Study method
How a study is run, and how to replicate it.
Every published benchmark follows the same five steps. The methodology version and evaluation window are stated on the study so the result can be checked and repeated.
01 — Sample selection
- The population is defined before any scoring begins, with the inclusion rule stated in the published study.
- Organizations are selected on public criteria — sector, size, market — never on relationship.
- Evaluation is unsolicited; organizations are not contacted before publication.
02 — Evidence capture
- Owned surfaces are crawled and captured with retrieval dates.
- Third-party surfaces — profiles, registries, reviews — are captured in the same window.
- Controlled prompts are run against named AI assistants, with the prompt text published.
03 — Scoring
- All fifteen dimensions are scored on the same 0–10 scale using published anchors.
- Brand Integrity and Signal Strength are calculated, and the Signal Gap derived.
- Every score is linked to the evidence that produced it.
04 — Review
- A sample is independently re-scored without sight of the first set.
- Disagreements greater than one point are re-evidenced before release.
- Findings are written from the scores, not the other way round.
05 — Replication
- The methodology version, evaluation window and prompt set are published with every study.
- Any reader following the same steps in the same window should reach the same range.
- Re-measurement uses the same version, so before-and-after comparisons remain valid.
Inclusion and corrections
Inclusion in a Brand & Signal benchmark is not an endorsement, and the organizations evaluated are not clients, sponsors or partners of the benchmark. Evaluation is unsolicited and based on publicly observable evidence.
Any organization may request a correction. We review the evidence, and where a factual error is confirmed we republish with a dated change note. Corrections address factual accuracy only — scores cannot be purchased, negotiated or removed, and no commercial relationship changes a result.