Snowflake Horizon

Snowflake Horizon data catalog

Horizon catalogs Snowflake objects automatically, and Horizon Context (preview) harvests 18 external sources, including Databricks and BigQuery.

Compare: data catalog tools

• Data Catalog: Snowsight's Governance & security area provides centralized access to governance features over Snowflake objects, with object tagging and classification (docs).

• Automated Metadata Harvesting: Native objects plus Horizon connectors for 18 external sources (Preview) (docs).

• Search & Discovery: Universal Search in Snowsight finds tables, views, semantic views, agents, and other objects by name, comment, or tag, with results filtered by the active role's privileges; it searches metadata, not object contents (docs).

• AI-Powered Search: Universal Search accepts conversational natural-language queries, matches objects whose names differ from the search terms, and tolerates misspellings (docs).

• Usage & Popularity Statistics: Connectors collect popularity, and the "behavioral" context layer covers popularity and query patterns (docs).

• Custom Fields & Templates: Object tagging supports multi-value tags, plus native and custom classification categories (docs).

• Asset Certification: `SNOWFLAKE.TAGS.CERTIFICATION_STATUS` (CERTIFIED/DEPRECATED), GA 2026-10-20 (docs).

• Metadata Propagation: Tag inheritance propagates tags through the object hierarchy; tag-based masking policies apply protection wherever a tag is assigned (docs).

• Catalog Usage Analytics: No catalog adoption dashboard is documented.

• Business Glossary: No glossary module. Business definitions live in semantic views (docs).

Snowflake Horizon data lineage

Horizon captures lineage for Snowflake objects and accepts OpenLineage events for external lineage.

Compare: data lineage tools

• Data Lineage: Data lineage tracks data flow relationships between objects (docs).

• Table-Level Lineage: Lineage tracks relationships between tables and views (docs).

• Column-Level Lineage: Lineage works at object and column level; column lineage shows upstream and downstream columns with their distance and is retained for one year (docs).

• End-to-End Lineage: Lineage from PostgreSQL through dbt and Snowflake to Tableau dashboards is stitched into one column-level graph (Preview) (docs).

• Impact Analysis: Object Dependencies audits object references via metadata, surfacing what depends on a given object (docs).

• OpenLineage Support: External lineage accepts OpenLineage-compatible events through a REST endpoint, with documented setup for dbt (dbt-ol) and Airflow 2.7+; column mappings are captured best-effort (docs).

Snowflake Horizon data stewardship

• Review & Approval Workflows: Request approval workflow for organizational listings, and Intent-Driven Governance gates changes behind review (narrow) (docs).

• Metadata Automation: Classification profiles auto-tag new objects on a schedule, and Autopilot manages timing (Preview) (docs).

• Data Stewardship: Not yet researched.

• Governance Campaigns: Not documented.

• Critical Data Elements: Not documented.

• Governance Scorecards: The Trust Center scores security posture, not governance maturity (docs).

Snowflake Horizon data classification

Compare: data classification tools

• Data Classification: Sensitive data classification categorizes personal and sensitive data using native and custom categories (docs).

• Sensitive Data Discovery: Sensitive data classification identifies personal and sensitive data for compliance and discovery (docs).

• Data Policy: Column-level security via masking policies, row-level security via row access policies, and tag-based masking policies (docs).

• Data Masking: Masking policies applied to columns, including tag-based masking policies, and row access policies on tables and views (docs).

• Native Policy Enforcement: Horizon enforces policy within Snowflake, the platform it is part of. Under this row's definition that is a native governance layer governing its own platform, not cross-platform push-down into systems the tool does not operate (docs).

Snowflake Horizon AI features

Cortex agents and Snowflake CoWork use Horizon context, and AI-mode classification tags sensitive data.

• AI Chat Assistant: Snowflake CoCo uses catalog context to write SQL, trace lineage and review governance posture (docs).

• AI Documentation Generation: Cortex generates table, view and column descriptions (docs).

• Agentic Metadata Enrichment: Classification profiles with AI mode (Preview) detect and tag sensitive columns without per-asset review (docs).

• AI Features: Not yet researched.

Snowflake Horizon data compliance

• Data Compliance: Access History audits user access through Account Usage views; Object Dependencies audits object references (docs).

• Audit Trail: Access History provides auditing of user access to data through Account Usage views (docs).

• Compliance Reporting: Trust Center Sensitive Data Entitlement and Access reports (GA 2026-10-20) and account posture PDF reports (Preview) (docs).

• Data Activity Monitoring: ACCESS_HISTORY, the Sensitive Data Access report and Trust Center detection findings for anomalous events (GA 2026-04-29) (docs).

• Data Retention Policies: Storage lifecycle policies archive or expire rows by condition "to help meet compliance requirements" (GA 2025-11-07) (docs).

• Regulatory Framework Mapping: Classification has privacy categories, but no regulation-to-asset mapping is documented.

• ROPA: Not documented.

• Data Subject Rights: Not documented.

Snowflake Horizon collaboration

• Alert Notifications: DQ incident notifications to email, Slack, Teams and PagerDuty (Preview) (docs).

• Collaboration: Not yet researched.

• Discussions & Annotations: Not documented.

• Ticketing Integrations: Not documented.

Snowflake Horizon data marketplace

Compare: data marketplace tools

• Data Marketplace: Snowflake Marketplace lets providers publish free and paid listings and consumers explore and access them, with live data sharing and no data copies (docs).

• Self-Service Access Requests: Request approval workflow for Internal Marketplace organizational listings: consumers request access in Snowsight (or via an external email/URL), and roles with OWNERSHIP or MODIFY on the listing approve or deny with comments (docs).

• Data Products: Listings are the unit of published data product, including organizational listings and Data Exchange listings (docs).

• Data Mesh Support: Organizational listings support domain sharing, but no domain model is documented.

• Subscription Management: Not documented for internal listings.

Snowflake Horizon data quality

• Data Quality: Data Quality Monitoring with system and user-defined data metric functions, data profiles, and anomaly detection (docs).

• SQL Table Tests: User-defined data metric functions run SQL-defined checks against tables to monitor state and integrity (docs).

• Anomaly Detection: Anomaly detection is documented as part of data quality monitoring (docs).

• Data Profiling: Data profiles are part of data quality monitoring (docs).

• Data Health Scores: The DQ dashboard shows a healthy-table % (Preview), not per-asset composite scores (docs).

• Data Contracts: Not documented.

Snowflake Horizon AI governance

• AI Governance: Horizon Catalog adds an AI governance and security layer — AI guardrails, agent policies and MCP management from a centralized dashboard (website).

• AI Asset Inventory: The Horizon catalog UI surfaces AI assets alongside Snowflake objects, external tables and dashboards, and models are registered with their versions and metadata in the Snowflake Model Registry (docs).

• AI Lineage: Lineage covers ML objects (datasets, feature views, models, model monitors), and Cortex Agents appear as downstream consumers of semantic views and Cortex Search services (docs).

• AI Agent Access Governance: Horizon enforces AI guardrails and agent policies centrally, governing what AI agents may access at query time rather than only exposing an MCP endpoint (website).

• AI Regulatory Compliance: No EU AI Act, NIST AI RMF or ISO 42001 assessment templates or evidence workflows documented for Horizon (website).

• AI Model Health Score: No composite per-model trust or health score (website).

Snowflake Horizon workflows

• REST / GraphQL API: Not documented (docs).

• MCP Server: Snowflake-managed MCP server GA 2025-11-04 (docs).

• dbt Integration: Horizon dbt connector collects model definitions and lineage. External lineage also accepts dbt-ol (Preview) (docs).

• Python SDK: Snowflake Python API (docs).

• Workflows: Not yet researched.

• Webhooks: Webhook notification integrations exist, but catalog-event webhooks aren't documented (docs).

• Browser Extension: Not documented.

Snowflake Horizon data context layer

• Semantic Models: Semantic views define metrics, dimensions and relationships, built in Semantic Studio (GA 2026-09-30) (docs).

• Data Context Layer: Not yet researched.

• Knowledge Graph: Not documented as a property graph.