- Home
- Model registries
- Snowflake Feature Store
Snowflake Feature Store
Menlo Park, California, United States
What it is
Snowflake Feature Store lets data scientists and machine-learning engineers create, maintain, and use ML features within Snowflake. It works with batch and streaming data, updating features as new data arrives, and supports backfill and point-in-time-correct features through ASOF JOIN. Transformations can be written in Python or SQL. Snowflake-managed feature views refresh incrementally on a schedule, while external views are maintained by a separate process such as dbt. The service integrates with Snowflake Model Registry and supports user-managed dbt pipelines. ML Lineage can trace the path from source data through features and datasets to a trained model, and is created automatically when the Feature Store is used. Snowsight includes a feature search and discovery interface, and the Python API can be used in a local IDE or Snowsight worksheet or notebook. Data remains under Snowflake governance with fine-grained role-based access control. Pricing is consumption-based, with no prices listed. Feature-view versions are limited to 128 characters; Postgres-backed online serving limits the combined feature-view name and version to 46 characters.
For and against
For
- Supports batch and streaming data with automatic updates.
- Creates point-in-time-correct features using ASOF JOIN.
- Integrates with Snowflake Model Registry and dbt pipelines.
- ML Lineage traces data through to trained models.
Against
- Feature-view versions have a 128-character maximum.
- Postgres-backed online serving has a 46-character combined name and version limit.
- Pricing is consumption-based, with no listed prices.
The verdict
Snowflake Feature Store provides feature creation, discovery, lineage, and governance within Snowflake. Its view-management options and integration with existing Snowflake ML tools are central considerations, alongside the stated naming limits and consumption-based pricing.
Who it suits
It is for data scientists and ML engineers building feature workflows within Snowflake, particularly those using Python or SQL and Snowflake Model Registry. Teams should check the name-length constraints if using Postgres-backed online serving.
Plans and prices
Business Critical
all Enterprise features · Tri-Secret Secure · private connectivity · failover and failback
Read 1 Oct 2026 · snowflake.com
Enterprise
all Standard features · multi-cluster compute · granular governance and privacy controls · extended Time Travel
Read 1 Oct 2026 · snowflake.com
Standard
core platform functionality · Snowpark · data sharing · optimized storage and Time Travel
Read 1 Oct 2026 · snowflake.com
Virtual Private Snowflake
all Business Critical features · completely separate isolated Snowflake environment
Read 1 Oct 2026 · snowflake.com
On the record
- API platforms
- The Snowflake Feature Store Python API is part of the snowflake-ml-python package and can be used in a local Python IDE or in a Snowsight worksheet or notebook.docs.snowflake.com · read 1 Oct 2026
- Compliance
- Snowflake lists global certifications and attestations including ISO 27001, ISO 27017, ISO 27018, SOC 1 Type II, and SOC 2 Type II.docs.snowflake.com · read 1 Oct 2026
- Cost model
- Snowflake-managed feature views use dynamic tables, while external feature views use views and incur no additional storage cost.docs.snowflake.com · read 1 Oct 2026
- Data support
- It supports batch and streaming data with automatic updates as new data arrives.docs.snowflake.com · read 1 Oct 2026
- Examples and support
- Snowflake quickstarts are examples only and are not guaranteed for accuracy or covered by a Snowflake Service Level Agreement.docs.snowflake.com · read 1 Oct 2026
- Feature views
- Snowflake-managed feature views refresh incrementally on a schedule, while external feature views are maintained by another process such as dbt.docs.snowflake.com · read 1 Oct 2026
- Integrations
- The Feature Store integrates with Snowflake Model Registry and supports user-managed pipelines with dbt.docs.snowflake.com · read 1 Oct 2026
- Limits
- Feature view versions have a maximum length of 128 characters, and Postgres-backed online serving limits combined feature-view name and version length to 46 characters.docs.snowflake.com · read 1 Oct 2026
- Lineage
- ML Lineage can trace data flow from source to feature to dataset to trained model and is automatically created when the Feature Store is used.docs.snowflake.com · read 1 Oct 2026
- Point-in-time features
- It supports backfill and point-in-time-correct features using ASOF JOIN.docs.snowflake.com · read 1 Oct 2026
- Purpose
- Snowflake Feature Store lets data scientists and ML engineers create, maintain, and use machine-learning features within Snowflake.docs.snowflake.com · read 1 Oct 2026
- Security
- Data remains under Snowflake governance and does not leave Snowflake, with access managed through fine-grained role-based access control.docs.snowflake.com · read 1 Oct 2026
- Transformations
- Feature transformations can be authored in Python or SQL.docs.snowflake.com · read 1 Oct 2026
- User interface
- Snowsight provides a Feature Store UI for searching and discovering features.docs.snowflake.com · read 1 Oct 2026
