People analytics — serving decisions, not surveillance
People analytics serves diagnosis, prediction, evaluation, equity, and investment guidance. It doesn’t serve surveillance. The line between insight and creep is what privacy and fairness disciplines protect.
What to measure
Workforce composition. Lifecycle metrics (time-to-hire, 90-day attrition, quality-of-hire, internal-fill rate). Engagement and wellbeing (privacy-respecting). Performance distribution. Equity (pay, promotion, recognition, PM application, development access).
Aggregate signal is what’s actionable; individual data is rarely insight.
What not to measure
Continuous individual monitoring (keystrokes, facial expression analysis, sentiment without consent). AI individual flight-risk scores acted on without agent awareness. Productivity metrics not tied to outcome. Individual engagement scoring (undermines candour). Personal-life inference from work patterns.
These tend to backfire — agents game, trust collapses, the data becomes unreliable, regulators engage.
Equity analytics
Pay equity by demographic. Promotion rates by demographic. Recognition (who gets recognised, who doesn’t). Performance management application by demographic. Development access.
Equity analytics often surfaces uncomfortable findings. The disciplined leader engages rather than rationalises.
Privacy and consent
Minimise data collection to what the work justifies. Anonymise where possible. Consent transparently. Restrict access. Audit periodically. Provide recourse. Align with regulation (GDPR and equivalents).
Not legal advice — engage DPO and HR specialists.
The closing principle
People analytics serves human judgement; it doesn’t replace it. Measure aggregate, respect privacy, audit for equity, refuse surveillance. The line between insight and creep is what the disciplines protect.
See also
- Attrition reduction diagnose by segment, act on cause
- Engagement vs satisfaction they aren t the same thing