Foundations of HR Analytics and Evidence-Based HRM

Introduction

Human resource management has moved from an administrative function concerned primarily with records, compliance and transactions to a strategic discipline concerned with organizational capability, employee experience and sustainable performance. This shift has been supported by the growth of digital HR systems, cloud platforms, workforce-management technologies and analytical methods. Yet possessing data does not automatically produce better decisions. HR analytics creates value only when reliable evidence is connected to a clearly defined organizational decision, interpreted within context and used responsibly.

HR analytics is therefore both a technical and managerial discipline. It requires knowledge of data structures, measurement, statistics and technology, but also an understanding of work design, motivation, leadership, organizational strategy, labour markets and employment law. Evidence-based human resource management (EBHRM) provides the broader decision philosophy: important HR decisions should be informed by the best available evidence, professional expertise, organizational context and the legitimate interests of stakeholders.

This chapter establishes the conceptual foundation for later work in workforce metrics, predictive modelling, experimentation, people dashboards and artificial intelligence. It distinguishes forms of analytics, explains the evolution and maturity of HR analytics, and shows how analytical findings should be translated into strategic decisions.

Learning Objectives

After studying this chapter, students should be able to:

1. Define HR analytics, people analytics, workforce analytics, HR metrics and evidence-based HRM.
2. Explain the historical evolution and maturity of HR analytics.
3. Distinguish reporting, descriptive, diagnostic, predictive and prescriptive analytics.
4. Relate analytical work to organizational strategy and decision theory.
5. Evaluate the quality, limitations and ethical implications of HR evidence.
6. Apply a structured approach to an HR analytics problem.

1. The Meaning and Scope of HR Analytics

1.1 HR data, metrics and analytics

HR data are recorded observations about employees, jobs, teams, processes and employment outcomes. Examples include absence records, recruitment dates, compensation, performance assessments, learning participation and employee survey responses. Data may be structured, such as fields in a human resource information system, or unstructured, such as interview notes, emails and open-text survey responses.

Information is data interpreted in relation to a question. A list of monthly exits is data; a comparison showing that voluntary exits are concentrated among employees with less than two years of tenure is information. Knowledge develops when information is connected to explanations, theory and experience. Decision value arises when knowledge improves a choice or action.

An HR metric is a defined measure used to monitor an HR activity, condition or outcome. Examples include time to fill, absence rate and training completion. Analytics goes further than measurement. It involves systematically examining data to understand patterns, test explanations, estimate future outcomes or recommend actions.

A useful distinction is:

• Metric: What is being measured?
• Report: What has been recorded or summarized?
• Analysis: What pattern, relationship or explanation can be identified?
• Decision: What should be done, by whom and under what conditions?

1.2 Related terms

The terms HR analytics, people analytics and workforce analytics are often used interchangeably, although their emphases differ.

HR analytics usually refers to the analysis of data generated by HR processes and systems, such as recruitment, compensation, learning and employee relations. People analytics is often broader and may include behavioural, collaboration, experience and operational data, while emphasizing employees as people rather than merely as administrative records. Workforce analytics commonly focuses on workforce composition, capacity, deployment, skills and labour supply, including workers outside conventional employment relationships.

These boundaries are not universal. The important issue is not the label but the decision scope, data provenance, analytical method and governance arrangements.

1.3 The unit of analysis

An analyst must identify the unit of analysis before calculating a metric or fitting a model. A unit may be an applicant, employee, position, team, business unit, month or employment episode. Confusing units creates serious errors. For example, an organization may have 1,000 employees but 1,150 employment records if some people changed roles or left and rejoined. A turnover calculation based on records rather than people may produce a misleading result.

The analyst should also distinguish between individual-level, team-level and organizational-level data. An association observed across departments does not necessarily apply to individual employees. This is known as the ecological fallacy. Conversely, an individual relationship may disappear when data are aggregated, a problem sometimes called the atomistic fallacy.

2. The Evolution of HR Analytics

2.1 Administrative HR and personnel records

Early personnel functions concentrated on payroll, attendance, employee files, industrial relations and legal compliance. Data were often paper-based, fragmented and retrospective. Measurement focused on activity and cost: headcount, payroll expenditure, absence days and vacancies. The dominant question was whether transactions had been completed accurately.

2.2 Human resource management and strategic alignment

The development of strategic HRM encouraged organizations to view employees and capabilities as contributors to competitive advantage. The resource-based view of the firm argued that valuable, rare, difficult-to-imitate and non-substitutable resources can support sustained advantage. Human capabilities, organizational knowledge and social relationships may meet these conditions when they are developed and coordinated effectively.

This perspective changed the analytical question from “How many employees do we have?” to “What capabilities are required to execute strategy, and how effectively are they being developed and deployed?”

2.3 HR scorecards and business-linked measurement

In the 1990s and early 2000s, HR scorecards and balanced-scorecard approaches encouraged links among HR activities, workforce outcomes and business results. The logic was not that every HR activity must have an immediate financial return. Rather, HR measures should clarify strategic contributions and causal assumptions.

For example, a service organization might hypothesize that supervisor coaching improves employee capability, which improves service quality, which increases customer retention. Each link requires conceptual justification and empirical examination. A dashboard that merely displays many indicators does not establish such a chain.

2.4 Digital HR and people analytics

Human resource information systems, integrated talent platforms and cloud technologies have increased the volume, speed and variety of workforce data. Organizations can now combine HR records with operational outcomes, learning systems, engagement surveys, scheduling data and external labour-market information. Advanced methods include text analysis, machine learning, network analysis and causal inference.

Technology has expanded analytical possibility but also increased risk. More data can produce more noise, proxy discrimination, surveillance and false precision. The central management challenge is therefore not simply acquiring data, but designing trustworthy decision systems.

2.5 A maturity perspective

Maturity is not a single ladder through which every organization progresses in a fixed sequence. Nevertheless, the following model is useful for diagnosis.

A mature organization does not necessarily use the most complex algorithm. It asks better questions, maintains trustworthy data, makes uncertainty visible, integrates professional judgment and evaluates whether decisions produce beneficial outcomes.

3. The Analytics Continuum

The familiar progression from reporting to prescriptive analytics is best understood as a set of increasingly decision-oriented questions. The categories can overlap; a single project may use several forms.

3.1 Reporting: What happened?

Reporting presents selected data in a structured form. A monthly report may show headcount, vacancies, absence and labour cost by department. Reports may be static or interactive, scheduled or produced on demand.

Reporting is essential for accountability and operational control. It is not analytically trivial: definitions, data quality, comparison periods and access controls matter. However, reporting normally does not explain why an outcome occurred or what action is best.

3.2 Descriptive analytics: What has happened and what patterns exist?

Descriptive analytics summarizes distributions, trends, differences and relationships. It may involve rates, averages, medians, cross-tabulations, visualization and segmentation.

A commonly used turnover rate is:

The numerator is the number of separations in the specified period. Average headcount is often calculated as the mean of opening and closing headcount:

These formulas assume that headcount is measured consistently and that the period and population are clearly defined. If workforce size changes sharply during the period, a monthly or daily average may be more appropriate. The analyst must specify whether the rate includes voluntary and involuntary separations, contractors, retirees or internal transfers.

Worked example. A business begins a quarter with 480 employees and ends with 520. It records 30 separations. Average headcount is (480 + 520) ÷ 2 = 500. The turnover rate is (30 ÷ 500) × 100 = 6%. This describes the period; it does not show whether turnover was harmful, avoidable or caused by a particular policy.

3.3 Diagnostic analytics: Why did it happen?

Diagnostic analytics investigates possible explanations. Methods include segmentation, cohort analysis, process analysis, interviews, comparative case analysis and statistical modelling. For example, exits may be compared by manager, tenure, shift, location, job family and pay position.

Diagnostic work must guard against spurious relationships. If employees who attend a particular course have lower turnover, the course may be effective, but participants may also differ in motivation, tenure or manager support. A correlation is not automatically a causal effect.

3.4 Predictive analytics: What is likely to happen?

Predictive analytics estimates future or unknown outcomes using historical data. Examples include forecasting staffing demand, estimating time to fill, predicting absence risk or identifying applicants likely to complete a recruitment process.

A prediction is not a fact and not a recommendation. It is an estimate conditional on data, model assumptions and a future environment. Predictive accuracy may decline when organizational policies, labour markets or employee behaviour change.

For binary outcomes, a model may estimate a probability such as P(Y=1|X), where Y is the outcome and X represents observed features. A probability of 0.70 means that the model estimates a 70% likelihood under its validation conditions; it does not mean that a particular employee will certainly experience the outcome.

3.5 Prescriptive analytics: What should be done?

Prescriptive analytics compares possible actions and their consequences. It may use scenario analysis, optimization, simulation, decision rules or formal experimentation. A scheduling model, for example, may recommend staffing levels subject to demand, working-time rules, skills, budget and employee preferences.

Prescriptive recommendations require explicit objectives and constraints. An apparently optimal solution may maximize coverage while damaging fairness, wellbeing or retention. Therefore, technical optimization must be combined with organizational values, legal requirements and stakeholder review.

4. Evidence-Based HRM

4.1 Meaning and foundations

Evidence-based HRM applies the principles of evidence-based management to people decisions. It does not mean that numerical data replace judgment. It means that decisions are made through a disciplined consideration of the best available evidence, rather than relying exclusively on habit, authority, fashion or anecdote.

Four evidence sources are particularly important:

1. Scientific evidence: Findings from credible research, including systematic reviews, high-quality studies and validated measures.
2. Organizational evidence: Internal data about employees, processes, performance and outcomes.
3. Professional expertise: The informed judgment of HR practitioners, managers and relevant specialists.
4. Stakeholder and employee perspectives: The experiences, preferences, values and interests of affected people.

Good practice considers the convergence and conflict among these sources. A research finding may not transfer directly to a particular workforce. Internal data may be precise but measure the wrong construct. Professional experience may identify context that a model misses. Employee voice may reveal harms not visible in organizational averages.

4.2 Evidence quality

Evidence should be evaluated according to relevance, credibility, validity, reliability, timeliness, completeness and ethical legitimacy.

• Validity concerns whether a measure or study supports the intended interpretation.
• Reliability concerns consistency of measurement.
• Representativeness concerns whether the data reflect the population to which conclusions will be applied.
• Causal credibility concerns whether an observed effect can reasonably be attributed to an intervention rather than to confounding factors.
• Practical relevance concerns whether the finding changes a decision sufficiently to justify its cost.

A statistically significant relationship may be operationally unimportant, while a modest average effect may matter greatly for a large workforce. Conversely, a strong association may be ethically unsuitable for individual decision-making.

4.3 Evidence and decision quality

Decision quality involves more than outcome quality because outcomes are partly affected by luck and external conditions. A well-reasoned decision can produce a poor outcome under adverse conditions; a poorly reasoned decision can appear successful by chance. Analytical governance should therefore assess the quality of the process: question definition, data selection, assumptions, alternatives, uncertainty, stakeholder effects and post-decision learning.

5. Strategy, Causality and Decision Theory

5.1 Linking analytics to strategy

Analytics should begin with a strategic or operational decision, not with the availability of data. A useful chain is:

Strategic objective → critical workforce capability → HR intervention → employee or workforce outcome → organizational result.

Suppose a hospital’s strategy emphasizes safe and responsive patient care. Relevant workforce capabilities may include adequate nurse staffing, clinical expertise, teamwork and supervisor support. Possible HR interventions include targeted recruitment, skill development, scheduling redesign and retention initiatives. Relevant outcomes could include vacancy duration, staffing stability, teamwork measures and patient-safety indicators.

The chain is a hypothesis, not a proven law. Each link should be tested with appropriate measures and time ordering. Analytics becomes strategically useful when it helps leaders choose among actions, allocate resources and learn whether the strategy is working.

5.2 A decision-theory perspective

A decision can be represented by alternatives, uncertain states of the world, consequences and preferences. Let a denote an action, s an uncertain state, and U(a,s) the utility or value of taking action a when state s occurs. If probabilities are available, expected utility can be expressed as:

Here, p(s) is the probability assigned to state s, with probabilities non-negative and summing to 1. The formula is useful when outcomes and probabilities can be meaningfully specified. In HR decisions, utility may include financial performance, service quality, fairness, legal exposure, employee wellbeing and reputation. These dimensions cannot always be reduced credibly to one monetary value.

A practical alternative is to compare options using a decision matrix. The matrix should identify benefits, costs, risks, affected groups, assumptions and reversibility. High-stakes and irreversible decisions require stronger evidence and more safeguards than low-risk, reversible experiments.

5.3 Causal reasoning

Managers often ask whether an HR intervention caused an outcome. Causal reasoning requires more than observing that two variables move together. A credible causal analysis considers:

• Temporal order: The intervention must precede the outcome.
• Confounding: Other variables may influence both intervention and outcome.
• Selection: Participants may differ systematically from non-participants.
• Comparison: A suitable counterfactual or comparison group is needed where feasible.
• Mechanism: There should be a plausible explanation for how the intervention produces change.
• Heterogeneity: Effects may differ across groups, roles or contexts.

Randomized experiments can provide strong causal evidence when ethically and practically feasible. Quasi-experimental methods, such as difference-in-differences, regression discontinuity and matched comparisons, may be useful when randomization is unavailable. Qualitative evidence remains valuable for understanding mechanisms and employee experience.

6. Measurement, Data Quality and Analytical Design

6.1 Start with a decision question

A well-formed analytics brief states:

1. What decision must be made?
2. Who will make it and by when?
3. Which population and time period are relevant?
4. What outcome indicates success?
5. What alternatives are available?
6. What constraints and ethical boundaries apply?
7. What evidence would change the decision?

For example, “Why is engagement low?” is too broad. “Which feasible changes to frontline scheduling are most likely to improve six-month retention without reducing service coverage?” is more decision-oriented.

6.2 Operational definitions

A construct is an abstract concept such as engagement, wellbeing, capability or inclusion. An operational definition specifies how it will be measured. Engagement might be represented by a validated survey scale, response behaviour, discretionary effort indicators or a combination. Each operationalization captures only part of the construct.

Analysts should document inclusion rules, missing-data treatment, calculation periods, data owners, transformations and known limitations. A metric without a definition is not a reliable management instrument.

6.3 Rates, averages and denominators

Rates are meaningful only when their denominators match the question. Absence days per employee, absence episodes per 100 employees and absence hours per scheduled hour answer different questions. An average can conceal unequal experiences; median pay, pay distribution and pay position may be more informative than mean pay alone.

Small groups create privacy and statistical problems. A dashboard showing a result for three employees may allow identification even if names are removed. Minimum reporting thresholds, aggregation and access controls are therefore part of analytical design.

6.4 Data quality dimensions

Important dimensions include accuracy, completeness, consistency, timeliness, uniqueness, validity and provenance. Data quality should be assessed in relation to use. A missing preferred name may be unimportant for payroll but important for employee experience analysis. An outdated job code may be critical for workforce planning.

7. Ethics, Governance and Regulation

HR analytics concerns people’s livelihoods, opportunities and dignity. Ethical governance should address at least the following principles:

• Purpose limitation: Collect and use data for specified legitimate purposes.
• Proportionality: Use no more data or intrusive method than necessary.
• Transparency: Explain what is collected, why it is used and how decisions are made.
• Privacy and security: Restrict access, protect identity and retain data appropriately.
• Fairness: Assess disparate impacts and avoid unjustified discrimination.
• Human oversight: Do not delegate high-impact employment decisions blindly to an algorithm.
• Contestability: Provide a meaningful route to question or appeal significant decisions.
• Accountability: Assign responsibility for data quality, model performance and consequences.

Legal requirements vary by jurisdiction and may include data-protection law, employment discrimination law, record-retention requirements, consultation duties and automated-decision safeguards. Compliance is a minimum, not a complete ethical standard.

Algorithmic systems can reproduce historical discrimination because past decisions may reflect unequal access, biased ratings or occupational segregation. Removing protected attributes does not necessarily remove bias: postcode, career gaps, educational institution or work pattern may act as proxies. Fairness assessment should examine selection rates, error rates, calibration and consequences across relevant groups, while recognizing that no single fairness measure is universally appropriate.

Surveillance is another concern. Collaboration metadata or sentiment analysis may be technically possible but inappropriate if employees cannot reasonably understand or challenge its use. The existence of data does not establish a legitimate right to use it.

8. Technology and Artificial Intelligence

Modern HR analytics commonly uses data warehouses, application programming interfaces, business-intelligence tools, statistical software and machine-learning platforms. Artificial intelligence may support résumé classification, conversational assistance, workforce forecasting, skills inference and text summarization.

The appropriate principle is “fit for purpose.” A transparent descriptive model may be preferable to a complex model when the decision is high stakes and explanation is essential. Machine learning can improve prediction, but it does not solve measurement error, causal ambiguity or ethical disagreement. Generative AI can accelerate drafting and exploration, but outputs require verification, confidentiality controls and human accountability.

A responsible analytical lifecycle includes:

1. Define the decision and governance boundaries.
2. Identify relevant data and lawful access conditions.
3. Assess quality, representativeness and potential bias.
4. Select a method proportionate to the question and risk.
5. Validate results technically and with domain experts.
6. Communicate uncertainty and limitations.
7. Pilot or implement with safeguards.
8. Monitor outcomes, subgroup effects and model drift.
9. Retire or revise the system when it no longer serves its purpose.

9. Illustrative Case: Reducing Early-Career Attrition

Background

The fictional organization Northstar Services employs 2,400 people across customer-support centres. Annual voluntary turnover is reported as 18%, but leaders are particularly concerned about employees leaving within their first twelve months. Replacement costs are increasing and service teams experience recurring shortages.

Problem

The executive team asks HR to “build a model to identify people who will leave.” This request is reframed as a decision problem: which interventions can improve early-career retention while respecting privacy and avoiding unfair treatment?

Analysis

The analytics team first defines early-career attrition as voluntary separation within twelve months of start date. It verifies that internal transfers are not incorrectly counted as exits. It examines cohorts by location, job family, shift, manager, pay position, training completion and schedule stability. The analysis shows that turnover is higher among employees with unstable schedules and limited early supervisor contact, but these variables are correlated with high-demand locations and difficult customer portfolios.

The team does not conclude that schedule instability causes all exits. It conducts interviews and finds that employees value predictable notice of shifts and timely coaching. It also discovers that the engagement survey response rate is lower among recent hires, limiting inference from survey scores.

Approach and solution

Rather than automatically flagging individuals for adverse action, Northstar pilots two team-level interventions: more predictable scheduling and a structured supervisor check-in during the first eight weeks. Teams are selected using operational criteria and matched, where feasible, with comparable teams. Outcomes include six- and twelve-month retention, service coverage, employee-reported schedule predictability and supervisor support.

The pilot specification is documented before implementation. HR monitors outcomes by location, role and demographic group. Employees receive information about the purpose of the pilot, and individual risk scores are not shared with managers.

Outcome

After two quarters, pilot teams show improved schedule-predictability scores and a lower early-exit rate than comparison teams, although the difference is smaller in the busiest centre. The organization therefore expands the scheduling intervention selectively and redesigns workload allocation in the busiest centre. It does not claim that the pilot proves universal causality because implementation differed across locations and the observation period was short.

Learning points

1. A prediction request may conceal a more useful intervention question.
2. Clear definitions are necessary before calculating turnover.
3. Correlation should guide investigation, not end it.
4. Team-level interventions can reduce privacy and stigmatization risks.
5. Evaluation should measure employee experience and operational outcomes together.
6. The practical value of analytics lies in improved decisions, not in model complexity.

Best Practices for HR Analytics Professionals

A professional standard for early analytics work includes the following:

• Begin with a decision and a theory of change.
• Define the population, unit, period and outcome precisely.
• Use the simplest method that can answer the question credibly.
• Combine quantitative evidence with qualitative and employee evidence.
• Separate prediction from causal explanation and recommendation.
• Report uncertainty, missing data and alternative interpretations.
• Test subgroup effects and practical consequences.
• Protect confidentiality and restrict access by role.
• Involve HR, legal, technology, managers and employee representatives where appropriate.
• Evaluate whether an intervention worked, for whom and at what cost.
• Maintain documentation so results can be reproduced and challenged.

Chapter Summary

HR analytics is the systematic use of workforce and organizational data to inform HR and business decisions. It differs from a metric or report because it seeks to interpret patterns, explain possible causes, anticipate outcomes or recommend actions. Its evolution has moved from administrative record keeping through operational reporting and descriptive analysis toward predictive and prescriptive decision support, although maturity is better understood as capability and governance than as technological sophistication.

Reporting asks what was recorded; descriptive analytics identifies patterns; diagnostic analytics investigates explanations; predictive analytics estimates what may occur; and prescriptive analytics compares possible actions. Each level has distinct uses and risks. Evidence-based HRM integrates scientific research, organizational data, professional expertise and stakeholder perspectives. It emphasizes validity, reliability, relevance, causal reasoning and contextual judgment.

Strategic analytics begins with organizational objectives and workforce capabilities rather than with available data. Decision theory clarifies alternatives, uncertainty, consequences and preferences, while causal reasoning prevents the common error of treating association as proof of intervention effects. Data quality, operational definitions and suitable denominators are essential. Because HR analytics affects rights and opportunities, privacy, fairness, transparency, accountability, human oversight and contestability must be designed into the analytical lifecycle. Technology and AI expand capability but do not replace judgment, ethical governance or responsibility for outcomes.

Key Terms

HR analytics: Systematic analysis of HR and workforce data to inform decisions.

People analytics: A broad approach to analysing data about people, work and organizational experience.

Workforce analytics: Analysis of workforce capacity, composition, skills, deployment and supply.

HR metric: A defined measure of an HR activity, condition or outcome.

Reporting: Structured presentation of recorded data.

Descriptive analytics: Analysis of historical patterns, distributions and trends.

Diagnostic analytics: Investigation of possible explanations for observed outcomes.

Predictive analytics: Estimation of future or unknown outcomes using data and models.

Prescriptive analytics: Analysis of actions and scenarios to support recommendations.

Evidence-based HRM: HR decision-making informed by the best available research, organizational evidence, expertise and stakeholder perspectives.

Unit of analysis: The entity being studied, such as an employee, team, month or organization.

Operational definition: A precise specification of how an abstract construct will be measured.

Validity: The extent to which evidence supports the intended interpretation or use of a measure.

Reliability: The consistency of a measurement process.

Confounding: A situation in which another variable influences both an apparent cause and an outcome.

Counterfactual: The outcome that would have occurred under an alternative action or no intervention.

Algorithmic bias: Systematic and unjustified differences in model outputs or consequences across groups.

Model drift: Decline in model performance when data, behaviour or circumstances change.

Data governance: Policies and responsibilities governing data quality, access, use, security and accountability.

Review Questions

Short Answer Questions

1. Distinguish an HR metric from HR analytics.
2. What is the central question addressed by diagnostic analytics?
3. Identify four sources of evidence in evidence-based HRM.
4. Why must the unit of analysis be specified?
5. State two ethical risks associated with HR analytics.

Long Answer Questions

1. Explain the evolution of HR analytics from administrative reporting to prescriptive and strategic analytics.
2. Compare reporting, descriptive, diagnostic, predictive and prescriptive analytics, including an example of each.
3. Discuss how evidence-based HRM balances scientific evidence, organizational data, professional expertise and stakeholder perspectives.
4. Explain why strategic alignment and causal reasoning are necessary for valuable HR analytics.
5. Evaluate the principal governance requirements for using AI in employment decisions.

Analytical / Application Questions

1. An organization reports a 25% turnover rate in one division and 10% in another. Design an analytical investigation that distinguishes descriptive, diagnostic and causal questions. Specify the data, comparisons and limitations you would examine.
2. A predictive model identifies employees at high risk of leaving. Propose a responsible decision process that uses the model without stigmatizing employees or allowing automated adverse action.
3. Design a small evaluation of a supervisor-coaching intervention intended to improve new-hire retention. State the outcome, comparison approach, possible confounders, ethical safeguards and criteria for organizational adoption.

Multiple Choice Questions with Answers

Question 1

Which statement best distinguishes HR analytics from an HR metric?

A. A metric is always predictive, whereas analytics is always descriptive.

B. A metric is a defined measure, whereas analytics systematically examines data to inform decisions.

C. A metric applies only to finance, whereas analytics applies only to employees.

D. A metric is qualitative, whereas analytics is exclusively quantitative.

Correct Answer: B. A metric is a defined measure, whereas analytics systematically examines data to inform decisions.

Question 2

Which question is most directly associated with diagnostic analytics?

A. What was the headcount last month?

B. What is the projected vacancy rate next year?

C. Why are voluntary exits concentrated among recent hires?

D. Which staffing schedule maximizes coverage?

Correct Answer: C. Why are voluntary exits concentrated among recent hires?

Question 3

An organization starts with 480 employees, ends with 520, and records 30 separations during a quarter. Using the average of opening and closing headcount, what is the turnover rate?

A. 5%

B. 6%

C. 6.25%

D. 12.5%

Correct Answer: B. 6%

Question 4

Which is a core principle of evidence-based HRM?

A. Organizational data should always override employee perspectives.

B. Managerial intuition should replace research when decisions are complex.

C. Decisions should consider the best available evidence, expertise, context and stakeholder interests.

D. Only statistically significant findings are legitimate evidence.

Correct Answer: C. Decisions should consider the best available evidence, expertise, context and stakeholder interests.

Question 5

Why does correlation between training participation and lower turnover not by itself prove that training caused the reduction?

A. Training data can never be measured reliably.

B. Participants may differ from non-participants in other relevant ways.

C. Turnover cannot be studied with quantitative methods.

D. Causality applies only to financial outcomes.

Correct Answer: B. Participants may differ from non-participants in other relevant ways.

Question 6

Which sequence most appropriately links HR analytics to strategy?

A. Available data → dashboard → algorithm → strategy

B. Strategic objective → workforce capability → HR intervention → outcome → organizational result

C. Employee survey → legal review → payroll → strategy

D. Technology purchase → data collection → headcount → mission

Correct Answer: B. Strategic objective → workforce capability → HR intervention → outcome → organizational result

Question 7

What does a probability of 0.70 from a predictive model mean under valid model conditions?

A. The individual will definitely experience the outcome.

B. The model estimates a 70% likelihood of the outcome, subject to its data and validation conditions.

C. The model has 70% accuracy in every subgroup.

D. The outcome has already occurred in 70% of cases.

Correct Answer: B. The model estimates a 70% likelihood of the outcome, subject to its data and validation conditions.

Question 8

Which practice best supports responsible HR analytics governance?

A. Use all available data because more data always improves fairness.

B. Remove human review from high-impact decisions to prevent inconsistency.

C. Provide transparency, access controls, monitoring and a route to challenge significant decisions.

D. Hide model limitations to improve managerial confidence.

Correct Answer: C. Provide transparency, access controls, monitoring and a route to challenge significant decisions.

Question 9

What is the ecological fallacy?

A. Assuming that a relationship observed across groups necessarily applies to individuals.

B. Using environmental data in workforce planning.

C. Treating missing data as evidence of employee misconduct.

D. Assuming that individual data are always more accurate than group data.

Correct Answer: A. Assuming that a relationship observed across groups necessarily applies to individuals.

Question 10

Why might a simple transparent model be preferred to a complex machine-learning model in a high-stakes employment decision?

A. Simple models never contain bias.

B. Complex models cannot make predictions.

C. Transparency and explainability may be more important for accountability and contestability.

D. High-stakes decisions do not require validation.

Correct Answer: C. Transparency and explainability may be more important for accountability and contestability.