Exposures

Exposures declare downstream consumers of the pipeline's outputs — dashboards, notebooks, ML models, applications, or other pipelines. They complete the lineage graph by connecting data production to data consumption.

Exposures are metadata-only. They do not affect pipeline execution.

Structure

exposures:
  - name: <string>
    type: <string>
    description: <string>
    url: <string>
    maturity: <string>
    owner:
      name: <string>
      email: <string>
    depends_on: [<AssetRef>]
    tags: [<string>]
    meta:
      <key>: <any>

Fields

Field	Type	Required	Default	Description
name	string	Yes	—	Exposure identifier.
type	string	Yes	—	Consumer type (see below).
description	string	No	—	What this consumer does with the data.
url	string	No	—	Link to the consumer (dashboard URL, notebook link, etc.).
maturity	string	No	—	Maturity level: "high", "medium", or "low".
owner	Owner	No	—	Consumer owner or team.
depends_on	NonEmptyList[AssetRef]	Yes	—	Assets consumed. Must reference asset names defined in the pipeline.
tags	List[string]	No	[]	Classification labels.
meta	Map[string, any]	No	{}	Custom metadata.

Exposure Types

Type	Description
dashboard	BI dashboard or report.
notebook	Jupyter or Zeppelin notebook, or exploratory analysis.
analysis	Ad-hoc analysis or query.
ml	Machine learning model or feature store.
application	Application or microservice consuming the data.
pipeline	Another data pipeline that depends on this pipeline's outputs.

The depends_on Field

The depends_on field lists the assets that the exposure consumes. These must reference assets defined in the pipeline (inputs, transformations, or outputs).

This field is what connects the exposure to the pipeline DAG and enables impact analysis.

exposures:
  - name: revenue_dashboard
    type: dashboard
    depends_on: [daily_revenue, monthly_summary]

When daily_revenue or monthly_summary changes schema or breaks, the implementation can trace the impact to revenue_dashboard.

Impact Analysis

Exposures enable downstream impact analysis. Given any asset in the pipeline, the implementation can answer:

• What consumes this asset? — Follow the depends_on references backward from exposures.
• What breaks if this asset changes? — Identify all exposures (and their owners) that depend on the affected asset.

This complements the upstream lineage that is implicit in the DAG. Together, they provide full bidirectional traceability from source data to end consumers.

Examples

Dashboard

exposures:
  - name: revenue_dashboard
    type: dashboard
    maturity: high
    url: "https://bi.company.com/dashboards/revenue"
    description: "Executive daily revenue tracking"
    owner:
      name: Analytics Team
      email: analytics@company.com
    depends_on: [daily_revenue, monthly_summary]
    tags: [executive, finance]

ML Model

exposures:
  - name: churn_model
    type: ml
    description: "Customer churn prediction model"
    owner:
      name: ML Engineering
      email: ml@company.com
    depends_on: [customer_features]
    tags: [ml, production]
    meta:
      model_version: "3.2"
      framework: xgboost

Notebook

exposures:
  - name: quarterly_analysis
    type: notebook
    maturity: low
    url: "https://notebooks.company.com/analysis/q1-revenue"
    description: "Quarterly revenue deep-dive analysis"
    owner:
      name: Sarah Chen
      email: sarah@company.com
    depends_on: [daily_revenue]
    tags: [analysis, finance]

Application

exposures:
  - name: billing_service
    type: application
    maturity: high
    url: "https://github.com/company/billing-service"
    description: "Billing microservice reads customer subscription data"
    owner:
      name: Billing Team
      email: billing@company.com
    depends_on: [customer_subscriptions]
    tags: [production, billing]

Downstream Pipeline

exposures:
  - name: marketing_pipeline
    type: pipeline
    description: "Marketing attribution pipeline consumes customer segments"
    owner:
      name: Marketing Engineering
      email: mktg-eng@company.com
    depends_on: [customer_segments, daily_revenue]
    tags: [marketing, production]
    meta:
      pipeline_id: "com.company.marketing.attribution"

Complete Example with Multiple Exposures

exposures:
  - name: revenue_dashboard
    type: dashboard
    maturity: high
    url: "https://bi.company.com/dashboards/revenue"
    description: "Executive daily revenue tracking"
    owner:
      name: Analytics Team
      email: analytics@company.com
    depends_on: [daily_revenue, monthly_summary]
    tags: [executive, finance]

  - name: churn_model
    type: ml
    description: "Customer churn prediction model v3.2"
    owner:
      name: ML Engineering
      email: ml@company.com
    depends_on: [customer_features]
    tags: [ml, production]
    meta:
      model_version: "3.2"
      framework: xgboost

  - name: billing_service
    type: application
    maturity: high
    description: "Reads subscription data for invoice generation"
    owner:
      name: Billing Team
      email: billing@company.com
    depends_on: [customer_subscriptions]
    tags: [production, billing]

  - name: marketing_pipeline
    type: pipeline
    description: "Downstream marketing attribution pipeline"
    owner:
      name: Marketing Engineering
      email: mktg-eng@company.com
    depends_on: [customer_segments, daily_revenue]
    tags: [marketing]

With these declarations, an implementation can determine that changing the daily_revenue asset impacts both the revenue_dashboard and the marketing_pipeline, and can notify the respective owners.