What Prizm Collects from dbt

Prizm runs two platform jobs against each dbt connector:

Technical
Operational

Type: CONTEXT — runs once after setup is complete.Parses manifest.json to build the full dbt catalog: models, sources, tests, exposures, columns, lineage, tags, and owners. All subsequent jobs and UI views depend on this completing successfully.

Object	Fields Captured
Model	Name, database, schema, materialization, dependencies, tags, owner, description, compiled SQL
Source	Name, database, schema, table, freshness thresholds, loaded-at field, tags, owner, description
Test	Name, type (generic / singular), severity, attached model/column, tags, description
Exposure	Name, type, upstream models, owner, maturity, tags, description
Column	Name, data type, description, parent model, associated tests, tags

catalog.json is optional and enriches column data types with warehouse-introspected values. semantic_manifest.json is ingested if present — full metric UI support coming soon.

Type: TRUST — re-runs on every artifact push (dbt Core) or webhook event (dbt Cloud).Ingests run_results.json and sources.json to power run history, test pass rates, model health trends, alerting, and freshness tracking.

Signal	Source	What It Powers
Job Run	`run_results.json`	Run history, job health cards
Model Run Result	`run_results.json`	Execution time graphs, failure alerting, rows affected
Test Run Result	`run_results.json`	Test pass rate, failure issues
Source Freshness	`sources.json`	Freshness SLA tracking on source assets

What You See in Prizm

dbt Job Page

Access from Home Page → View All → Jobs → DBT

Overview
Metric
Run
Model
Test

Health snapshot of the dbt job powered by the last 30 days of run data.Key Metric Cards: Failure Rate (30D), Avg Duration (30D), P95 Duration (30D), Runs (Last 7D), Runs (Last 30D).Prizm also surfaces AI risk signals — duration regression, recurring failures, high-impact job, and test degradation.Right Side Panel — More Info: Qualified Name, Source Name, Type (JOB), Execution Mode, Job ID, Last Run At, Modified At,DBT Cloud Specific Metadata: Environment, Job Schedule

Metric	Category
Execution Status	Availability
Job Run Freshness	Availability
Job Duration	Usage
Job Success Rate	Queries

dbt Model Asset Page

Key Metric Cards: Volume, Schema, Freshness, Metrics, Total Runs, Failure Rate (%).Right Side Panel: Qualified Name, Source Name, Type, Materialization, Materializes To (linked warehouse table), Alias, dbt Owner, dbt Tags, Database, Schema, Execution Mode.

Metric	Category
Run Success Rate	Queries
Execution Time	Usage
Rows Affected	Availability
Freshness	Availability

dbt Test Asset Page

Each dbt test is a standalone asset in Prizm.

Overview
Metric
Run

Key Metric Cards: Total Executions, Failures, Failure Rate (%).Also shows Description, Semantic context, Owners, Audience, Who should use this, and inline Lineage showing the upstream model and column.

Metric	Category
Test Execution Time	Usage
Test Status	Conditional

Right Side Panel — More Info: Qualified Name, Source Name, Type (TEST), Execution Mode, Severity (error / warn), Test Type (generic / singular), Created At, Modified At.

dbt Exposure Asset Page

Each dbt exposure represents a downstream consumer (dashboard, notebook, application, ML pipeline). Tabs: Overview, Lineage, Audit, Documentation, Conversation.

Overview
Lineage

Key Metric Cards: Volume, Schema, Freshness, Metrics (inherited from upstream models).Also shows Description, Semantic context, Owners, Audience, Who should use this, and inline Lineage showing the upstream model chain.

Exposures have no Run, Test, Metric, or Compiled Code tabs — they are not executed by dbt. All quality and metric signals come from upstream models.

Get Started

Core Concepts

Platform

Architecture

Data Sources

Integrations

Security & Access Control

Deployment

Help

Settings

MCP

Release Notes

API Reference

What Prizm Collects from dbt