The analytics decision
Freeze a core panel and version every intentional change. Keep raw snapshots so analysts can revisit a metric without rewriting history.
Build the data contract
- Document the panel version and activation date
- Record scheduled, successful, failed, and missing observations separately
- Normalize brand and citation identities with reversible rules
- Compute metrics from immutable snapshots and stated denominators
- Annotate product, model, prompt, and extraction changes on the timeline
Example data flow
Adding 50 purchase prompts to a 20-prompt awareness panel will raise the weight of commercial questions. The next total is not directly comparable unless the series marks the panel break.
Evidence to retain for consistent AI search analysis over time
Keep these fields with the decision:
- source object
- stable key
- time grain
- metric definition
- missing-data state
- transformation
- reconciliation
- owner
Limits and UnderAI's role in consistent AI search analysis over time
Even a stable method observes a changing system. Consistency improves interpretation; it does not make AI answers deterministic.
UnderAI stores dated Answer Snapshots and immutable prompt text after creation, supporting a versioned longitudinal analysis.