HR Metrics, Benchmarking, Dashboards and Data Storytelling

Introduction

People analytics becomes valuable when organizational questions are translated into credible measures and then connected to decisions. A human resources department may possess extensive data about recruitment, absence, performance, learning, compensation, mobility and turnover. Data alone, however, does not establish what is happening, why it is happening, or what managers should do next. Metrics provide disciplined representations of workforce conditions; benchmarking supplies context; dashboards support monitoring; and data storytelling turns analysis into shared understanding and action.

This chapter treats HR measurement as a decision system rather than a reporting exercise. A useful metric has a clear definition, an appropriate denominator, a known population and period, reliable data, and a decision purpose. It should also be interpreted cautiously. A rise in voluntary turnover, for example, may signal a retention problem, improved reporting, workforce restructuring, or a change in the composition of employees. The number is a starting point for inquiry, not a complete explanation.

Learning Objectives

By the end of this chapter, you should be able to:

• define HR metrics, KPIs, ratios and funnels;
• distinguish activity, efficiency, effectiveness and outcome measures;
• calculate and interpret common recruitment, workforce and retention metrics;
• design internal and external benchmarking exercises;
• construct dashboards that are accessible, focused and decision-oriented;
• communicate trends, uncertainty, limitations and recommended actions through data storytelling; and
• identify ethical, governance and validity risks in people measurement.

1. From HR Data to Decision Measures

An HR metric is a quantified representation of a workforce phenomenon or HR process. A key performance indicator (KPI) is a metric judged especially important for a defined objective. Thus, every KPI is a metric, but not every metric deserves KPI status. A recruitment team may track dozens of measures, while its quarterly KPIs might be time to fill critical roles, quality of hire and candidate acceptance rate.

A strong measurement system begins with a decision question:

What decision will this measure inform, who will use it, and what action could follow?

The question prevents the common error of collecting convenient data without a use case. It also clarifies the level of measurement. A board may need workforce risk and succession indicators; a recruitment manager may need stage conversion rates; a hiring manager may need candidate response time and interview quality.

1.1 Constructs, operational definitions and populations

Many important HR concepts are constructs rather than directly observable objects. Engagement, inclusion, leadership quality and employability require operational definitions. For example, “retention” might mean employees remaining employed after 12 months, whereas “regrettable attrition” might mean the exit of employees in specified critical roles with strong performance ratings. These are not interchangeable.

Every metric definition should specify:

1. Numerator: what is counted.
2. Denominator: the population or exposure against which it is compared.
3. Time period: the start and end dates.
4. Unit: people, cases, days, currency, percentage or score.
5. Inclusion and exclusion rules: for example, whether contractors, interns and employees on leave are included.
6. Data source and refresh schedule.
7. Owner and intended use.

A rate without a denominator can be misleading. “Twenty resignations” has a different meaning in a 200-person organization than in a 20,000-person organization. Similarly, average absence days should not be compared across groups with different employment fractions unless the calculation accounts for contracted working time.

1.2 Descriptive, diagnostic, predictive and prescriptive uses

HR analytics is often described through four analytical purposes:

• Descriptive analytics summarizes what happened, such as monthly turnover.
• Diagnostic analytics investigates why it happened, such as comparing turnover by manager, tenure and shift.
• Predictive analytics estimates what may happen, such as the probability of an employee leaving within six months.
• Prescriptive analytics evaluates possible actions, such as alternative staffing or retention interventions.

The distinction matters because a descriptive metric should not be presented as proof of causation. A department with high turnover is not necessarily causing turnover; it may employ workers in a scarce occupation or operate in a difficult labor market. Causal interpretation requires stronger research designs, controls, and attention to confounding variables.

2. Metric Families and the Performance Chain

A practical classification separates measures by their relationship to work and results.

These categories form a performance chain rather than a hierarchy in which outcomes always replace other measures. Activity measures are useful for operational control; efficiency measures expose resource use; effectiveness measures assess process success; and outcome measures connect HR work to organizational consequences.

A balanced recruitment measurement set might include applications screened (activity), recruiter hours per hire (efficiency), qualified-candidate conversion (effectiveness), and first-year retention or performance (outcome). Monitoring only activity can reward busyness. Monitoring only outcomes can delay detection of a failing process.

2.1 Leading and lagging indicators

A leading indicator changes before the result of interest and may provide an opportunity to intervene. Examples include interview scheduling delays, employee sentiment or the proportion of critical roles without successors. A lagging indicator records an outcome after it has occurred, such as regrettable turnover or revenue per employee.

Leading indicators are not automatically causal or accurate predictors. Their value should be tested. A survey item may correlate with later turnover in one population but not another, particularly when response rates are low or organizational conditions change.

2.2 Ratios, rates and averages

A ratio compares two quantities, while a rate usually describes events relative to a population or exposure over time. The interpretation depends on whether numerator and denominator are logically related.

For an average headcount over a period with beginning and ending headcount, a simple approximation is:

This approximation assumes that workforce changes are reasonably distributed or that a more detailed headcount series is unavailable. If headcount fluctuates substantially, use a daily or monthly average:

where headcount_t is headcount in period t and n is the number of observations. The observation interval must be consistent.

A turnover rate commonly uses separations divided by average headcount:

This measure assumes that separations and headcount are defined consistently. It is not a universal formula: organizations may use average employees, opening headcount, or annualized monthly rates. The definition must therefore accompany the value.

Worked example. A business has 480 employees at the start of a quarter and 520 at the end. During the quarter, 18 employees leave. Average headcount is (480 + 520) ÷ 2 = 500. Quarterly turnover is (18 ÷ 500) × 100 = 3.6%. The result should not automatically be multiplied by four to claim annual turnover, because annualization assumes comparable rates and may exaggerate or understate the actual annual value.

A voluntary turnover rate restricts the numerator to voluntary separations. A regrettable turnover rate requires a documented organizational definition, such as voluntary exits from critical roles or exits of high-performing employees. “Regrettable” is a managerial classification and should not be treated as an objective property of an individual.

3. Recruitment Metrics and Funnels

Recruitment is a process with stages, making it suitable for funnel analysis. A funnel displays the number of candidates moving from one stage to the next and the conversion rate between stages.

For a stage-to-stage conversion:

The denominator must be the number entering that stage, not necessarily the original applicant count. A recruitment funnel may include applicants, screened applicants, interviewed candidates, offers and accepted offers. If 800 applications produce 160 screens, 48 interviews, 12 offers and 9 acceptances, the application-to-acceptance yield is (9 ÷ 800) × 100 = 1.125%, while the offer acceptance rate is (9 ÷ 12) × 100 = 75%.

Funnels help locate process loss, but they do not explain it. Low screening conversion may reflect a targeted vacancy announcement, poor sourcing, an overly restrictive specification, or an unsuitable applicant pool. Funnel comparisons should control for role family, labor market, recruitment source and selection standards.

Common recruitment metrics include:

• Time to fill: elapsed time from an agreed vacancy opening date to accepted offer or filled position. The organization must define the start and end events.
• Time to hire: often measured from candidate application or entry into the process to acceptance; this is not identical to time to fill.
• Cost per hire:

Total costs may include advertising, agency fees, assessment, technology, travel and allocated recruiter labor. Inclusion rules are essential.

• Offer acceptance rate: accepted offers divided by offers made.
• Source yield: hires attributed to a source divided by applicants or qualified candidates from that source.
• Quality of hire: a composite measure that may combine validated performance, retention and hiring-manager assessment after a defined period. Because composite scores depend on weighting and scale, the construction should be documented.

Quality of hire is especially vulnerable to attribution and measurement bias. A new employee’s performance reflects onboarding, manager support, job design and labor conditions, not only recruitment quality.

4. Workforce, Absence and Retention Measures

4.1 Headcount and full-time equivalent

Headcount counts people; full-time equivalent (FTE) represents contracted work capacity relative to a full-time standard. If two employees each work half-time, they count as two people but equal one FTE. Headcount is appropriate for questions about individuals and legal populations; FTE is often more suitable for labor capacity and cost analysis.

4.2 Absence

An absence rate can be defined as:

Available workdays must reflect the same employee population and period. The organization should specify whether vacation, parental leave, occupational injury and unauthorized absence are included. A frequency measure, such as absence episodes per 100 employees, answers a different question from a severity measure, such as lost days per absence episode.

4.3 Internal mobility and promotion

Internal fill rate measures the proportion of vacancies filled by existing employees:

It can indicate career mobility and reduced external hiring costs, but a high value may also signal limited external perspectives or restricted access to opportunities. Promotion rates should be disaggregated by relevant groups and adjusted for eligibility and job structure; raw comparisons can be misleading when groups have different tenure or occupational distributions.

4.4 Engagement and survey measures

Survey indicators require attention to response rate, question wording, scale construction and nonresponse bias. A mean score can hide polarized responses, and a high response rate does not guarantee representative responses. Report distributions, confidence intervals where appropriate, and subgroup sizes. Small groups create both statistical instability and confidentiality risks.

5. Benchmarking: Context Without False Comparability

Benchmarking is the systematic comparison of a measure with a reference point. It may be internal, external, historical, functional or normative.

• Internal benchmarking compares teams, locations, periods or job families within the organization.
• External benchmarking compares an organization with peer organizations or published labor-market data.
• Historical benchmarking compares current performance with its own past.
• Functional benchmarking compares similar processes across different industries.
• Normative benchmarking compares performance with a target or standard, such as a service-level agreement.

Internal benchmarking is often more actionable because definitions and systems are more likely to be consistent. However, internal comparisons can be unfair when groups face different labor markets, work designs or risk exposures. External benchmarking supplies context but introduces differences in definitions, industry, geography, organization size, workforce composition and data collection.

A sound benchmarking procedure is:

1. Define the decision and metric precisely.
2. Confirm numerator, denominator, period and population equivalence.
3. Assess data quality and missingness.
4. Select a relevant comparison group rather than an attractive one.
5. Compare distributions or ranges, not only a single average.
6. Investigate plausible drivers of gaps.
7. Decide whether the gap is material and actionable.
8. Record limitations and repeat the comparison consistently.

Percentile comparisons require care. If an organization is at the 75th percentile for time to fill, this generally means its value is greater than or equal to approximately 75% of observations in the reference distribution—not that it has achieved 75% of a target. For measures where lower is better, a high percentile may indicate poor performance; the direction must be stated.

A benchmark is not automatically a goal. Copying a competitor’s turnover rate without understanding job mix, pay structure and labor market conditions can produce harmful decisions. Benchmarking should stimulate explanation, not substitute for it.

6. Dashboard Design for HR Decisions

A dashboard is a visual interface that presents selected measures for monitoring, exploration or decision support. It is not simply a collection of charts. Its design should follow the information needs of a defined audience.

6.1 Dashboard layers

A useful HR dashboard often has three layers:

1. Executive layer: a small set of strategic KPIs, trends, thresholds and risks.
2. Managerial layer: comparisons by department, role, location, tenure or manager, with filters and drill-down.
3. Analytical layer: detailed records, definitions, statistical tests and diagnostic views for analysts.

The layers should use consistent definitions. An executive turnover value should reconcile with the value in the analyst view.

6.2 Visual encoding and accessibility

Select a chart according to the analytical task. Lines show trends; bars compare categories; stacked bars show composition when categories are limited; scatterplots examine relationships; and tables are preferable when exact values matter. Avoid three-dimensional charts, excessive decoration and dual axes that imply relationships not supported by the data.

Accessibility requires more than color. Use direct labels, adequate contrast, meaningful alternative text, readable font sizes and patterns or symbols alongside color. Red–green distinctions can exclude users with color-vision deficiency. Dates, currencies, percentages and time zones should be formatted consistently.

A dashboard should display:

• metric name and plain-language definition;
• period and refresh date;
• numerator, denominator or population where relevant;
• comparison baseline or target;
• filters and their current state;
• notes on missing data and exclusions;
• responsible owner; and
• a link or reference to methodology.

6.3 Targets, thresholds and uncertainty

Targets are normative; forecasts are estimates; confidence intervals express sampling uncertainty under specified assumptions. These must not be conflated. For a proportion p estimated from n independent observations, a simple approximate standard error is:

This approximation is most defensible when observations are independent and the sample is not extremely small or near 0 or 1. For small samples, clustered data or sensitive proportions, analysts should use more appropriate methods, such as exact or Wilson intervals and cluster-robust approaches where justified.

For example, if 75 of 300 employees respond positively, p = 75 ÷ 300 = 0.25. The approximate standard error is √(0.25 × 0.75 ÷ 300) ≈ 0.025. A rough 95% interval using p ± 1.96 × SE is approximately 0.201 to 0.299. This interval describes sampling uncertainty under the stated assumptions; it does not correct nonresponse bias or measurement error.

Dashboards should avoid false precision. Displaying 25.0% may imply more certainty than 25%, particularly for small groups. Suppression rules are also necessary when subgroup counts could reveal individuals or produce unstable estimates.

7. Data Storytelling: From Finding to Action

Data storytelling combines evidence, explanation and a decision. A useful narrative commonly follows this structure:

1. Context: What decision or organizational problem matters?
2. Finding: What reliable pattern is visible?
3. Meaning: What are the plausible explanations and consequences?
4. Uncertainty: What is not known, and how strong is the evidence?
5. Action: What should be done, by whom and by when?
6. Follow-up: Which measures will indicate whether the action worked?

The narrative should distinguish observation from interpretation. “Turnover increased from 8% to 12%” is an observation. “Compensation caused the increase” is a causal claim requiring evidence. A more defensible statement might be: “Turnover rose most sharply among employees with less than two years’ tenure, coinciding with a change in shift patterns; exit-survey and pay data should be examined before selecting an intervention.”

Effective stories use a meaningful comparison, such as prior period, target, peer group or expected range. They avoid cherry-picking, truncated axes that exaggerate differences, and emotionally loaded labels. They also make the next decision explicit. A report that ends with “more analysis is needed” without specifying the analysis or decision is incomplete.

A concise executive narrative might say:

Voluntary turnover among first-year warehouse employees increased from 10% to 17% over two quarters, while other tenure groups remained stable. The increase is concentrated on night shifts, but the group is small and the data do not establish causation. HR and operations should review schedule stability, supervisor assignment and local pay competitiveness within 30 days. The pilot intervention will be evaluated using six-month retention, absence and employee feedback.

This statement integrates trend, segmentation, limitation, action and evaluation.

8. Illustrative Case: Northstar Services’ Recruitment Dashboard

Background

Northstar Services, a fictional customer-support organization, has 1,200 employees across three locations. Managers report that recruitment is slow and expensive, while the executive team is concerned about early-tenure turnover. The HR information system contains applicant-stage dates, recruitment costs, employee records and six-month retention data, but definitions differ between locations.

Problem

The existing monthly report lists applications, interviews, hires and total recruiting expenditure. It does not distinguish roles, source, location or time-to-event definitions. Managers therefore disagree about whether the central issue is insufficient applicant volume, slow selection, weak offers or poor onboarding.

Analysis

The analytics team first creates a metric dictionary. It defines time to fill as calendar days from approved requisition date to accepted offer, cost per hire as recruiting expenditure divided by completed hires, and six-month retention as the proportion of hires still employed 180 days after start date. Contractors and internal transfers are excluded from the external recruitment funnel.

In one quarter, 1,000 external applications produce 220 screens, 66 interviews, 20 offers and 15 acceptances. Stage conversion rates are:

• screen-to-application: not a stage conversion; the appropriate application-to-screen rate is 220 ÷ 1,000 = 22%;
• interview-to-screen: 66 ÷ 220 = 30%;
• offer-to-interview: 20 ÷ 66 ≈ 30.3%; and
• acceptance-to-offer: 15 ÷ 20 = 75%.

The team finds that the night-shift source has a higher applicant-to-screen rate but lower six-month retention. It also finds that Location C records requisition approval several days earlier than the other locations, making its time-to-fill value appear longer even when candidate-facing processing time is similar. This demonstrates why metric standardization must precede benchmarking.

Approach and solution

Northstar redesigns the dashboard around three decisions: where to improve funnel conversion, where to reduce process delay, and whether recruitment changes improve early retention. The first page displays hires, median time to fill, cost per hire, offer acceptance and six-month retention. Median time is used because a small number of very long vacancies distort the mean; both statistics remain available for analysts.

The second page displays funnel counts and stage-to-stage conversion by role and source. The third page shows a cohort view of retention by start month, shift and location. A footnote identifies small groups and excludes them from public managerial views when confidentiality or reliability thresholds are not met.

The organization pilots a realistic job preview and a structured night-shift scheduling discussion. It does not claim that these interventions caused improvement solely because retention rises in the next month. Instead, it defines a six-month evaluation period and compares the pilot cohort with a suitable contemporaneous comparison group, while monitoring role mix and hiring volume.

Outcome

After the pilot, offer acceptance increases from 75% to 82%, and six-month retention improves from 68% to 74% in the pilot cohort. The evidence is encouraging but not conclusive because the pilot was not randomly assigned and seasonal hiring patterns changed. Northstar therefore continues the intervention while conducting a stronger evaluation and reviewing employee experience data.

Learning points

The case illustrates five principles. First, metric definitions must be harmonized before comparisons. Second, a funnel identifies locations of loss but not causes. Third, outcome measures should be linked to the original process rather than assumed. Fourth, dashboards should support distinct managerial decisions. Fifth, responsible storytelling includes uncertainty and avoids overstating a pilot’s effect.

9. Governance, Ethics and Responsible Use

People data concerns identifiable individuals and can affect employment opportunities. Governance should address lawful basis, purpose limitation, data minimization, access control, retention periods, security and auditability. Applicable requirements vary by jurisdiction, but privacy principles and employment law should be integrated into metric design rather than added after deployment.

Important risks include:

• Proxy discrimination: location, school, shift or career gaps may indirectly encode protected characteristics.
• Selection bias: historical hiring data reflect previous decisions and may reproduce them.
• Small-cell disclosure: a dashboard can reveal identity through rare combinations of attributes.
• Function creep: data collected for payroll or administration may be repurposed for surveillance without adequate justification.
• Metric gaming: employees may optimize a measured target while undermining the underlying objective.
• Automation opacity: a predictive score may be treated as a fact rather than an uncertain estimate.

Use role-based access, documented definitions, data-quality checks, impact assessments and human review for consequential decisions. Predictive or AI-supported tools should be validated for performance and disparate impact, monitored after implementation, and accompanied by an appeal or correction process. A model should not be used merely because it is technically accurate if its use is unfair, unlawful or inconsistent with employee expectations.

10. Implementation Guidelines and Contemporary Practice

Organizations usually improve measurement by taking a staged approach:

1. Map strategic priorities to decisions and stakeholders.
2. Build a metric dictionary and data lineage record.
3. Audit data completeness, consistency, timeliness and accuracy.
4. Establish baseline values before introducing targets.
5. Create a small prototype dashboard with representative users.
6. Test definitions, accessibility and interpretation.
7. Introduce governance, ownership and review routines.
8. Evaluate whether dashboard use changes decisions or outcomes.

Contemporary people analytics increasingly combines HRIS, applicant-tracking, learning, scheduling and employee-listening data. Integration can improve context but also increases privacy and linkage risks. Natural-language interfaces and generative AI may help users query dashboards or draft narrative summaries, but generated explanations require verification. AI should not invent causal interpretations, conceal uncertainty or expose sensitive records through prompts. Human analysts remain responsible for definitions, validation and communication.

The most mature organizations treat metrics as hypotheses embedded in a learning cycle: measure, interpret, act, evaluate and revise. They retire measures that no longer support decisions and resist the temptation to equate more metrics with better management.

Chapter Summary

HR metrics convert workforce phenomena and processes into defined measures for decisions. Reliable measurement requires explicit numerators, denominators, populations, periods, units, inclusion rules and ownership. Activity, efficiency, effectiveness and outcome measures serve different purposes and should be combined rather than confused. Recruitment funnels reveal stage conversion but do not by themselves explain process failure. Turnover, absence, mobility and recruitment formulas are meaningful only when definitions and populations are consistent.

Internal and external benchmarking provide context, but comparisons must account for workforce composition, labor markets, definitions and data quality. Dashboards should be audience-specific, accessible, transparent about methods and designed around decisions. Data storytelling connects context, finding, interpretation, uncertainty, action and follow-up. Ethical practice requires privacy protection, fairness assessment, governance, human oversight and resistance to metric gaming. Ultimately, people analytics creates value when measurement improves the quality of organizational decisions rather than merely increasing the volume of reporting.

Key Terms

Metric: A quantified representation of a defined workforce phenomenon or process.

Key performance indicator (KPI): A metric selected as especially important for a specified objective.

Numerator: The quantity counted in the upper part of a ratio or rate.

Denominator: The reference population or exposure used in a ratio or rate.

Activity measure: A measure of work performed or actions completed.

Efficiency measure: A measure of resources or time used relative to output.

Effectiveness measure: A measure of whether an immediate process objective was achieved.

Outcome measure: A measure of a subsequent organizational or workforce result.

Leading indicator: A measure that may change before a result of interest.

Lagging indicator: A measure of a result after it has occurred.

Funnel: A sequence of process stages showing counts and conversion rates.

Conversion rate: The proportion advancing from one defined stage to the next.

Turnover rate: Separations expressed relative to a defined average workforce population.

Benchmarking: Systematic comparison with an internal, external, historical or normative reference.

Dashboard: A visual interface presenting selected measures for monitoring or decision support.

Data storytelling: Communication that combines evidence, context, interpretation, uncertainty and action.

Data lineage: Documentation of a metric’s sources, transformations and owners.

Small-cell disclosure: Risk of identifying individuals from very small or distinctive groups.

Proxy discrimination: Unfair disadvantage arising when a variable indirectly represents a protected characteristic.

Metric gaming: Behavior that improves a measured value while weakening the underlying objective.

Review Questions

Short Answer Questions

1. Why is a metric definition incomplete if it specifies only a numerator?
2. Distinguish an activity measure from an outcome measure and give one HR example of each.
3. What is the difference between internal and external benchmarking?
4. Why should a recruitment funnel use stage-specific denominators?
5. Identify two accessibility requirements for an HR dashboard.

Long Answer Questions

1. Explain the performance chain from activity to outcome and discuss why an organization should monitor more than one type of measure.
2. Describe a rigorous process for conducting an external benchmarking exercise, including threats to comparability.
3. Explain how a dashboard can communicate uncertainty without becoming unusable for managers.
4. Discuss the distinction between observation, interpretation and causal claim in HR data storytelling.
5. Evaluate the ethical and governance risks of using predictive employee turnover scores.

Analytical / Application Questions

1. An organization begins a month with 740 employees, ends with 780, and records 25 separations. Calculate approximate average headcount and the monthly turnover rate. State one limitation of the calculation.
2. A recruitment funnel has 1,500 applicants, 300 screened candidates, 90 interviews, 30 offers and 21 acceptances. Calculate the application-to-screen, screen-to-interview, interview-to-offer and offer-acceptance rates. Identify the stage that deserves further investigation and explain why conversion alone does not establish the cause.
3. A dashboard shows that Department A has 14% turnover and Department B has 8%. Department A employs mostly new, part-time workers in a highly competitive labor market, while Department B employs longer-tenured professionals. Design a fairer comparison and propose a data-storytelling narrative that communicates both the observed gap and its uncertainty.

Multiple Choice Questions with Answers

Question 1

Which measure is best classified as an outcome measure?

A. Number of interviews completed

B. Recruiter hours per hire

C. Offer acceptance rate

D. Six-month retention of new hires

Correct Answer: D. Six-month retention of new hires

Question 2

A team receives 800 applications and makes 12 hires. What is the application-to-hire yield?

A. 0.67%

B. 1.5%

C. 6.7%

D. 66.7%

Correct Answer: B. 1.5%

Question 3

Which denominator is most appropriate for a conventional turnover rate?

A. Number of job advertisements

B. Average headcount during the period

C. Number of interviews

D. Ending payroll cost

Correct Answer: B. Average headcount during the period

Question 4

If 9 of 12 offers are accepted, the offer acceptance rate is:

A. 25%

B. 33.3%

C. 75%

D. 133.3%

Correct Answer: C. 75%

Question 5

What is a primary advantage of internal benchmarking?

A. It guarantees causal conclusions

B. It usually uses more comparable organizational definitions

C. It eliminates workforce differences

D. It always provides labor-market context

Correct Answer: B. It usually uses more comparable organizational definitions

Question 6

Which chart is generally most appropriate for showing monthly turnover over two years?

A. Line chart

B. Pie chart

C. Three-dimensional area chart

D. Radar chart

Correct Answer: A. Line chart

Question 7

Which statement is a causal claim rather than a descriptive observation?

A. Turnover increased from 8% to 12%

B. Turnover was highest among first-year employees

C. Night-shift turnover was 5 percentage points higher

D. The schedule change caused the increase in turnover

Correct Answer: D. The schedule change caused the increase in turnover

Question 8

For a proportion estimated from independent observations, which expression gives the simple approximate standard error?

A. p ÷ n

B. √(p(1 − p) ÷ n)

C. n ÷ p

D. p(1 + p) × n

Correct Answer: B. √(p(1 − p) ÷ n)

Question 9

Which practice best reduces small-cell disclosure risk?

A. Publishing every subgroup regardless of size

B. Removing all metric definitions

C. Suppressing or aggregating very small groups

D. Replacing percentages with more decimal places

Correct Answer: C. Suppressing or aggregating very small groups

Question 10

What is the main purpose of a recruitment funnel?

A. To prove why every candidate rejected an offer

B. To show movement and conversion across recruitment stages

C. To replace all outcome measures

D. To guarantee that a source is unbiased

Correct Answer: B. To show movement and conversion across recruitment stages