Reliable measurement needs coverage of relevant decisions and enough repetition to interpret variation. The question count depends on the market, range of intentions and change you want to detect. One hundred questions may map one scope well and inadequately cover another.
Universe quality matters as much as size. Duplicating equivalent questions increases the count without necessarily broadening coverage. Adding questions far from the buying decision can dilute the indicator.
Start with the decision the report supports
For content planning, cover problems, requirements, comparisons and hiring. For identity, maintain a separate branded-question group. For growth monitoring, define a stable core of commercial questions.
Declare market and language. Brazilian Portuguese research does not automatically represent English queries or buying decisions in other countries. Preserve the original scope in translated reports.
Assign questions to families before collection. Explain any weights and keep them consistent. Without demand data, a weight is a design choice, not a proven estimate of buyer volume.
Questions and repetitions serve different purposes
More questions broaden coverage. Repetition tests recurrence within a condition. Different engines broaden product coverage but are not identical repetitions of one experiment.
The SparkToro study included 600 volunteers and 2,961 executions across 12 categories and documented recommendation inconsistency. It supports attention to variability, not a universal minimum repetition count.
Combining coverage and depth
murmur’s August 6, 2026 Brazilian campaign, before this article collection was published, used:
| Layer | Design | Captures |
|---|---|---|
| Coverage | 100 Portuguese questions × four engines × one execution | 400 |
| Repetition | 25 of those questions × four engines × four additional executions | 400 |
| Total | The 25 priority questions had five executions each | 800 |
This maps initial presence and recurrence for priority questions. Repeating 25 questions does not establish equal stability for the other 75. AI Overview was Google’s surface; the Gemini app was excluded.
A smaller pilot, such as three questions × five executions × four engines = 60 captures, can assess a narrow hypothesis. This is a sizing example, not a reliability certification.
Choosing questions for repetition
Prioritize questions influencing hiring, differentiating the company or yielding ambiguous results. Include meaningful intent variations while keeping paraphrases distinct if comparing them.
Record attempts, valid captures and failures. Error screens or still-loading answers should not silently become brand absences. If coverage differs across periods, report the total collected and the comparable subset.
Assessing meaningful change
Repeat the baseline when needed to understand variation in your universe. Examine differences by question, intention and engine as well as the aggregate. Repeated questions are not necessarily independent observations; a simple margin-of-error calculation may imply precision the design cannot support.
A threshold observed for one client is not a universal significance rule. The review arXiv 2607.14035 likewise emphasizes evidence design. For contractual SOV, agree on the indicator, universe, window, eligibility and change handling in advance.
Frequently asked questions
Are 100 questions enough?
It depends on market diversity and the objective. Assess intent coverage, repetition and period comparability, not just count.
Do five repetitions guarantee stability?
No. They describe recurrence within that set. Variability and decision stakes may require more observation.
Can questions change between rounds?
Yes, to explore new needs. Keep a common core for SOV comparison and report expansion separately.
How should collection failures be handled?
Record reasons and counts. Declare the eligible denominator and do not turn technical failure into brand absence.
How murmur works
murmur works through a SWAS model, combining proprietary measurement software with specialists who diagnose, plan and execute GEO to grow Share of Voice (SOV). Eligible contract models may include an SOV-growth guarantee, with the metric, baseline, period, scope and conditions agreed in writing. This does not guarantee an individual answer, a fixed position or sales.
Author: Mateus Gomes, murmur operator. This article explains our approach and discloses our commercial interest.