Methodology · v2.12

PQIProduktqualität

Expert review aggregates for {legal_name}'s OWN FLAGSHIP PRODUCTS. EXTRACTION RULES: Return rating_aggregate with one entry per outlet (each: platform, score, scale, source_url, review_count). CATEGORY-APPROPRIATE OUTLET SELECTION (agnostic, sector-adaptive): pick the TIER-1 specialist expert outlets RELEVANT TO THE PRODUCT CATEGORY — do NOT restrict to tech press. Examples by category (illustrative, not exhaustive): tech/electronics → Consumer Reports, Wirecutter, The Verge, Tom's Guide, DXOMARK, PCMag; fashion/apparel/luxury → Business of Fashion, Vogue Business, GQ, Highsnobiety, Hypebeast; watches → Hodinkee, WatchTime; beverages/food → Wine Spectator, Decanter, sector tasting panels; design/furniture → Architectural Digest, Dezeen; beauty/cosmetics → Allure, expert dermatology testers; music/media → Pitchfork, specialist critics. Each rating MUST be that outlet's direct expert review of one of {legal_name}'s OWN flagship products. QUALITATIVE-VERDICT → NORMALIZED RATING (when an outlet gives an editorial verdict WITHOUT a numeric score): map the documented verdict to a normalized 0-100 tier using this rubric — editor's-pick / best-in-class / award → 85-90, recommended / positive → 75-80, mixed → 60, negative / pan → 35 — and emit it as the entry's score on a 0-100 scale (set scale=100). ANTI-FABRICATION: a qualitative verdict counts ONLY when ≥2 INDEPENDENT specialist outlets converge on the same direction; cite each outlet's URL. SUBJECT VALIDATION: the source_url MUST be a review of {legal_name}'s product, NOT a review of a THIRD-PARTY content app that happens to appear on {legal_name}'s app store. Specifically forbidden: ratings of any third-party publisher's mobile app or web app distributed through {legal_name}'s own OS / app-store / marketplace surface — those ratings measure the app, not the company's product. If fewer than 2 TRUE expert-review aggregates (numeric OR ≥2-convergent qualitative) can be sourced on the outlets' own domains, RETURN type="absent" with reason="no_outlet_reviews_for_flagship_products" instead of substituting OS app-store reviews of unrelated third-party apps. For numeric outlets use each outlet's native published expert rating on its native scale (e.g., Tom's Guide 1-10, DXOMARK 0-200, PCMag editor rating 1-5); do NOT translate a numeric score. Server takes the trimmed, log-review-count weighted median × 20 — DO NOT compute any 0-100 score.