People Analytics: The Complete Guide
People analytics is the use of workforce data to improve HR decisions. It measures what employees do and what they say. This guide explains what people analytics covers, where it structurally falls short — particularly around retention prediction — and how emotional AI closes the gap.
Published May 2026 · By Jonathan Prescott · Cavefish
What is people analytics?
People analytics (also called HR analytics, workforce analytics, or talent analytics) is the use of data and statistical methods to improve decisions about people in organisations. It encompasses a wide range of data types and analytical approaches:
- Recruitment and selection data: Application volumes, interview outcomes, time-to-hire, source quality, offer acceptance rates
- Performance metrics: Objective-setting outcomes, performance review scores, 360-degree feedback, promotion rates
- Engagement and sentiment: Annual engagement surveys, pulse surveys, eNPS scores, open-text qualitative responses
- Retention and turnover: Voluntary and involuntary turnover rates, tenure by role and team, exit interview data
- Learning and development: Training completion, skill acquisition, internal mobility, career progression
- Workforce planning: Headcount forecasting, skills gap analysis, succession planning
The discipline has expanded significantly as HR analytics platforms — Workday, SAP SuccessFactors, Qualtrics, Culture Amp, Glint, Peakon — have made large-scale data collection and analysis accessible to non-specialist teams.
People analytics vs HR analytics
These terms are often used interchangeably. Where a distinction is made:
- HR analytics typically focuses on operational HR metrics: time-to-hire, cost-per-hire, headcount planning, absence rates, compensation benchmarking
- People analytics often implies a broader scope: culture health, organisational effectiveness, workforce behaviour, and the connection between employee experience and business outcomes
Both disciplines share the same structural limitation: they cannot access the emotional experience that drives the behaviours they observe. They measure what employees do and what employees say — not what employees feel.
The three layers of people analytics data
Understanding what people analytics can and cannot measure requires recognising three distinct data layers:
Layer 1 — Behavioural
What employees do
Attendance, performance ratings, promotion velocity, tenure, mobility, absence patterns. Well-captured by HRIS systems. Lagging signal — behaviour changes after decisions are made.
Layer 2 — Stated sentiment
What employees say they feel
Engagement scores, eNPS, pulse survey responses, open-text comments. Captured by Culture Amp, Qualtrics, Glint. Lagging signal — socially moderated responses underestimate negative affect.
Layer 3 — Felt sentiment
What employees actually feel
The emotional signal beneath written responses — text-emotion divergence, masked dissatisfaction, the pre-exit emotional pattern. Captured by emotional AI. Leading signal — detectable 3-6 months before resignation.
Traditional people analytics operates on layers 1 and 2. Emotional AI adds layer 3 — the leading indicator that predicts what the other layers will show months later.
The people analytics toolkit
Understanding what each tool type contributes — and what it cannot see — helps position emotional AI correctly within a broader people analytics programme.
| Tool type | What it measures | Structural limitation |
|---|---|---|
| HRIS (Workday, SAP) | Attendance, performance, tenure, mobility | Behavioural lag — signals appear after decisions made |
| Engagement platforms (Culture Amp, Glint, Qualtrics) | Stated sentiment — what employees write in surveys | Social moderation — masked dissatisfaction invisible |
| Exit interview tools | Retrospective stated reasons for leaving | Post-decision rationalisation — not predictive |
| Sentiment analysis (NLP) | Linguistic sentiment of written responses | Word-level only — cannot detect text-emotion divergence |
| Emotional AI (EchoDepth) | Felt sentiment — emotional signal beneath written response | Depth tool, not scale tool — 20-500 respondents |
Why people analytics fails to predict retention
People analytics is very good at describing what has happened. Turnover was 14% in Q3. The engineering team's engagement score dropped 0.6 points. Voluntary exits in sales increased year-on-year. These are accurate, useful observations.
People analytics is less reliable at predicting what is about to happen — particularly voluntary turnover, which is the most costly and disruptive outcome most programmes are designed to prevent. The standard predictors (engagement score decline, absence increase, performance dip) are all lagging signals. By the time they appear in the data, the employee has already made their decision.
The structural reason: people analytics measures stated sentiment, not felt sentiment. Employees who have already decided to leave often continue to perform normally, attend reliably, and write neutrally in surveys for months before the resignation arrives. The data shows nothing unusual. The emotional reality — anxiety, disconnection, loss of trust in leadership — is not visible in any metric that people analytics currently captures.
How emotional AI adds the missing layer
The emotional AI layer does not require new surveys, new participants, or new platforms. It is applied to the open-text data your existing people analytics programme already collects — the comments fields in your Culture Amp, Qualtrics, Glint, or Peakon surveys.
EchoDepth applies VAD (Valence, Arousal, Dominance) emotional analysis to these responses — independently of the words chosen. The comparison between written content and emotional signal produces:
- Text-emotion divergence: Where an employee writes positively but the emotional signal is negative
- Masked dissatisfaction: Responses flagged as containing suppressed negative sentiment
- Employee Culture Index (ECI): 0-10 composite emotional health score
- Pre-exit pattern detection: The emotional signature that precedes resignation by 3-6 months
The output is not a replacement for your existing people analytics dashboard. It is an additional layer: a board-ready culture health report identifying which teams, themes, and time periods show the largest gap between what employees write and what the emotional signal suggests they feel.
Further reading
What is People Analytics?
Deep dive into people analytics definitions, tools, and the emotional AI layer that closes the prediction gap.
How to Measure Employee Sentiment
Step-by-step guide to measuring sentiment accurately, including the limitations of traditional surveys.
Employee Sentiment Survey
Question templates, design best practices, and how to analyse survey data with emotional AI.
Employee Sentiment Guide
Complete guide to employee sentiment: definition, measurement, and the metrics that matter.
Predicting Employee Turnover
How emotional AI detects the pre-exit condition 3-6 months before traditional people analytics.
Culture Measurement
How EchoDepth adds the emotional AI layer to your existing culture survey programme.
Add the emotional AI layer to your people analytics
Send us your most recent culture survey open-text export. We will show you the emotional AI layer — ECI scoring, text-emotion divergence, and the pre-exit pattern — within five working days.
Start the conversation →