Trust Score Methodology

# Trust Score Methodology Source: Qlooto Development Handoff Brief, sections 5 and 6, reproduced verbatim — the methodology_version shown next to every score on this platform corresponds to the formulas below. Two independent scores are computed, 0.0-10.0 to one decimal. A developer score is NEVER an average of its project scores — separate calculations over different inputs. ## Project score components - RERA compliance — 30% weight, never absent - Delivery record — 25% weight; pre-possession projects use the developer's historical pattern - Resident signal — 25% weight; excluded and the remaining weights re-normalised when absent - Pricing transparency — 20% weight; excluded and re-normalised when absent ## Developer score components - Corporate governance — 20% - RERA history across portfolio — 20% - Delivery pattern — 20% - Financial health — 20% - Post-possession conduct — 20% Missing components are excluded and re-normalised, never imputed. The interface states "score computed on N of M components" whenever a component is missing — filling a gap with an assumed value would be fabrication. ## Bands - 8.0-10.0 Sound record - 6.5-7.9 Caution - below 6.5 High risk The numeral and the word are the primary encoding; colour is third and is never the only signal. ## The firewall The scoring engine runs as a separate scheduled process, never inside a web request. Its database role holds SELECT on verified_data and community, INSERT on trust_engine, and no grant at all on the developer schema. No HTTP endpoint can trigger a score write. Every write emits a score_changes row recording the old value, the new value, which component moved, the source record ids, the methodology version, and the timestamp. ## Formulas ``` Information value: V(i) = S x A x D x F x R S = 1/(1+n) scarcity, n = others who already reported this A in [0.30..1.00] source authority (0.3 member .. 1.0 government) D in {0.2, 0.6, 1.0} 0.2 general, 0.6 answers an open question, 1.0 moves a score or contradicts declared data F = e^(-t/tau) tau = 90d for price, 365d for possession, infinite for legal R = log(1+w) reach, w = number of watchers Reciprocity (gates DERIVED intelligence only, never the record itself): R(u) = (0.3 x Passive + 1.0 x Active) / Consumption basic tier requires R >= 0.05, deep tier requires R >= 0.20 Connection matching: M(p,q) = P x C x Av x Fg Fg = 1/(1 + requests_last_7d) the fatigue term P: 10 same tower, 8 same project, 5 same developer, 3 same locality, 1 same segment ```