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.

Important: Operator Cluster and Related-site Hypothesis are probabilistic research concepts. They do not establish personal identity, beneficial ownership, legal control, or wrongdoing.