Appendix A. The ten-question buyer bank¶
Reference material for Chapter 8 at /valid-measurement/. This appendix reproduces the question bank design at the level the source volumes publish it, and no deeper.
What a question bank is for¶
Reasoned. A visibility measurement is a measurement of performance on a set of questions. The bank therefore defines what "visibility" means for that measurement, and two banks over the same company on the same engine in the same window will produce different scores. That makes the bank the single most consequential design decision in the instrument, and the one most often left undeclared.
Rule one: questions are phrased the way a buyer types them¶
Measured, Volume I. The Volume I bank used ten standardised buyer questions per category, a mix of best-of, direct comparison and use-case questions, phrased the way a buyer types them. Reasoned. The alternative, questions containing the subject's brand name, measures whether the engine knows the company exists rather than whether it competes. A bank of brand-name questions will produce a flattering score that has no relationship to whether a buyer researching the category encounters the vendor.
Rule two: the bank is fixed before collection and never regenerated¶
Measured, Volume III. Of 280 questions, 276 came from the study's existing question bank and 4 from the published Volume I dataset. No question was regenerated. Reasoned. Regeneration between measurements introduces a degree of freedom that can be tuned, deliberately or otherwise, toward questions the subject performs well on. Fixing the bank removes it, and publishing the bank makes the fixing checkable.
Rule three: every category covers the shapes that matter¶
Measured, Volume III. Each category was required to carry at least one best-of, one alternatives, one comparison, one use-case and one pricing or evaluation question. 33 of 40 categories met that in full. The 7 that did not are listed explicitly in the published sample manifest rather than quietly padded, because the question bank for those categories did not contain the missing shape.
Rule four: the realised mix is reported, not just the intended one¶
Measured, Volume III. The realised shape mix across the 280 questions was 80 alternatives, 80 best-of, 47 comparison, 38 evaluation, 33 use-case and 2 other. Reasoned. Intended coverage and realised coverage differ whenever a bank is drawn from a finite pool. Reporting both is what allows a reader to judge whether a score is weighted toward shapes that are easy or hard for the subject, and Chapter 7 at /overview-trigger-rate/ shows that the mix also moves the trigger rate and therefore the denominator.
The six shapes¶
Measured, Volume II. The classification scheme, its rules and its precedence order are reproduced in full in Appendix B at /absence-rules/. In summary the shapes are head-to-head comparison, alternatives-to-incumbent, best-of shortlist, evaluation criteria, use case or problem, and definitional.
A ten-question template, offered as construction¶
Reasoned. The volumes publish the shapes and the coverage requirement but do not publish a fixed ten-question allocation, so what follows is an explicit construction rather than a reproduction. A ten-question bank that satisfies the published coverage requirement and reflects the realised mix would carry roughly three best-of or shortlist questions, three alternatives-to-incumbent questions, two head-to-head comparisons naming real rivals, one evaluation or pricing question, and one use-case question. Adjust the weighting to your category, declare the weighting you used, and then never change it.
| Slot | Shape | What it tests |
|---|---|---|
| 1 to 3 | Best-of shortlist | Whether you are assembled into the candidate set at all |
| 4 to 6 | Alternatives-to-incumbent | Whether you are reachable from the incumbent's name |
| 7 to 8 | Head-to-head comparison | Whether you win preference against a named rival |
| 9 | Evaluation or pricing | Whether you appear when the buyer is comparing on criteria |
| 10 | Use case or problem | Whether you appear when the buyer describes a problem rather than a category |
Rule five: the questions name real rivals, not aspirational ones¶
Measured, Volume II. For every answer the study recorded which competitor was named most often. Reasoned. The rivals a buyer pairs you with are an observable fact rather than a positioning choice, and comparison questions built around a rival nobody pairs you with will measure nothing. Build the comparison slots from the pairings that appear in your own absence list, which Chapter 10 at /comparison-gate/ discusses.
Rule six: the bank is published with the score¶
Reasoned. Without the question text, a reader cannot tell whether a high score reflects broad presence or a bank weighted toward definitional questions where almost everyone appears. Publication is what makes a score checkable by somebody who did not run it, and it costs nothing.
The limitation the source volumes state about their own bank¶
Measured, Volume II. The ten questions per category were generated to reflect realistic buyer research rather than sampled from real search logs. Volume II states that a different question set would produce a different absence mix, and that the shapes are drawn from that set. Reasoned. So the absolute proportions in any published result are set-dependent, while the tier comparison is internally consistent because every tier faced the same bank. Read the ordering as the finding and the decimals as properties of the instrument.
What is deliberately not in this appendix¶
Reasoned. This appendix publishes the design rules and the shape taxonomy at the level the source volumes already publish them. It does not describe how any commercial measurement is operated, scheduled or delivered, and a reader looking for that will not find it in this manual.
Sources¶
The bank design, provenance and coverage requirement, the 33 of 40 disclosure and the realised shape mix are from The 2026 State of GEO, Volume III. The ten-questions-per-category design and its phrasing rule are Volume I. The shape taxonomy and the limitation on question generation are Volume II. The ten-slot template is an explicit construction by this manual and is labelled as such. Everything is at github.com/Broadcastwell/state-of-geo-2026.
About this manual¶
Author. Sairam Sivakumar, Broadcastwell.
Operator disclosure. Broadcastwell ran this measurement and sells services in the category it measures. Broadcastwell is excluded from the measured sample and from every ranking. The mitigation is not that the conflict is absent, it is that the raw data and the code are public and the result can be recomputed by anyone who disagrees.
Self-audit, July 2026. In the four-engine, five-run self-audit published alongside Volume II in July 2026, Broadcastwell was named in 0 of 200 answers and cited 0 times among 663 citations. That published figure stands with its date and is never replaced.
Self-audit, 18 August 2026. Re-measured on the same ten published questions across the same four engines, at one run per question rather than five, between 03:51 and 04:03 UTC on 18 August 2026: named in 1 of 40 answers, and cited once among 674 citations. The single naming and the single citation are the same answer, in which the engine quoted Broadcastwell's own published visibility page as a source. The two lines are not directly comparable, because one rests on five runs per question and the other on one.
Not peer reviewed. This is an independent industry study published as an open dataset with the analysis code that produced every figure in it. It has not been through academic peer review. Read it as measurement, and check the measurement. If you disagree with a number here, recompute it from the public data and publish what you get.
Licence. Prose and figures CC BY 4.0. Site code MIT.
The three volumes.
| Volume | What it covers | DOI |
|---|---|---|
| Volume I | 85 companies, 61 categories, 860 scored answers and 5,160 citations, one engine held constant. The dataset README additionally records 1,753 unique domains cited | 10.5281/zenodo.21537014 |
| Volume II | The Absence Ladder. All 616 absence records classified by question shape | 10.5281/zenodo.21586091 |
| Volume III | Cross-engine divergence. 280 questions, 40 categories, four engines, 853 answers | 10.5281/zenodo.21789120 |
Data and analysis code for all three volumes: github.com/Broadcastwell/state-of-geo-2026.
Version 1.0, August 2026.