People analytics has graduated from HR dashboards into a genuinely consequential decision-support discipline. Workforce planning, retention modelling, compensation equity analysis, performance prediction, location-and-presence reporting — the technical capability exists to do all of these at scale. Whether your organisation should is a separate question, and one that determines whether your analytics function builds trust or quietly destroys it.
The Capability Has Outrun the Norms
A modern HRIS, combined with productivity tooling, calendar systems, security telemetry, and survey platforms, generates an enormous quantity of employee data. Most of it is technically usable for analysis. Some of it is ethically questionable to use. A smaller share is illegal in some jurisdictions to use without specific safeguards. The line between these categories shifts depending on country, employment contract, and analysis purpose, and it is easy for an analytics team to drift across the line without anyone noticing until something visible breaks.
Three Categories of Analysis That Hold Up
Aggregate workforce analytics — headcount trends, turnover by department, hiring funnel conversion, learning programme uptake, demographic representation in aggregate — are well-established, useful, and broadly accepted. The data sources are mostly the HRIS, and individuals are not identifiable in the analysis. This is the foundation. Most organisations under-invest here and over-reach into more sensitive areas before getting the basics right.
Voluntary survey-based analytics — engagement, manager effectiveness, inclusion experience, learning preference — produce useful signal when employees genuinely consent and the survey design is sound. The discipline here is statistical (avoid drawing conclusions from underpowered cuts) and operational (act on the findings, or trust degrades fast).
Compensation and progression equity analysis is increasingly required by regulators in many jurisdictions and is one of the highest-leverage uses of people analytics. Salary distribution by demographic, promotion rate parity, time-to-promotion gaps. The ethical case is strong; the operational case (unaddressed equity gaps surface in lawsuits and regulatory action) is also strong. Most organisations should be doing more of this.
Three Categories That Should Stay Off the Backlog
- Productivity surveillance — keystroke logging, screen capture, application time tracking aggregated to individual identification
- Sentiment analysis on internal communications — reading email and chat to predict morale or flight risk
- Predictive flight-risk scoring on individual employees presented to managers as actionable
- Location and presence tracking beyond legitimate access control purposes
- Inferring protected characteristics (health, religion, sexual orientation) from any data the employee did not voluntarily provide
- Performance prediction tied to individual employees, used for management action
A pattern that often surfaces: a vendor pitches an analytics product that crosses into one of these categories, framed as "early intervention to help managers retain talent" or "objective performance support." The framing is plausible. The reality of how managers actually use individual flight-risk scores or performance predictions is rarely benign — they become inputs to decisions that should be based on the manager's own observation and the employee's actual work. Decline these tools regardless of how they are marketed.
The Operating Discipline That Earns Trust
A people analytics function that builds trust shares structural features. Data minimisation is enforced — analyses use the smallest viable dataset, not the largest available. Employee notice is genuine — staff know what is collected, why, and how it is used. Aggregation thresholds prevent re-identification — no cell with fewer than ten employees, more depending on context. Independent review evaluates new analytical use cases before deployment. And the function partners closely with privacy and works counsel rather than treating them as obstacles.
The Question That Should Guide Every Analysis
"If our employees saw the exact analysis we are running, with the exact data sources, would they consider this acceptable?" If the honest answer is yes, the analysis is probably fine. If the answer is "they would object but they probably will not find out," do not run it. The discovery happens eventually, and the cost to the analytics function and to HR more broadly outweighs whatever insight the analysis would have produced.