Summary

Precon sizing evidence methodology explains how documents become sizing claims. The platform indexes evidence, selects relevant pages and sections, extracts signals, applies precedence, synthesizes claims, and validates quality before the estimator workspace treats the result as useful.

Reader Question

How do documents, precedence, OCR/extraction, candidate evidence, source quality, and synthesis produce sizing claims?

Methodology

The method starts with document indexing and eligibility. Selection services rank pages, sections, and documents by relevance and budget. Inference extracts structured signals with prompts and parsing contracts. Precedence logic then decides which evidence should win when sources conflict.

Synthesis composes selected evidence into sizing outputs, while guardrails, citations, and quality services preserve traceability. Preview and reporting jobs let engineers evaluate the methodology without mutating production-facing state.

Boundaries

Indexing, selection, inference, synthesis, precedence, guardrails, citations, and quality are separate packages because each answers a different reasoning question. The estimator workspace consumes the read model; it does not own the evidence methodology.

Tradeoffs

The sizing method is more complex than summarizing every document. The benefit is bounded context, inspectable evidence, and explicit conflict handling.