What question did this study set out to answer?

This research aims to develop and validate the Brand Spectrometer for measuring brand perception across different cohorts.

July 1, 2026Open Access

The Brand Spectrometer: A Reproducible Instrument for Cohort-Resolved, Multi-Dimensional Brand-Perception Measurement from Public Artifacts

Key Points

This research aims to develop and validate the Brand Spectrometer for measuring brand perception across different cohorts.
Introduced metameric psychometrics to validate instruments without a true score.
Utilized an open pipeline to reconstruct eight-dimensional perception vectors from public artifacts.
Conducted a pre-registered battery across two time windows to assess validity and reliability.
Achieved convergent validity and four-of-five test-retest reliability across cohort comparisons.
Resolved distinct owner cohort perceptions from the press with a signal-to-noise of 3.6, permutation p = .003.
Validated measurement properties indicate reliable cohort metameric variance without relying on criterion accuracy.

Abstract

This paper introduces metameric psychometrics, a validation framework for instruments that operate without a latent true score. In place of an oracle it substitutes instrument-computed noise floors – operator variance across cross-family model pairs and source sensitivity under signal-source resampling – so every cohort comparison becomes a falsifiable signal-to-noise test with a computed interval. The framework is realized in the Brand Spectrometer, an open pipeline that reconstructs cohort-resolved, eight-dimensional perception vectors from dated public artifacts by aggregating per-artifact reflections over signal sources; the unit of analysis is the cohort pair. Because cohort metameric variance is the measurement, not error around a true specification, the instrument is validated for measurement properties, never criterion accuracy. A pre-registered battery across two windows shows convergent validity, reproducibility, and four-of-five test-retest reliability. Discriminant resolution depends on the metric: a scale-invariant mean-cosine signal-to-noise abstains, whereas a distribution-level, operator-floored separation between whole cohort clouds resolves the owner cohort from the press in one window (signal-to-noise 3.6, permutation p = .003, triangulated) while abstaining elsewhere. The cohort signal lives in magnitude, not mean shape; the instrument resolves real separations and abstains where none clears. The framework may extend to perceptual constructs measured from public text without ground truth. Includes zharnikov-2026ax-brand-spectrometer.yaml (Paper Spec v0.1.0) – a machine-readable specification of the paper's claims, assumptions, and dependencies. The paper's full machine-first bundle (the SPINE claim/dependency graph and the ONTOLOGY term module) lives in the public repository; see https://github.com/spectralbranding/paper-spec for the standard. This PDF is generated programmatically from that machine-first source under a research-as-repository model.

Read Full Paperexternally

Perguntar à IA

Bookmark

View Full Paper