Targeted Classification

Classifying a Snowflake data source samples every table in its database, including the tables you already know hold nothing sensitive. Targeted classification narrows a job to one schema or table and skips objects you name inside it, so a job reads only the data you want classified.

Targeted classification is available for Snowflake data sources. Scan Scope appears in the classification dialog alongside the data source you select.

Scan Scope

Scan Scope sets where a classification job starts. Selecting Entire Database – the default – classifies the whole database the data source points at. Selecting A Specific Object narrows the job to a single schema, or to a single table within a schema, and adds two steps to the job configuration.

The Select An Object step is where you name that object. Select a schema to classify it and everything in it; select a table within that schema to classify only that table. Leave both empty to classify the entire database with the Exclude Objects step still available – that’s how you exclude whole schemas from a database-wide job.

A job has one scope. To classify two schemas, run two jobs. The database comes from the data source itself, so a scope narrows within that database and never reaches another one.

Excluded Objects

Everything inside the scan scope is classified unless you exclude it. The Exclude Objects step lists the objects one level below the scope you selected: schemas when the job is scoped to the database, tables when it’s scoped to a schema. A job scoped to a table has nothing left to exclude, so the step offers nothing to select and you continue past it.

Excluded objects are removed before sampling begins, so ALTR never reads them. This is different from filtering results out of a finished report – data in an excluded object is never sampled in the first place. A job can exclude up to 100 objects.

Run a Targeted Classification Job

To narrow a classification job to one object:

1. Select Data Classification > Classification Reports in the navigation menu.
2. Click Classify Data. ALTR displays a dialog to configure the job.
3. Select a Snowflake connection type and data source.
4. Under Scan Scope, select A Specific Object. ALTR adds the Select An Object and Exclude Objects steps to the job configuration.
5. On Select An Object, select the schema to classify, and a table within it if you want to narrow the job further. Leave both empty to classify the entire database and exclude schemas from it on the next step.
6. On Exclude Objects, select any object in scope that shouldn’t be classified. Objects you select appear as a list you can remove entries from.
7. Configure the rest of the job – classification method, collection, processing location, Human-in-the-Loop Review, and sample size – as described in Data Classification.
8. On Review details, check the job’s scope level and its excluded objects, then click Classify Data. ALTR runs the job against the scope you selected.

Scan Scope on a Report

A classification report states the scope its job ran at. The Scan Scope banner above a report and its findings names the scoped object in full and the level it ran at. A job run without a targeted scope shows the data source’s own name.

When a job excluded objects, the banner also states the number of exclusions, with a View link that opens Excluded Objects – the excluded objects grouped by the object that contains them, with a search box. On a completed job, the count covers the exclusions that matched an object in the data source; an exclusion that matched nothing isn’t counted or listed.

The report list carries a Scan Scope column stating the scope each report ran at, so a targeted job is distinguishable from a whole-database one without opening it. Individual runs inside an expanded row don’t repeat the scope.

Targeted Jobs and Report History

A targeted job keeps its own report history. ALTR groups reports by scope alongside the data source, collection, and classification method, so a schema-scoped job and a whole-database job on one data source are two rows in the report list, each with its own runs. Jobs run at the same scope are runs of the same report.

Re-running a targeted job from its report starts the configuration with the original scope and its excluded objects already selected, including any that matched nothing when the job ran.

Caution

If a re-run can’t load the original job’s excluded objects, Review details states that instead of listing them, and the re-run classifies those objects. Select them again on Exclude Objects before starting the job.

Decisions on a Targeted Job

A targeted job you run with Human-in-the-Loop Review opens its findings tree at the scope it ran at, and its decisions carry forward by object rather than by scope. See Decisions on Later Jobs for how that match is made and when findings come back as new.