Snowflake Conditions

The Snowflake Native condition lets an ALTR Native classifier match a column against the output of Snowflake’s own classification engine, rather than against sampled values (a sampled subset; see Sampling) directly. It’s a condition type available to any Snowflake ALTR Native classifier, on either processing location.

Note

Snowflake Native classification as a standalone classification method is discontinued. It can no longer be selected when starting a classification job. Jobs that previously ran with it keep their results in the classification report, but no new jobs run using it. Snowflake’s classification output is available today only through the Snowflake Native condition described below.

Fields

The condition uses the categories Snowflake’s EXTRACT_SEMANTIC_CATEGORIES function produces, exposing 4 fields:

• Privacy category — one of Snowflake’s privacy categories: IDENTIFIER, QUASI_IDENTIFIER, or SENSITIVE.
• Semantic category — a Snowflake semantic category value, such as EMAIL, US_SSN, or PHONE_NUMBER. ALTR matches the value case-insensitively and doesn’t restrict the set, so the available values are whatever Snowflake’s classification returns.
• Minimum confidence — LOW, MEDIUM, or HIGH (default LOW). A detection below the chosen confidence doesn’t satisfy the condition. This is a separate gate from Match Threshold, below.
• Match Threshold — the share of sampled values that must carry the detection for the condition to pass, shown as a percentage in the UI (default 70%) and stored as a fraction from 0-1 in the API/JSON (minimum_coverage, default 0.70).

The condition’s own Minimum confidence and Match Threshold fields are the only gates it uses. The classifier-level Minimum Match Threshold and Sample size settings don’t apply to Snowflake Native conditions, since the condition matches against Snowflake’s classification output rather than evaluating sampled row data the way Regex, Data Length, Column Content, Google DLP, or Amazon Comprehend conditions do.

Example

This condition matches columns Snowflake identifies as an email address at high confidence, on at least 70% of sampled values:

{ "target": "SNOWFLAKE_NATIVE", "semantic_category": "EMAIL", "privacy_category": "IDENTIFIER", "minimum_confidence": "HIGH", "minimum_coverage": 0.70 }

Behavior

In Match Confidence scoring, a Snowflake Native match outranks a Regex, Column Name, Data Location, Data Length, or Column Size match on the same column, and is outranked only by Google DLP.

Availability

Because it depends on Snowflake’s classification function, the Snowflake Native condition evaluates only on Snowflake ALTR Native classification jobs, in both ALTR Hosted and In-Warehouse processing. It isn’t evaluated on OLTP, and it isn’t available on Databricks. See Classification Processing Location for the full picture.