Methodology
Evidence before conclusions.
RivalAtlas produces Website Relationship Intelligence from bounded, public observations.
Collection boundaries
Scans retrieve a limited set of public HTTP pages, JavaScript resources, response headers, DNS records, RDAP data, and public certificate transparency records. Requests use strict byte, page, redirect, port, and execution-time budgets. Robots directives and reasonable request pacing are respected.
Normalization
Raw observations are normalized into identifier families such as analytics, advertising, authentication, backend projects, support identities, payment catalogs, infrastructure, and public legal entities. Secret keys, cookies, login state, private accounts, access-controlled content, and write operations are out of scope.
Scoring
score = baseWeight × IDF × freshness × sourceReliability
Rare, durable identifiers receive more weight. Template defaults, public SaaS instances, shared services, and high-frequency identifiers are heavily discounted. Only the strongest observation in each identifier family contributes, preventing duplicate accumulation.
Confidence
A high-confidence Related-site Hypothesis requires at least two independent evidence families. A single GA, GTM, Clarity, DNS, or certificate signal never confirms a common operator. Cloudflare nameserver pairs carry only low supporting weight.
Ownership epochs
Domains can be sold or repurposed. Observations are attached to operating epochs so historical identifiers do not automatically describe the current operator.