Every useful WALPHA output should be answerable with three questions: what was observed, when was it observed, and where did it come from?
Evidence classes
Canonical source evidence
Original posts, thread context, authorship, timestamps and media references. This is the primary social record.
Contextual evidence
Token metadata, market observations, wallet labels and other typed enrichments associated with a source event or entity.
Attributed perspectives
Public support, criticism, warnings, claims and disagreements attached to the account and source that expressed them. WALPHA preserves these perspectives as evidence; it does not convert them into unattributed facts or inferred private opinions.
Derived measurements
Counts, rates, independent author breadth, momentum, attention share, relationship strength and causal event-time outcomes computed from bounded windows.
Model findings
Interpretations, narrative states and hypotheses generated from compact evidence packs. They retain a report reference, version and lineage and never overwrite canonical facts.
Time and capture context
Market context can be captured close to the source event or reconstructed from historical data. The interface labels that distinction because present-day market state must not be mistaken for event-time state.
Source manifests
Research outputs carry an input window and source manifest. This makes a report reproducible at the level that matters: which evidence was eligible, which entities were resolved and which enrichments were available.
Corrections and review
New metadata or reviewed relationship proposals can supersede an earlier resolution. Prior evidence remains auditable; the current product view reflects the active resolution state rather than deleting history.