Research method
Evidence hierarchy, explicit uncertainty and deterministic publishing are part of the research product, not post-production decoration.
Evidence before narrative
Primary and official sources are preferred for quantities, policy decisions and institutional facts. High-quality secondary reporting is used for events, context and triangulation.
Factual observation, inference and scenario analysis are separated whenever the distinction materially changes the reader's interpretation.
Causal structure
Research is organized around transmission mechanisms, constraints and second-order effects. Cross-asset and geopolitical analysis should explain how a shock travels rather than merely list simultaneous events.
Publishing pipeline
Research source files are validated before rendering. HTML, search metadata, structured data, author pages, topic pages, feeds and sitemaps are generated by the build rather than written independently by an agent.
Visual evidence
Tables remain the audit layer. Charts, causal flows, dependency maps, regional maps and mind maps may be added when they improve comprehension. Visuals must be derived from the report's own data or explicit analytical structure.
Revision and uncertainty
Publication time, revision time and confidence are preserved when available. Recoverable formatting errors are normalized automatically; missing evidence or missing source files block publication.