How SI separates signal, interpretation and proof.

Repeated or material changes in search, technical conditions, reviews, demand, representation, competitors or journeys.

Not automatically a cause, conclusion or action priority.

Reported with source, period, scope, recurrence, magnitude, context and confidence.

May trigger monitoring, validation, materiality assessment or escalation.

Random volatility, measurement error, duplication, irrelevant scope, short-lived anomalies and non-material variation.

Not merely an inconvenient result; an apparent anomaly may later prove to be a signal.

Documenting volatility, uncertainty, data quality and reasons for exclusion or continued monitoring.

Correctly identifying noise prevents wasted action and weak governance decisions.

Data, captures, logs, research, customer statements, technical outputs, reports, records and validated observations.

Not the same as interpretation, certainty or proof of causation.

Every material item retains source, date, scope, method, context, limitations and confidence.

Gives decisions an auditable trust path.

Contextual analysis, comparison, causal hypotheses, materiality, competing explanations and decision implications.

Not fact; must not be presented without its evidence basis and uncertainty.

Reported separately from observed facts and labelled with confidence and limitations.

Enables action while preserving evidence discipline and challenge.

Commercial value, customer harm, trust, risk, technical severity, category position, recurrence, reversibility and strategic relevance.

Not determined by metric size alone; may differ by sector, product, journey or risk appetite.

Reported through defined criteria, assumptions, evidence confidence and rationale.

Supports prioritisation, escalation and proportionate response.

Source quality, recency, consistency, triangulation, relevance, repeatability and known limitations.

Not Customer Decision Confidence; does not state how confident the customer felt.

Reported as a defined level with rationale, not as an unexplained score.

Determines how strongly a decision may rely on the evidence and whether further validation is required.

Result presence, captured AI output, log event, technical error, transaction, stated response or direct interaction evidence.

Does not mean causal, complete or representative beyond the capture conditions.

Reports must state what was observed, where, when and how.

Observed facts form the base of the Evidence Trail.

Likely encounters, confidence movement, causal pathways and customer transitions supported by patterns or triangulation.

Must not be presented as observed fact.

Reports show the supporting evidence, alternative explanations and confidence level.

Enables useful interpretation while protecting credibility.

Demand opportunity, avoided loss, probability, scenario, impact range and forecast.

Not observed fact or guaranteed outcome.

Reports disclose model logic, inputs, assumptions, ranges, sensitivity and limitations.

Supports decisions where direct measurement is incomplete but must not create false certainty.

Triangulated data, repeated tests, customer research, operational confirmation and outcome evidence.

Does not remove all uncertainty or prove universal causation.

Reports state the validation method and remaining limitations.

Validated findings can support stronger governance action and control design.

Unmeasured effects, missing data, inaccessible journeys, unresolved causal questions and unverified claims.

Must not be converted into zero, no risk or assumed safety.

Reports state what is unknown, why and what evidence would reduce uncertainty.

Naming unknowns prevents false confidence and supports responsible risk decisions.

Competitor movement, technical deterioration, representation change, demand shift, review pattern and category disruption.

Not a prediction or certainty that harm or opportunity will occur.

Reported with evidence confidence, lead time, plausible consequence and monitoring or response recommendation.

Enables earlier mitigation, protection or opportunity capture.

Multi-touch influence, trust support, validation, demand creation, journey progression and risk reduction.

Not attribution; does not claim that one factor caused the outcome alone.

Reported with evidence basis, confidence, alternative contributors and limits.

Supports value interpretation without overstating causation.

Single-touch, multi-touch, experimental, econometric or validated customer evidence approaches.

Not automatically established by a conversion path, last click or correlation.

Reports disclose the attribution method, assumptions, data limitations and uncertainty.

Governance should use attribution cautiously for investment and accountability decisions.