Operational guide · Data lifecycle

From a known input to a governed analytical result.

Use this workflow when a recurring analytical result must retain known input, quality, version, permission and execution context across separate canonical modules.

Use this workflow when a recurring analytical result must retain known input, quality, version, permission and execution context across separate canonical modules.
Decision question

Can the team identify the exact input, accepted quality outcome, Dataset contract, Report contract and execution record behind the result—or is one of those hand-offs still informal?

Objects and controls used

The workflow uses objects owned by different modules. Their references connect the story; they are not collapsed into one synthetic run.

Operational flow

Perform each transition explicitly and retain the identity returned by the responsible module.

Failure and partial paths

Failure is handled by the module that owns the attempted action; a later screen must not overwrite or reinterpret that terminal state.

Evidence produced

Depending on the chosen path and retention bounds, the flow can retain FileVersion and transform identity, QualityRun, DatasetVersion, ReportVersion, AnalyticsRun, diagnostics, artifact hashes and the canonical references projected into Assurance.

Current boundary

This is an operator journey, not one automatic workflow. Pipeline does not create a Dataset automatically, Schedule does not currently dispatch runs, a virtual DatasetVersion does not freeze live rows, and there is no universal rollback or exactly-once guarantee.

Decision question

Can the team identify the exact input, accepted quality outcome, Dataset contract, Report contract and execution record behind the result—or is one of those hand-offs still informal?

Objects and controls used

The workflow uses objects owned by different modules. Their references connect the story; they are not collapsed into one synthetic run.

Input identity

Project-scoped Source or immutable FileVersion with server-owned tenant and Project scope.

Publication and quality

Optional TransformationRun and TableVersion, followed by a source-oriented QualityRule and persisted QualityRun.

Governed data contract

Dataset and exact DatasetVersion with source/query lineage and recorded column metadata.

Analytical contract

ReportVersion bound to that DatasetVersion and an AnalyticsRun under current permission, RLS and runtime-filter context.

Operational flow

Perform each transition explicitly and retain the identity returned by the responsible module.

  1. Register the input

    Select the project Source, or upload/import a file so an immutable FileVersion and SHA-256 identity exist before interpretation.

  2. Inspect before execution

    Review Source metadata or file parser, delimiter, header, schema preview and warnings; do not queue a file transform while interpretation is unresolved.

  3. Publish a managed table when needed

    Load into a shadow physical version, validate row/schema results and promote the stable Sandbox alias only after success.

  4. Run the quality decision

    Execute the accepted assertion and retain its QualityRun; a preview is useful feedback but is not the persistent run record.

  5. Promote the governed Dataset

    Accept the SQL or table lineage as a stable Dataset identity and exact immutable DatasetVersion.

  6. Accept the analytical contract

    Select dimensions and metrics, validate metadata, compile server-owned read-only SQL and save a ReportVersion bound to the exact DatasetVersion.

  7. Execute and inspect

    Run the report and inspect AnalyticsRun status, bindings, filters, permission/RLS context, diagnostics, artifacts and truncation before using the result.

Failure and partial paths

Failure is handled by the module that owns the attempted action; a later screen must not overwrite or reinterpret that terminal state.

Preflight or transform fails

No new alias version is promoted; the prior stable table remains current and the attempted TransformationRun stays inspectable.

Quality fails

The QualityRun and bounded finding persist. A supported Master gate may stop downstream work, but no universal automated remediation follows.

One Analytics binding fails

Per-binding diagnostics remain visible while successful bindings can retain their own output; the overall run remains partial where applicable.

Monitoring disconnects

Reconnect or poll the canonical run. A lost browser stream is not proof that a queued state-changing operation never reached its worker or target.

Evidence produced

Depending on the chosen path and retention bounds, the flow can retain FileVersion and transform identity, QualityRun, DatasetVersion, ReportVersion, AnalyticsRun, diagnostics, artifact hashes and the canonical references projected into Assurance.

Current boundary

This is an operator journey, not one automatic workflow. Pipeline does not create a Dataset automatically, Schedule does not currently dispatch runs, a virtual DatasetVersion does not freeze live rows, and there is no universal rollback or exactly-once guarantee.

Next step

Examine the operating modules behind the flow.

Review source, file, transform and execution boundaries before mapping the flow to your environment.

Explore Data Operations