Signal acquisition without surveillance: a minimal telemetry blueprint
Published: 14 Jan 2026 · Category: Signals · Back to Notes
Most teams do not need emotion detection. They need a way to spot friction and recovery points early, then redesign the interaction to reduce pressure.
This note is a minimal blueprint for “signal acquisition” that keeps you away from surveillance patterns. It focuses on system-level signals and opt-in feedback. It is compatible with privacy-by-design and it scales across web and mobile.
Principle 1: Measure the interaction, not the person
Prefer signals like loops, time-to-complete, retries, and abandon points. Avoid biometric inference, sentiment prediction, or any attempt to guess mental state.
Principle 2: Keep telemetry explainable
If you cannot explain a metric in one sentence, do not collect it. Your goal is to make design decisions, not to build a hidden risk model.
Principle 3: Separate “quality signals” from “security signals”
Security logging and fraud detection are different problems. If you mix them, you create pressure and you create mistrust. Treat calm design telemetry as a quality signal with short retention and strict minimisation.
Minimum viable telemetry (MVT)
Start with a small set of events that identify friction and recovery. Keep identifiers ephemeral.
| Event | What it tells you | Why it is safe |
|---|---|---|
step_view |
Which step users reached | No personal content, only flow position |
step_complete |
Time-to-complete and drop-off points | Pure interaction timing |
error_shown |
Where the system fails users | Captures the system state, not user identity |
retry |
Loops and repeated attempts | Highlights design pressure and ambiguity |
help_open |
Confusion points | Tracks need for clarity |
save_and_return |
Where people need time back | Shows pacing pressure without profiling |
feedback_quick |
User-controlled “clear / unsure” input | Explicit consent and agency |
A simple event schema you can ship fast
Store only what you need to redesign. Everything else is optional.
{
"event": "error_shown",
"flow_id": "onboarding-v3",
"step_id": "verify_email",
"timestamp": "2026-01-14T12:34:56Z",
"session_id": "ephemeral-uuid",
"attempt": 2,
"meta": {
"error_code": "EMAIL_TIMEOUT"
}
}
Important: Keep session_id short-lived. Avoid collecting typed input, free-text logs, clipboard content, photos, or “emotion scores”.
Consentful feedback that does not annoy people
- Ask after a natural completion point, not mid-task.
- Use one click options like “Clear” and “Unsure”.
- Make the prompt easy to dismiss without penalty.
- Explain why you are asking in one sentence.
Operational guardrails (these prevent drift)
- Retention: keep quality telemetry short, for example 14 to 30 days.
- Sampling: measure enough to learn, not enough to profile.
- Access: product and design teams only, not broad internal sharing.
- Purpose: redesign for calm, not behavioural targeting.
Where to go next
For a ready-to-use mapping from signal to meaning to response, open Signals. For design checks that reduce pressure before measurement is needed, open the Toolkit.