How Auraqu sentiment panels are built

Sample frame

The panel unit is the named individual, not the company. At top-tier respondents, coverage spans five operational roles — CFO and finance, commercial and wholesale, digital commerce, operations, and marketing partnership. Multi-respondent coverage at the same firm enables triangulation; brands are a derived grouping at analysis time, never the atomic unit.

Recruiting and verification

Respondents are sourced from public partner registries, trade directories, and verified professional records. Each contact is matched to a defined role and seniority before outreach. Generic shared inboxes are excluded. Vol. II onward augments the sample frame with peer referrals captured at the close of each completed interview.

Interview instrument

Two-track design. Four banded calibration questions establish distributional comparability across the cohort (category, geography, channel share, payment-term band). Eight open-ended questions capture the sentiment corpus — what is changing, in which direction, and why. The open block is the product; the calibration block makes it cross-sectionally comparable.

Field operations

Interviews are conducted by an AI voice agent on business telephone lines only. AI status is disclosed at the open. Verbal recording consent is obtained per applicable jurisdictional standards (§201 StGB in the DACH region; TCPA-compliant disclosure for US panels). Average completion time: ten minutes. Refusal at any point ends the call cleanly; opt-outs are recorded and honored permanently.

Signal extraction

Each completed interview is transcribed and indexed. A structured pipeline tags topics raised by the respondent (payment terms, returns friction, channel allocation, competitive positioning, and sector-specific themes), assigns sentiment polarity and directional vector (worsening / stable / improving) per topic, and links every extracted signal to the verbatim quote and transcript ID it derives from. No signal is published without its evidentiary anchor.

Publication standard

A confidentiality threshold of fifty respondents per published figure governs all standard releases. Cells below fifty are either marked low-confidence (10–49) or withheld entirely (under 10). Brand-attributable findings require three or more respondents at the same firm; smaller cells appear only in tier-level aggregates. Every figure carries the underlying respondent count, wave date, and a transcript-level audit path on request.