Infinite AudienceIdentity Resolution & Enrichment

Resolve fragmented consumer records into persistent, deduplicated identities using a hybrid matching engine, featuring deterministic graph traversal for exact matches and vector embeddings for probabilistic fuzzy matching, and then enrich every identity with 525+ attributes across 12 categories.

0B+

Identifier Nodes

0M+

Resolved Identities

0+

Enrichment Attributes

🔍 Click to Expand

Graph Architecture

A Graph Database Built for Identity

Unlike traditional identity solutions that rely on flat table joins and key-value lookups, the Infinite Audience Graph stores consumer identities as a true property graph. Nodes represent distinct entities (people, households, addresses, buildings, emails, phones), and typed edges encode the relationships between them.

This structure enables multi-hop traversals that traditional databases can't perform efficiently: starting from a single email, the engine can walk edges to find the person, then traverse to their household, discover co-residents, and resolve address history within in a single graph query.

Hybrid Matching Engine

Two Phases. Maximum Match Rates.

Every record passes through a two-phase hybrid resolution pipeline that combines the precision of deterministic graph matching with the reach of vector-based probabilistic matching to produce maximum match rates and persistent, deduplicated identities.

⬡

Phase 01

Deterministic Graph Matching

Graph Database · Exact Match

The first pass traverses the identity graph using exact, verified identifiers. Names, postal addresses, emails, and phone numbers are resolved via direct edge traversal, not table joins. Each node in the graph represents a distinct identifier (a name, an address, an email), and edges encode known relationships between them (LIVES_AT, HAS_EMAIL, HAS_PHONE, MEMBER_OF). When your record's identifiers match existing graph nodes, the engine walks the edges to find the canonical person identity, producing high-confidence, deterministic linkages.

Direct node-to-node traversal via graph edges — O(1) lookup per identifier
Address-anchored resolution using USPS-normalized postal addresses
Multi-hop edge walking: email → person → household → address
Deduplication via persistent iag_person_id across all source systems

◈

Phase 02

Probabilistic Fuzzy Matching

Vector Search · Embeddings

Records that don't produce an exact graph match enter the probabilistic layer. Here, each unmatched record is encoded into a high-dimensional vector embedding that captures its semantic identity signature — combining phonetic encodings and contextual signals. These embeddings are searched against the graph's vector index using approximate nearest-neighbor (ANN) search with Tree-AH, surfacing candidate matches ranked by cosine similarity. This process catches typos, name variations, maiden names, address formatting differences, and transliterations that deterministic matching would miss.

Semantic vector embeddings encode name + address + contextual features
Approximate Nearest Neighbor search via Tree-AH algorithm
Cosine similarity scoring with configurable confidence thresholds
Phonetic encodings (Soundex, Double Metaphone) as embedding features

Infinite Audience Graph Categories

Identity

Resolved person and household identifiers, contact information, and physical address derived from cross-source entity resolution

Ex.

IAG Person ID

Demographics & Household

Individual and household-level demographic attributes describing life stage, family composition, and generational cohort

Ex.

Gender

Financial & Property

Estimated financial capacity, property ownership characteristics, and professional profile indicators

Ex.

Estimated HH Income

Media & Behavioral

Modeled digital behavior, media consumption preferences, and lifestyle engagement signals

Ex.

Active Social Networks

Geography

Geographic boundaries and geography-based classifications

Ex.

Census Block

Land Context

Physical environment surrounding the address — land-use classifications, adjacency to natural and built features, and distance to major infrastructure

Ex.

Lives On Military Base

Census & Socioeconomic

Population characteristics from the American Community Survey covering income distribution, educational attainment, housing stock, commuting patterns, and labor force composition

Ex.

Total Population

Health

Community-level health outcomes and risk factors, including chronic disease prevalence, preventive care utilization, and disability rates

Ex.

Lack of Health Insurance

Environment & Risk

Natural hazard exposure and energy vulnerability metrics from federal risk assessment programs

Ex.

Energy Burden

Economic Indicators

Regional economic conditions — employment dynamics, public safety indicators, housing market trends, and tax-reported income distributions

Ex.

Unemployment Rate

Brand Proximity

Physical proximity to tracked anchor brands — continuous distance measurements and an array tracking nearby anchor brands

Ex.

Distance to Whole Foods

Neighborhood Lifestyle

Neighborhood personality profiling through point-of-interest density and concentration indexes across 20 lifestyle themes

Ex.

Wellness Hub Density

🔗

See the Full Graph

Explore the full graph architecture to research all attributes available for enrichment.

Explore the Graph →

Processing Modes

Batch or Real-Time. Same Graph. Same Match Quality.

Whether you need to resolve and enrich millions of records in a single batch operation or process individual records on-demand via API, both modes utilize the same hybrid matching engine to deliver the exact same depth of enrichment.

📊

Batch Resolution & Enrichment

Sync your first-party data, including names, addresses, emails, phone numbers, or pre-hashed identifiers, to resolve an entire file against the graph in a single batch operation. The pipeline enriches every matched record with your chosen attributes.

LatencyMinutes to hours

VolumeMillions of records

InterfaceREST API, MCP, Agents

Use CaseCampaign builds, CRM enrichment

⚡

Real-Time Resolution & Enrichment

Resolve and enrich individual records on-demand via the REST API or MCP. An endpoint or tool handles identity resolution against the graph, enriches the data with selected attributes, and returns the full profile in a single API response. This setup is ideal for point-of-interaction enrichment, real-time personalization, and event-driven workflows.

LatencySub-second

VolumeIndividual records or microbatches

InterfaceREST API, MCP, Agents

Use CaseReal-time personalization

Privacy-First Architecture

Decentralized Clean Rooms. Zero-Copy Data Collaboration.

We offer decentralized clean room architecture to enable seamless data joins without ever moving, copying, or exposing your raw data. By using cryptographic tokenization and zero-copy sharing protocols, your first-party datasets can be matched against the identity graph in a completely isolated environment, ensuring both parties retain absolute control of their proprietary records. This setup guarantees maximum privacy and security, completely eliminating the need for data transfer or intermediary exposure.

☁️

Zero-Copy Sharing

Data remains natively inside your storage bucket or database. No records are ever copied, transferred, or exposed to third parties to execute identity matching.

🔒

Cryptographic Hashing

Identities are resolved via secure, one-way SHA-256 hash matching of tokenized records, keeping raw customer PII fully protected during traversal.

📊

Differential Privacy

Applies mathematical privacy frameworks to query outputs, blocking single-user re-identification and preventing database inference attacks.

🛡️

Aggregation Control

Enforces strict minimum thresholds of 50 matching records to completely block single-consumer targeting and profile scraping.

📋

Certified Templates

Restricts matching operations to pre-approved analysis templates, blocking arbitrary SQL execution and preventing data exfiltration.

📝

Immutable Audit Logs

Generates detailed, cryptographically signed transaction logs tracking every single join and matching run automatically.