Emotional Intelligence Pharma · FMCG · Healthcare · People & Culture

What people feel is not what they tell you.

EchoDepth Insight captures the involuntary emotional response that focus groups, surveys and post-session interviews will never reveal. Real-time. Remote. No guesswork.

GDPR-compliant · No wearables · Remote delivery · Built by Cavefish

Featured in The Daily Telegraph · DSEI 2025 · ITEC 2026 speaker
Pharma | FMCG | Healthcare | people and culture | FCA Sandbox | 13 Awards | Research Insights | ISO 9001 | GDPR Compliant

People & Culture

Your engagement survey says 7.2. People keep leaving. Here is why.

Culture surveys systematically underestimate negative affect. Employees moderate their language when leadership behaviour is involved. EchoDepth analyses the emotional signal beneath the text — surfacing masked dissatisfaction, flagging text-emotion divergence, and producing board-ready culture health reports from data you already collect.

ECI Score

4.4/10

↓ from 6.6 six months ago

Emotional Risk

67/100

90-day forecast

Masked dissatisfaction

34%

of responses flagged

Turnover prediction

8–16wks

before behavioural signal

Latest research

All insights →

People analytics

What is employee sentiment analysis?

Why NLP-based tools measure the output of a filtering process — and what emotional AI measures instead.

Read →

People analytics

Employee engagement vs employee sentiment

Engagement is what employees are willing to tell you. Sentiment is what they actually feel. The gap is where retention risk hides.

Read →

People analytics

How emotional AI predicts employee turnover before it happens

The pre-exit emotional pattern surfaces in survey data 8–16 weeks before any behavioural indicator appears.

Read →

People analytics

How to measure employee sentiment accurately

Why self-report systematically underestimates negative affect — and what emotional AI measures instead.

Read →

Culture measurement

Why your culture survey score doesn't predict what's coming

The masked dissatisfaction mechanism — why scores look stable before resignations arrive.

Read →

Pharma research

Social desirability bias in pharma research

Why patients say what researchers want to hear — and how FACS-based emotional AI detects the divergence.

Read →

Research methodology

Why focus groups fail to predict market success

Three structural flaws no facilitator can fix — and what the evidence says to do instead.

Read →

People analytics

What is people analytics — and where does it fall short?

The three layers of people analytics — and the emotional layer that traditional programmes cannot see.

Read →

Methodology

What is FACS analysis — and why does it matter?

The Facial Action Coding System explained — 44 Action Units, VAD scoring, and why it is the scientific standard.

Read →

The Problem with Traditional Research

Your participants are performing, not responding.

Every traditional research method puts the observer effect at its centre. The moment people know they're being studied, the data is compromised.

Traditional methods — what you get

  • Social desirability bias — participants answer what they think researchers want to hear, not what they actually feel
  • The articulation gap — people simply cannot accurately describe their emotional responses in words
  • Post-rationalisation — conscious reflection after stimulus exposure distorts and rewrites emotional memory
  • Peer pressure in groups — dominant voices suppress minority responses; groupthink contaminates qualitative data
  • Recall bias — post-session surveys ask people to remember how they felt, not how they actually felt
  • Geographic lock-in — facility-based research limits participant diversity and drives up cost
  • No moment-to-moment data — you get a rating at the end, not a signal throughout

Read: why self-report systematically underestimates negative affect →

EchoDepth Insight — what you get instead

  • Involuntary physiological signal — facial Action Units cannot be consciously controlled; they reveal genuine response
  • No articulation required — VAD scores are computed from behaviour, not language
  • Real-time, in-stimulus capture — emotional data is recorded at the moment of exposure, not recalled afterwards
  • Individual, not group — every participant is analysed independently; no peer contamination
  • Remote delivery — global participant access from any device with a camera
  • Moment-to-moment timeline — see exactly when and where emotional response peaks and drops during the stimulus
  • Quantified, comparable output — VAD scores enable statistical analysis, segment comparison and longitudinal tracking

How EchoDepth Insight Works

Four stages. One session. Truth throughout.

1

Recruit & brief

Participants join remotely via any device with a camera. Consent is captured digitally. No facility. No travel. Global access.

2

Expose & capture

Stimuli — concepts, packaging, messaging, video, or interview questions — are delivered in-session. EchoDepth captures 44 facial Action Units continuously throughout exposure.

3

Score & flag

VAD scores are computed in real time. Key moments — emotional peaks, drops, surprise signals — are automatically flagged with timestamps. Moment-to-moment emotion curves are generated per stimulus.

4

Report & recommend

Structured session reports with emotion timelines, group aggregates, segment comparisons and confidence scores. From insight to marketing recommendation in a single workflow.

Client perspective

"Using EchoDepth from Cavefish allows us to validate ideas quickly, minimising the risk of launching a product or idea."
GT

Gethin Thomas

CEO, Iterate

Why it matters

"The gap between what consumers say in research and what they do in market has always been the central problem of insights. EchoDepth measures the signal that language cannot — the involuntary physiological response that drives behaviour."

EchoDepth Insight does not claim to read minds. It measures Valence, Arousal and Dominance — the three quantifiable dimensions of emotional state — at the moment of stimulus exposure. Read our ethics approach →

44
Facial Action Units tracked per frame
14
Cultural cohorts in training data
6
Countries in training dataset
13
Industry awards

Get Started

Stop asking what people think. Measure what they feel.

Book a 30-minute discovery call. We'll walk through your research challenge and show you what emotional capture looks like in practice.