Methodology · v3.4

CSCSentiment des consommateurs

Composite consumer sentiment for {legal_name}. EXTRACTION RULES: return type="composite_signal" with sub-signals, each 0-100 + source_url. Required: (a) review_aggregate_score — 0-100 trimmed median rating across whitelisted platforms (major map/local-business directories, ConsumerAffairs/G2/Capterra company pages, marketplace brand-storefront aggregates, mobile-app-store ratings of the company's flagship app) ×20. Single-product listings NOT acceptable; Trustpilot/BBB excluded for mega-brands due to complaint-magnet bias. (b) nps_proxy_score — 0-100 mapped from published NPS specifically for the COMPANY (CustomerGauge company page, consulting-firm case study, market-research company chart, IR/sustainability report; <-50→10, -50..0→30, 0..30→55, 30..60→75, 60+→90). (c) review_velocity_ratio — 0-100 from ratio of last-12mo reviews/month vs prior baseline (Trustpilot company history, mobile-app-store history via app-analytics platforms, Archive.org snapshot diff; <0.5→30, 0.5-1→55, 1-1.5→70, 1.5-2.5→85, 2.5+→95). (d) product_user_satisfaction — 0-100 trimmed median rating ×20 from product-level platforms when applicable (third-party customer-satisfaction studies, market-research satisfaction indices, Trustpilot product if not bias-skewed). Server applies composite_signal_score. CREDIBILITY BAR — every sub-signal MUST come from a real numeric rating / NPS / review-count datapoint on a named platform. DO NOT estimate or infer sub-signal values from qualitative articles, single stock-sentiment blog posts, generic "consumers like the brand" commentary, or a single source reused across all four sub-signals. If you cannot find at least the (a) review_aggregate_score from real platform ratings AND one other real numeric sub-signal for {legal_name}, return type="absent" with reason="not_disclosed" — many companies (B2B, industrial, premium/luxury, sports clubs) have no published multi-platform review/NPS numerics, so absence is expected and the server marks the index not_applicable. NEVER fabricate sub-scores to avoid an absent. PRESERVES methodology of legacy customer-reviews-aggregate, product-user-satisfaction, review-velocity, net-promoter-proxy.