Score competitive signals on two separate axes: evidence confidence and decision materiality. Do not average them into one magical priority number. A verified low-impact change and an uncertain high-impact possibility require different treatment, even if a formula gives them the same score.
What this guide covers
- why one-dimensional signal scores mislead;
- an evidence-confidence rubric;
- a decision-materiality rubric;
- routing rules, overrides, and calibration.
Separate “is it true?” from “does it matter?”
Evidence confidence asks whether the public record supports the scoped observation. Materiality asks whether the observation could change a named decision for a relevant segment.
| Low materiality | High materiality | |
|---|---|---|
| High confidence | retain or summarize | route to decision owner |
| Low confidence | reject or defer | research urgently; do not state as fact |
This simple matrix prevents two common errors: flooding leaders with verified trivia and presenting an important rumor as established movement.
Score evidence confidence with defined states
Use ordinal states rather than a mysterious decimal:
E0 — Fault or no evidence
The source failed, captures are incomparable, or the claim is unsupported.
E1 — Single ambiguous observation
A real public difference exists, but baseline, scope, availability, or interpretation remains unclear.
E2 — Direct comparable evidence
A primary source and healthy before/after record support the exact movement.
E3 — Corroborated movement
Multiple relevant sources or repeated observations support the same scoped conclusion.
Official pages should be matched to the claim. Cursor pricing can establish current public plan fields; Cursor status can establish the incidents and components that the provider disclosed. Neither supports conclusions about private motive or total customer experience.
Score materiality against a named decision
M0 — No current decision relevance
No identified segment, deal, assumption, product boundary, or risk is affected.
M1 — Context
Useful background or a weak leading indicator. No owner action required.
M2 — Review
Could change sales guidance, comparison language, a customer question, or a scheduled product review.
M3 — Decision
Could materially alter pricing, product scope, active deals, risk posture, resource allocation, or category strategy.
Require a sentence: “This is M2 because it affects [decision] for [segment] owned by [role].” Without that sentence, materiality is not scored.
Add urgency as a clock, not a third quality score
Urgency depends on decision half-life:
- active deal or public correction: hours to days;
- pricing or launch response: days;
- roadmap or positioning review: next operating cycle;
- category trend: monthly or quarterly synthesis.
A high-materiality item can have low urgency if the decision is distant. Do not confuse importance with interruption.
Route the matrix
| Evidence | Materiality | Route |
|---|---|---|
| E0–E1 | M0–M1 | reject, retain privately, or recheck |
| E2–E3 | M0–M1 | ledger or periodic context |
| E0–E1 | M2–M3 | evidence request with deadline; no public claim |
| E2 | M2 | owner review |
| E2–E3 | M3 | decision memo and named deadline |
Human reviewers can override routing, but they must record why. Overrides become training material for the rubric.
Use a scorecard that preserves reasons
# Signal review — [ID]
Observation:
Source / observed:
Before / after:
Comparability:
## Evidence confidence
State: E0 / E1 / E2 / E3
Reason:
Corroboration:
Unknown:
## Decision materiality
State: M0 / M1 / M2 / M3
Affected segment:
Named decision:
Owner:
Reason:
## Urgency and route
Decision half-life:
Due:
Route:
Override/reason:
Outcome:Calibrate against reviewed history
Monthly, sample:
- high-confidence items reviewers later reversed;
- high-materiality items that created no useful decision;
- rejected items later found material;
- source types with recurring false positives;
- score disagreements among reviewers;
- overrides that reveal missing rubric language.
Update definitions before adjusting thresholds. NIST's AI Risk Management Framework encourages explicit mapping, measurement, and management. W3C PROV-O reinforces the importance of retaining provenance across transformations. Together they support a reviewable system; neither validates your specific score.
Use the EXVIV company directory, market map, and Evidence Compare for public context; the monitoring guide for routing; and the AI evidence architecture for generated analysis boundaries.
Related EXVIV research
Sources and further reading
Method note
The states are governance aids, not statistical probabilities. Validate and calibrate them against reviewed outcomes in the actual operating context.