Evidence-bound. Doctrine-locked. Human-reviewed.
NuX Holdings™ does not estimate empathy. It measures human experience quality against an evidence standard — NuX v0.1 — and rolls it up under Directive 6. Every score on every executive surface carries its confidence provenance, in plain sight.
The six measured dimensions
Each interaction is scored on six orthogonal dimensions. Reviewer notes are stored as evidence and surfaced on the score detail view.
Was the person seen as themselves, not as a ticket type?
Was the next step unambiguous and honest about constraints?
Was the interaction free of coercion, condescension, or scripted dishonesty?
Did a human or system take accountability — or pass the problem on?
Did the answer hold across channels, agents, and AI handoffs?
Did AI assist within its competence and hand off cleanly when not?
From interaction to executive surface
- 01Ingest
Interaction signal lands via CSV, connector, or AI agent log. Field mapping is explicit; nothing is inferred about identity that wasn't provided.
- 02Score (NuX v0.1)
Each interaction is scored across the six dimensions on a sampled cohort, with reviewer notes attached as evidence.
- 03Composite (Directive 6)
Platform-wide composite — Empathy, Warmth, Precision, Dialogue, Resilience, Safety — rolls NuX evidence into executive-grade exposure.
- 04Calibrate
Per-tenant calibration corrects for channel mix, business unit, and language. Calibration coefficients are versioned in scoring_config.
- 05Human review
Anything that crosses the judgment threshold is human-reviewed before it lands on an executive surface. Reviewer identity is recorded.
- 06Publish
Scores ship to Watchtower tagged ● Measured · ◆ Human-reviewed. Everything else carries its honest provenance flag.
Every number carries its provenance
No NuX Holdings surface shows a number without telling you how it was produced. The four levels below appear inline on every score, chart, and executive briefing.
Real ingested data — human-audited under NuX v0.1. Defensible to CFO, auditor, and regulator.
Directional projection or demo data. Not a measurement. Replaced once real data lands.
Why your number is your number
Channel mix, business unit, and language shift the raw distribution. Calibration coefficients are versioned in scoring_config so every score can be reproduced against the exact coefficients used to publish it.
Scores that escalate exposure — attrition, complaints, regulatory, AI-human transition — are gated behind named reviewer sign-off. Reviewer identity is logged to the audit ledger alongside the publish event.
Sampled cohorts are double-scored to measure reviewer drift. Drift beyond threshold triggers re-calibration before further publishing.
Reviewer notes and the evidence chain are retained for the life of the engagement and exportable for audit. The score is never separated from the evidence that produced it.