22B keeps tighter conditions.

                Two practical points. A decision is solely automated only where there is no meaningful human involvement, and a manager rubber-stamping a system's ranking is unlikely to count. And the ICO's guidance in this area is being revised following the Act, so check the current position on the ICO site rather than relying on older material.
                 

                Trust

                Tell employees what people data you hold, what you use it for and what you do not use it for. This is a transparency obligation and it is also self-interested: declaration rates on ethnicity and disability depend almost entirely on whether people believe the answer will be used fairly. A workforce that does not trust the process gives you data you cannot report on.

                  This guidance is general

                  This article explains how UK data protection and reporting rules apply to people analytics in general terms. What applies to your organisation depends on the data you hold and what you do with it, so take your own data protection advice before relying on a particular condition or starting a new programme of analysis.

                    Predictive analytics and AI: when you are ready

                    Prediction needs clean history, and most organisations do not have it yet. A turnover model needs several years of consistent leaver data with reliable dates and reasons. If your leaver dates are still disputed, a model will simply repeat the problem with more confidence.

                    Three things to weigh before starting. Bias: a model trained on past decisions learns past patterns, including the ones you are trying to correct. Explainability: if you cannot explain to an employee or a tribunal why a model flagged them, you should not be acting on it. Proportionality: predicting individual attrition is a far heavier intervention than forecasting attrition at team level, and the team-level version answers most planning questions anyway. Wider uses are covered in AI in HR.

                      Getting started: a practical roadmap

                      1. Write down the three questions the business is asking. Not the metrics. The questions, in the words the board or finance used.
                      2. Choose the smallest set of measures that answers them. Five or six. Define each one precisely, including the denominator and the period.
                      3. Audit the data behind those measures only. Reconcile HR against payroll for a single date. Fix leaver dates. Ignore everything the chosen measures do not need.
                      4. Build one dashboard for one audience. Make it monthly, make it trended, and put a sentence of commentary next to it.
                      5. Set a review cycle. Monthly to produce, quarterly to ask what it changed. A measure that has not informed a decision in two quarters comes off.

                      Capability matters more than headcount here. Data literacy in the HR team, a working relationship with finance for cost definitions and with IT for access, and someone who owns each number. That combination gets further than a new hire with a title.

                        Why SD Worx

                        We have been running payroll and HR for 80 years, and we process more than 6 million payslips a month for over 90,000 customers worldwide. That scale is relevant here for one reason: most of the people data worth analysing passes through payroll, and joining it up is where the practical difficulty sits.

                        Our technology is HMRC-recognised and GDPR-compliant, and we hold ISO 27001, ISAE 3000 and ISAE 3402 Type II certification, which matters when the data includes health and diversity information. Our UK payroll professionals are CIPP-certified, and SD Worx representatives sit on UK Government consultation panels including the Rep Body Group and the Employer Payroll Group.

                        Briggs Equipment worked with us on payroll and HR automation, and their case study covers the operational side of getting people data into one place. We also publish the HR and Payroll Pulse on our publications page.

                          FAQ

                          What is HR analytics?

                          HR analytics is the practice of using data about your workforce to answer business questions, drawing on HR, payroll, time and recruitment systems. It covers everything from reporting headcount and turnover accurately to understanding why absence or attrition differs between teams. It is also called people analytics, and the two mean the same thing in practice.

                          What is the difference between HR analytics and people analytics?

                          There is no meaningful difference. People analytics is more common in larger organisations and in academic writing, HR analytics in everyday UK usage. Workforce analytics sometimes signals a wider scope including contractors and operational data. What matters is data quality and whether anyone acts on the output.

                          What HR metrics should I track?

                          Start with five: headcount and FTE, turnover split into voluntary and involuntary, absence rate, time to hire and total workforce cost. Those answer most questions a board or finance team will ask. Add diversity, engagement and internal mobility once the first five are reliable and agreed.

                          How do you calculate staff turnover?

                          Divide the number of leavers in a period by the average headcount over the same period and multiply by 100. Average headcount is best calculated as the mean of monthly headcounts rather than the average of the opening and closing figures. Use the same denominator and period every time, because changing either changes the answer significantly.

                          What should be on an HR dashboard?

                          Five to eight measures, chosen for one audience, shown as trends rather than single-month snapshots, with context such as the previous year or an organisation-wide average. A board dashboard typically covers headcount, turnover, workforce cost, absence and one or two priority measures. Anything that would not change a decision belongs in a report.

                          Is HR analytics allowed under GDPR?

                          Yes, provided you have a lawful basis under Article 6 of the UK GDPR and, where you use special category data such as health or ethnicity, a condition under Article 9. Paragraph 8 of Schedule 1 to the Data Protection Act 2018 provides a condition for equality monitoring, with limits attached including an appropriate policy document and a bar on decisions about individuals. Aggregate your reporting and complete a data protection impact assessment where the processing is high risk.

                          What tools are used for HR analytics?

                          Four categories: spreadsheets, the reporting built into your HR or payroll system, general business intelligence tools such as Power BI or Tableau, and specialist people analytics platforms. Most organisations should exhaust what their existing systems can report first, because the usual constraint is data quality rather than visualisation.

                          How do I get started with people analytics?

                          Write down the three questions the business is actually asking, choose the smallest set of measures that answers them, and audit only the data those measures need. Reconciling your HR headcount against payroll for a single date is usually the most useful first exercise. Then build one dashboard for one audience.

                            Want help turning your people data into something the board will act on?

                              Talk to us about your people data