Elementary OSS Data Observability

Elementary OSS data testing

Elementary OSS adds tests to your dbt project for SQL tables and DataFrames, compares tables, checks unstructured text with warehouse LLM functions (beta), and stores the rows that fail.

• SQL Table Tests: Wraps all native dbt tests — not_null (completeness), unique (uniqueness), accepted_values (validity), relationships (consistency/RI) — plus dbt-utils for date ranges and custom SQL assertions (docs).

• DataFrame Tests: Python tests allow writing custom test logic as Python scripts executed inside the dbt environment (docs).

• Unstructured Data Tests: The elementary.unstructured_data_validation dbt test (Beta, v0.18+) uses warehouse-native LLM functions to validate free-text columns against plain-English expectations — no data leaves the warehouse (docs).

• Table Comparison Tests: Custom SQL dbt tests can assert aggregate consistency across multiple models (e.g., row counts match between source and target) (docs).

• Inspect Failed Rows: Elementary captures and surfaces the specific rows that triggered each failing dbt test for direct inspection (docs).

• Data Diff: No row-level dataset comparison tool found in documentation (docs).

• Data Contracts: No formal data contract feature (machine-readable producer-consumer SLA agreements) found in documentation (docs).

Elementary OSS data observability

Elementary OSS detects anomalies in volume, freshness, schema, columns and dimensions against historical baselines. All monitors are defined as code in dbt YAML.

Compare: data observability tools

• Data Profiling: Column anomaly tests compute null rate, distinct count, min, max, avg, and standard deviation per column continuously (docs).

• Anomaly Detection: Anomaly detection tests use historical baselines to detect volume, freshness, schema, column, and dimension anomalies without hardcoded thresholds (docs).

• Freshness Monitors: Freshness anomaly and Event freshness anomaly tests detect stale tables and delays in event ingestion (docs).

• Volume Monitors: Volume anomaly test detects drops or spikes in row counts compared to historical baselines (docs).

• Schema Monitors: Schema change and schema change from baseline tests detect added, dropped, or retyped columns (docs).

• Field Monitors: Column anomaly tests track null count, distinct count, min, max, avg, and standard deviation per column as time series (docs).

• Distribution Monitors: Column anomaly tests monitor value distributions per column and detect abnormal shifts in expected patterns (docs).

• Segment Monitors: Dimension anomaly test tracks counts per dimension value (e.g., by country or status) and detects distribution shifts between them (docs).

• Custom SQL Monitors: Custom SQL dbt tests allow any SQL query to be used as an assertion; also supports custom singular tests (docs).

• Monitors-as-Code: All tests and monitors are defined in dbt YAML/SQL files, version-controlled and deployed through dbt CI/CD (docs).

• Window-Based Monitors: Monitors run at dbt model granularity (per run), not over configurable tumbling or sliding time windows (docs).

• UI-Based Monitors: OSS is entirely code-based (dbt YAML/SQL); there is no UI for configuring tests or monitors (docs).

• Pipeline Monitoring: No pipeline job runtime tracking in OSS; performance features require Elementary Cloud (docs).

• Partition-Based Monitors: Elementary runs dbt tests which naturally scope to incremental model partitions, but this is a dbt model concern rather than a dedicated partition monitor feature (docs).

Elementary OSS incident management

Elementary OSS sends alerts to Slack or Teams. It has no incident workflow: triage, ticketing and resolution are in Elementary Cloud.

• Alert Notifications: Sends dbt test failure alerts to Slack, email, and other configured channels via the CLI (docs).

• Incident Triage: No structured triage workflow, owner assignment, or incident grouping in the OSS version (docs).

• Alert Feedback: Elementary OSS fires dbt test alerts via integrations but has no feedback or muting mechanism on individual anomaly signals (docs).

• Incident Resolution: No resolution lifecycle tracking or false-positive management in the OSS version (docs).

• Ticketing Integrations: No Jira or ServiceNow ticket creation in the OSS version; only Slack and email alerts (docs).

• On-Call Integrations: No PagerDuty or Opsgenie on-call paging in the OSS version (docs).

Elementary OSS data quality analytics

Elementary OSS generates a self-hosted HTML report from its CLI, with dbt test results, run history and anomaly trends.

• Data Quality Dashboards: Generates a self-hosted HTML observability report showing dbt test results, run history, and anomaly trends (docs).

• Data Health Scores: No per-table quality score in the OSS version; the report shows individual test results but no aggregate health score (docs).

• Incident Resolution Metrics: OSS provides alerting and a static self-hosted report, not an incident-management system with MTTD/MTTR or resolution-rate tracking; incident management is listed as a Cloud-only feature (docs).

Elementary OSS data lineage

Elementary OSS shows table- and column-level lineage of your dbt project in its report, with impact analysis. Lineage stops at dbt: it doesn't reach BI tools.

• Table-Level Lineage: Table-to-table lineage derived from the dbt manifest, showing model dependencies (docs).

• Column-Level Lineage: Column-level lineage available through the dbt manifest and the Elementary dbt package (pricing).

• Impact Analysis: dbt lineage graph can be used to trace which downstream models are affected by an upstream failure (docs).

• End-to-End Lineage: OSS lineage covers only dbt models; no integration with ingestion tools (Fivetran) or BI tools (Tableau, Looker) (docs).

Elementary OSS workflows

Elementary OSS runs as part of your dbt jobs and dbt CI, in GitHub Actions, GitLab CI or any orchestrator. A failed test fails the run.

• Scheduling: Runs on the dbt schedule; can be orchestrated via Airflow, Prefect, or any scheduler that triggers dbt runs (docs).

• CI/CD Integrations: Runs as part of the dbt CI job in GitHub Actions, GitLab CI, or any orchestrator; a failed dbt test exits the job non-zero, actively blocking the pipeline (docs).

• dbt Integration: Elementary IS a dbt package — tests run as dbt tests, lineage comes from the dbt manifest, and scheduling follows the dbt run (docs).

• Webhooks: No HTTP webhook delivery in the OSS version; only Slack and email alerts (docs).

• REST / GraphQL API: OSS is a CLI tool; no REST API server is provided (docs).

• MCP Server: MCP server is an Elementary Cloud feature only (docs).

• Python SDK: Elementary OSS is a dbt package invoked via dbt CLI — there is no separate pip-installable Python SDK for the OSS tier; the Python SDK is Cloud-only (docs).