AI Prompts for HR Analytics and People Metrics

HR analytics is not a dashboard of headcount and attrition. It is the discipline of answering the questions leaders actually have — is the new manager program working, why is time to hire rising in one function, are we losing the people we want to keep — with data that is defined consistently, analyzed honestly, and reported in a way that protects individuals. Most HR data is messy; the first job is usually definitions and quality.

These prompts define the metrics, structure an analysis around a question, and design the reporting. The guardrail is the same as elsewhere in this cluster: aggregate, above threshold, for policy and support, never for scoring or predicting individuals.

Before you use these

Have these ready to replace the highlighted [variables]:

• The questions leaders ask HR, and the decisions behind them
• Current metrics, definitions and data sources (HRIS, ATS, survey, LMS, payroll)
• Data quality issues you know about
• Anonymity thresholds and access rules

The prompts

• 1. Define the people metrics and their data sources
• 2. Structure an analysis around a leader's question
• 3. Design the HR dashboard for decisions and privacy
• 4. Reconcile headcount across systems

1. Define the people metrics and their data sources

Act as a people analytics lead defining the people metrics for [company].

Decisions to support: [e.g. hiring investment by function; manager capability programs; retention actions; workforce cost planning; DEI monitoring where lawful]
Current metrics: [list with definitions as understood]
Systems: [HRIS, ATS, survey, LMS, payroll — and their known data problems]

1. Metric dictionary: for each metric that serves a decision — definition, formula, inclusions and exclusions (e.g. contractors, interns, transfers), source system and field, frequency, owner, and known quality issues. Cover at least: headcount and FTE, hires and time to fill/hire, attrition (total, voluntary, regretted, first-year), internal mobility, span of control, engagement outcome items, learning participation and application, workforce cost.
2. Definitions that must be reconciled across systems (e.g. 'start date', 'leaver date', 'role family').
3. Metrics to retire: those with no decision behind them or no reliable source.
4. Data-quality fixes in priority order, with the metric each unlocks.
5. Privacy and access: which metrics are reported at which aggregation, the anonymity threshold, who sees what — to verify with policy and law, especially for any demographic monitoring.
6. A one-page glossary for managers.

Define before you measure. A metric without an agreed definition produces arguments, not decisions.

2. Structure an analysis around a leader's question

You are structuring a people analytics investigation into: [the leader's question, as asked].

Data available: [systems, fields, periods, group sizes]
Context: [what changed, what the leader suspects, the decision they will make]
Anonymity threshold: [n]

1. Reframe: the analyzable question(s) behind the leader's question, and the decision the answer will inform.
2. Hypotheses: three or four candidate explanations, and the evidence that would distinguish them.
3. Data and method: the metrics, the comparison (before/after, group vs group, cohort), the controls for obvious confounders (role family, tenure, location), the period; where the method is a comparison of associations, say so.
4. Limits: sample sizes, data quality, what cannot be concluded (causation, individual-level claims); groups that cannot be reported.
5. Answer format: the one-page structure — the finding, the evidence, the confidence, the alternative explanations, the recommended decision and its risk.
6. Follow-up: the data to collect if the answer is inconclusive.

Answer the question at group level. If the honest answer is 'the data cannot tell us', say so and say what would.

3. Design the HR dashboard for decisions and privacy

Act as a people analytics lead designing the HR dashboard for [company].

Metric dictionary: [paste or summarize]
Audiences: [executives, HR business partners, function leaders, managers — with the decisions each makes]
Privacy rules: [anonymity threshold; access by role; demographic data rules; retention — to verify]

1. Views by audience: for each, the decisions served, the metrics shown, the dimensions available (function, location, level, tenure band), the time comparisons, and the aggregation floor.
2. Suppression and drill-down: cells below the threshold suppressed; drill-down stops above the level at which individuals could be identified; no individual-level scoring or 'risk' lists; how small teams are handled.
3. Alerts: the conditions that prompt a conversation (regretted attrition in a group above threshold; time to fill rising; engagement item falling), phrased as prompts to investigate, not conclusions.
4. Data freshness and cadence by view; the definitions link on every page.
5. Governance: owner, change control for definitions, access reviews, the audit of who accessed what.
6. Manager view specifically: what a manager sees about their own team (aggregate above threshold) and what they do not, and why.

Design for decisions and for trust. A dashboard that lets a manager see who answered a survey how has ended the survey.

4. Reconcile headcount across systems

Act as a people analytics lead reconciling headcount across systems for [company] as at [date].

Figures: [HRIS: n, definition; payroll: n, definition; finance: n, definition; other]
Known differences: [contractors, interns, employees on leave, pending starters and leavers, transfers between entities, date conventions, part-time as heads vs FTE]

1. Bridge: starting from one system, the adjustments that reconcile to each other system, category by category, with the count for each; unexplained residual stated.
2. Definitional differences: for each category, which system is 'right' for which purpose (cost, capacity, legal headcount).
3. Canonical definition: propose one headcount and one FTE definition for reporting, with inclusions and exclusions, and the owner.
4. Fixes: the process or data changes that would prevent recurring differences (start/leave date conventions, contractor tagging, transfer handling).
5. Reporting: how each purpose's figure is derived from the canonical one, so numbers reconcile by construction.
6. Verification: the monthly check that keeps the bridge small.

Show the bridge as a table. A headcount nobody agrees on undermines every other people metric.

Related prompts

• Employee Retention Analysis
• Workforce Planning
• Engagement Survey Analysis
• Exit Interview Analysis
• HR Policy Drafting
• Talent Reviews
• Operational KPI reviews — Operations & Supply Chain
• CRM data analysis — Sales
• Data Analysis prompts

Logical next step

After this, most HR teams move on to Employee Retention Analysis.

All Human Resources prompts · Search the full library