Methodology · v2.7

BLIBrand Loyalty

Customer retention / loyalty rate for {legal_name} (12-month, % retention-equivalent). ACCEPTED EVIDENCE (any ONE suffices; these published {legal_name}-specific loyalty metrics ARE the official extraction method — when you cite a figure that NAMES {legal_name} do NOT return type="rejected_peer_substitute"): (1) published 12-month customer-retention or subscription-renewal rate %; (2) repeat-purchase / repeat-customer rate % (use directly as a retention-equivalent); (3) 100 − published churn rate % (convert churn to retention); (4) analyst / market-research retention estimates that explicitly name {legal_name} (CIRP brand-retention surveys, CustomerGauge, SQ Magazine retention benchmarks, Bain loyalty studies, company investor decks). SCOPE GATE (vincolante, anti-inflation): the figure MUST be a COMPANY-WIDE retention covering the TOTAL customer base. REJECT sub-segment / cohort / tier-limited retention claims like "100% retention rate per clienti multi-unit", "retention of multi-unit accounts", "top-tier / strategic / enterprise / premium customer retention", "retention among the top X% of clients", or any figure scoped to a slice of the customer base — these are NOT comparable to a company-wide retention. If the only retention figure available is sub-segment-scoped and no company-wide figure is published, return type="absent" reason="not_disclosed" (do NOT use the sub-segment number as a proxy for the whole company). CREDIBILITY BAR — the figure MUST be an explicit numeric percentage (or churn % convertible to one) from the company's own IR/annual report OR a named reputable research/analyst source. REJECT anecdotal or qualitative material: marketing blogs, case-study videos, "retention doubled / improved" statements with no stated base percentage, vendor testimonials. A relative claim without an absolute % is NOT a value. If multiple credible sources disagree, use the median. Return exact_numerical{value: <retention_equivalent_pct_0_100>, unit: "percent", source_urls: [...]}. Server applies ratio_band (<50 → 30, 50-70 → 55, 70-85 → 75, 85+ → 90) — DO NOT compute any 0-100 score. If NO credible published retention / repeat-purchase / churn figure for {legal_name} exists, return type="absent" with reason="not_disclosed" — most premium / B2C / industrial companies do NOT publish this metric, so absence is expected and the server marks the index not_applicable. NEVER invent a number from generic NPS sentiment, qualitative claims, or peer/industry averages.