No camera · No biometric data · No emotion inference

Engagement analytics that don't need a camera

Signals measures where your content loses people using behavioural telemetry alone. Deployable in EU workplaces and education institutions, where facial analysis is prohibited.

Request a demo Check your compliance position
#1
Tab-switching ranks as the strongest single predictor of online disengagement
ScienceDirect, 2024
+3.1pt
All that facial analysis adds over behavioural signals alone (91.5% → 94.6%)
OUCI/DNTB study
0
Biometric data points processed, stored or transmitted
By design
Why it exists

The rule that changed this market

Since 2 February 2025, Article 5(1)(f) of the EU AI Act has prohibited AI systems that infer emotions from biometric data in workplaces and education institutions. No consent exception. Penalties to 7% of global turnover.

Facial analysis is out

If your engagement tool infers emotional states from a camera feed, it cannot lawfully be deployed to EU employees or students — regardless of consent, contract wording or on-device processing.

Behavioural signals are not

The prohibition hangs on biometric data. Tab-switching, window focus and dwell time are browser events, not physiological characteristics, and they infer no emotional state.

And they predict better

The strongest single predictor of disengagement in the literature is behavioural, not facial. The camera was never where most of the signal lived.

What it measures

Seven signals, no camera

All captured through native browser APIs. No hardware, no permissions prompt, no installation.

01
Tab-switching
Page Visibility API — the strongest single disengagement predictor
02
Window focus loss
Focus and blur events — attention lost without a tab change
03
Dwell time per section
Intersection Observer — real time on each slide, focus-adjusted
04
Scroll depth & velocity
Whether content was read or skimmed past
05
Interaction latency
Response time to prompts — a cleaner signal than click counts
06
Replay & rewind
Sections needing a second pass — a confusion indicator
07
Session abandonment
Where people actually quit — the ground truth for prediction
What you get

The same answer, without the risk

Signals produces content diagnostics, not personal dossiers. That is both the compliance position and the commercially useful output.

Content-level engagement heatmaps

Every section scored. "Module 3, slide 14 loses 60% of learners" — a fix list, not a verdict on individuals.

Drop-off point identification

Exactly where sessions end, and what the behavioural pattern looked like in the minutes beforehand.

Per-learner behavioural baselines

Deviation from each person's own pattern rather than a population average, so habitual multitaskers are not permanently flagged.

Anonymous mode

Aggregate at the point of collection. No individual identifiable, no session linked to a person, no personal data processed at all.

Cohort comparison and A/B testing

Prove version B held attention better than version A on real behaviour rather than survey scores.

Disengagement prediction

Flag at-risk sessions from behavioural pattern breaks while there is still time to intervene.

Compliance

Why this passes where others don't

| | Facial emotion analytics | **Emotuit Signals** | |---|---|---| | Camera required | Yes | **No** | | Biometric data processed | Yes | **None** | | Infers emotional state | Yes | **No** | | EU AI Act Art. 5(1)(f) — workplace | ❌ Prohibited | **✅ Not caught** | | EU AI Act Art. 5(1)(f) — education | ❌ Prohibited | **✅ Not caught** | | GDPR special category data | Yes | **No** | | Can run with zero personal data | No | **Yes, in anonymous mode** | | Works council approval | Frequently refused | **Routinely approved** | | Consent prompt needed | Yes, and it does not cure the ban | **No camera permission at all** |
Still get advice

Signals is designed to sit outside the Article 5(1)(f) prohibition, and we will support your assessment with a written vendor position and a completed supplier section for your DPIA. That is not a substitute for your own legal advice on your specific deployment. Use the applicability checker to work through the test.

Deployment

One script tag

Under 50KB. Works with any LMS, SCORM package or xAPI platform. No migration, no server-side integration, no infrastructure change.

Any LMS

Canvas, Moodle, Docebo, Cornerstone, TalentLMS, Absorb, Blackboard, Open edX — anything browser-based.

SCORM & xAPI

Emits xAPI statements to your existing Learning Record Store, or runs standalone.

On-premise & air-gapped

Zero external dependencies. Runs fully offline for government, defence and regulated environments.