Methodology · v2.10

PSTTransparence des spécifications

Spec-doc completeness assessment for {legal_name}'s flagship product line. EXTRACTION RULES: return type="composite_signal" with one entry per dimension, each 0-100 value + a NUMERIC weight + source_url. SECTOR-AWARE WEIGHTING (post-#592 audit) — assign weights based on the company's primary product category, since the relative importance of these dimensions differs materially by sector: (1) B2C consumer electronics / mobile / wearables: full_spec_table weight=2.0, downloadable_datasheets=1.0, machine_readable_format=0.2 (largely irrelevant for consumers), support_repair_docs=1.5, firmware_driver_archive=1.0. (2) B2B enterprise hardware / networking / industrial: full_spec_table=1.5, downloadable_datasheets=2.0, machine_readable_format=2.0 (critical for integrators), support_repair_docs=1.0, firmware_driver_archive=1.5. (3) Software / SaaS: full_spec_table=1.5, downloadable_datasheets=1.0, machine_readable_format=1.5, support_repair_docs=0.5 (N/A), firmware_driver_archive=0.3. (4) FMCG / retail / fashion: full_spec_table=1.0, downloadable_datasheets=0.5, machine_readable_format=0.1, support_repair_docs=0.5, firmware_driver_archive=0.1. Required sub-dimensions: (a) full_spec_table = how complete the per-SKU spec table is on the official product page; (b) downloadable_datasheets = availability + completeness of datasheet PDFs; (c) machine_readable_format = presence of XML / JSON / structured spec endpoints; (d) support_repair_docs = hardware support / repair documentation breadth (iFixit, official repair manuals); (e) firmware_driver_archive = depth of firmware / driver / manual archive across product generations. Each signals[].value MUST be a real assessment based on cited evidence (NOT 0 placeholder). State which sector category was used in _reasoning. Server applies composite_signal_score (weighted mean using the LLM-emitted weights).