1. Define the question universe
Agree the decision, audience, indication, brand or asset, competitors, lifecycle, geographies, and priority question groups before collection.
Transparent methodology
Know what the numbers mean.
AI answers change. We show the scope, cited sources, and limits behind every result.
Method
Five steps from scope to action
2. Run a controlled protocol
Use an approved provider and model set, consistent prompts, recorded timestamps, and repeated runs where variability needs to be understood.
3. Structure each answer
Capture the answer, brand mentions, positioning, messages, citations, domains, and the relevant evidence context at question level.
4. Aggregate carefully
Roll answer-level observations into useful views while keeping sample size, question mix, provider mix, and lifecycle interpretation visible.
5. Review before action
Inspect the underlying answer and evidence, record a human interpretation, assign an owner, and track the disposition.
Scope is part of the result
Scores are meaningful only with the question set, provider and model set, geography, run window, and lifecycle context used to produce them.
Variance is information
Repeated runs and provider differences can be reported rather than hidden. The necessary repeat structure depends on the decision and the cost of uncertainty.
Evidence outranks the index
An aggregate helps teams navigate. The captured answer, cited source, and documented interpretation remain the basis for review.
Lifecycle handling
Comparable discipline does not require identical language.
In-market reporting can discuss brand visibility, positioning, and message performance. Pre-launch reporting should focus on narrative formation, category language, evidence readiness, and whitespace. Portfolio views must label those differences rather than flatten them.
In-market
- Brand visibility
- Competitive positioning
- Message presence and credit
- Question-level opportunity
Pre-launch
- Narrative formation
- Category and mechanism framing
- Evidence readiness
- Emerging whitespace
Build the benchmark after the method is ready.
Any future public index should show its cohort, collection window, exclusions, and current methodology.