Identity Resolution Platform

Combines exact-match logic (hashed emails, login credentials) with statistical inference (IP address, browser fingerprint, behavioral patterns) to maximize match rates. Deterministic rules provide absolute certainty for authenticated users, while probabilistic models assign confidence scores to anonymous sessions. The engine applies Fellegi-Sunter record linkage to calculate match weights, classifying record pairs as matches, non-matches, or requiring review. This hybrid approach ensures both precision and recall, linking known customers with high fidelity while inferring anonymous device ownership to fill gaps in the customer journey.

A specialized database optimized for storing and traversing many-to-many relationships between identifiers, devices, and households. Unlike relational databases, the graph natively models complex connections—a single canonical ID links to multiple email addresses, device IDs, and partner identifiers. The layer supports temporal edge weighting, where the strength of a connection decays over time if not re-validated, preventing stale identifiers from polluting profiles. Built on graph databases or wide-column stores, it enables millisecond-latency lookups for real-time personalization engines.

Converts raw PII into irreversible cryptographic tokens before matching. Email addresses and phone numbers are hashed using SHA-256 or HMAC with secret salts, creating match keys that are consistent across systems but cannot be reversed to reveal the original identifier. The platform integrates with frameworks like Unified ID 2.0 (UID2) and RampID, generating interoperable tokens for authenticated addressability in cookieless environments. All match key operations occur in a data clean room or isolated enclave, ensuring raw PII never leaves the brand's control.

An event-driven stream processor that ingests raw clickstream and app events, then algorithmically connects discrete sessions into a continuous behavioral journey. The pipeline handles session timeouts, device switches, and anonymous-to-known transitions. Using Apache Kafka or Amazon Kinesis for ingestion and Apache Flink for stateful stream processing, it assigns a temporary session ID that is later merged into the persistent identity graph. This enables mid-session personalization—a user browsing anonymously on mobile can be recognized immediately after logging in on desktop.

Embeds privacy preferences directly into the identity resolution workflow. Before linking identifiers, the platform checks the Consent Management Platform (CMP) for the user's granular opt-in or opt-out status. It respects Global Privacy Control (GPC) signals and enforces purpose limitation—an identifier collected for analytics may not be used for advertising without separate consent. The system maintains an immutable audit trail of consent states, ensuring that identity stitching is legally defensible under GDPR, CCPA, and evolving state privacy laws.

Applies configurable survivorship rules to resolve conflicting attribute values when merging multiple source records into a single golden record. Rules prioritize data freshness, source trustworthiness, and completeness. For example, a mobile app's last-known location may override a stale CRM address, while a verified billing address always trumps a self-reported one. The engine outputs a canonical ID with the best-version-of-truth attributes, and retains all source values in a history ledger for auditability and downstream system reconciliation.

A centralized data structure that links all known identifiers—such as email addresses, device IDs, and usernames—to a single unified customer profile. The graph forms the backbone of cross-device personalization by maintaining edges between nodes that represent different identity signals.

• Stores deterministic links (hashed email, login credentials)
• Stores probabilistic links (IP address, browser fingerprint)
• Enables real-time lookup during online inference

A method of identity resolution that relies on exact, verified matches of personally identifiable information (PII) to link user activity across devices with absolute certainty.

• Uses hashed email keys or login credentials as anchors
• Produces 100% confidence match scores
• Requires authenticated touchpoints to establish linkage
• Forms the high-trust foundation of any identity spine

A statistical approach that uses non-personal signals like IP address, browser type, operating system, and behavioral patterns to infer device ownership. Each match is assigned a confidence score rather than a definitive link.

• Employs Fellegi-Sunter models for weight calculation
• Critical for linking anonymous sessions to known profiles
• Scores decay over time via identity decay models

A secure, neutral environment where multiple parties can combine and analyze first-party data sets for identity resolution without exposing raw, user-level data to external stakeholders.

• Enables second-party data matching with retail partners
• Applies differential privacy and k-anonymity constraints
• Critical for walled-garden identity resolution (Google, Amazon)

A marketer-managed system that aggregates first-party data from multiple sources to build a unified, persistent customer database. The CDP consumes the canonical IDs produced by the Identity Resolution Platform and makes them accessible to engagement tools.

• Ingests golden records from the identity spine
• Syndicates segments to email, advertising, and personalization engines
• Maintains real-time profile updates from streaming pipelines

A technology that captures, stores, and syndicates a user's granular privacy choices to downstream vendors. The CMP ensures that identity resolution logic respects the specified legal basis for processing.

• Propagates Global Privacy Control (GPC) signals
• Integrates with IAB Transparency & Consent Framework
• Determines which identifiers are eligible for matching
• Prevents unauthorized stitching of opted-out profiles