Skip to main content

Node Types by Scenario

Pick Node Types by modeling scenario, with platform notes and links to package listings. For the compact scenario-to-package table, start from Choosing the Right Node.

Modeling Baseline With Built-In Node Types​

Built-in Node types ship with Coalesce and cover staging, conformed modeling, persistent handoff, and simple views. Enable them under Build Settings > Node Types. Behavior, defaults, and how templates run: Nodes and Node Types.

Base Node Types on Snowflake, Databricks, and BigQuery​

Marketplace Base Node Types packages supply Work, Persistent Stage, Dimension, Fact, View, and Code templates per platform. Snowflake also includes Factless Fact and SQL Stage in the same package. Install the package for your engine, then use the listing as the source of truth for Node names and options: Snowflake Base Node Types, Databricks Base Node Types, BigQuery Base Node Types.

Advanced Deploy Merge and Tracking​

Advanced Deploy variants add merge, change tracking, last-modified comparison, and related options on Dimensions, Facts, Persistent Stages, Work, and Factless Facts when each listing supports them. Snowflake Advanced Deploy and BigQuery Advanced Deploy both ship these patterns. The BigQuery Advanced Deploy listing also includes Factless Fact Advanced Deploy where documented. Package references: Snowflake Base Node Types Advanced Deploy, BigQuery Base Node Types Advanced Deploy. For a summary by Node family on this page, see Advanced Deploy Options by Node Family.

Snowflake Change Feeds and Tasks​

Use Streams and Tasks Node Types when you orchestrate change feeds and scheduled work in Snowflake. Node names and options: Snowflake Streams and Tasks.

When You Land Each Batch in a Staging Layer​

You need an intermediate table or view that usually holds only the current extract or one run's slice, not the long-lived conformed row set.

Use: built-in Stage on any data platform, or Work from Base Node Types on your data platform.

When You Keep Curated Rows Across Runs on a Business Key​

You need a table that retains loaded rows, supports merge-style loads, and can track change history for selected columns when the package exposes those options.

Use: Persistent Stage from Base Node Types, or Persistent Stage Advanced Deploy for the full merge and tracking matrix on the listing.

Pair with When You Load Only New or Changed Rows when each Job should touch only new or changed sources.

When You Load Only New or Changed Rows​

You run on a schedule or trigger Jobs often, have a watermark or equivalent, and want each Job to touch only new or changed rows while a persistent table holds prior loads.

Use: Incremental Load and related types from Marketplace packages, wired to a Persistent Stage, Dimension, or Fact from Base Node Types.

When You Describe Business Entities and Slowly Changing Attributes​

You model customers, products, dates, or other descriptive context, usually one row per entity. Dimension Nodes require a business key.

Use: Dimension from Base Node Types, or Dimension Advanced Deploy when you need merge, change tracking, and last modified options on the Advanced Deploy listing.

  • Type 1: Attributes that should not version stay on one row. Do not select change tracking for those attributes; updates overwrite the current row.
  • Type 2: Select change tracking on columns that should version when values change. Type 2 increases row count and merge work compared with Type 1.

Orders and Statuses​

If you only need current status on the order row, keep status non-tracked on an order-level Dimension or use a conformed status Dimension without Type 2 on the order. If you need full status history, use Type 2 on an order-keyed Dimension or a Fact table with one row per status change plus order, time, and status keys. Pick the shape your reports query.

Package Listings for Dimensions​

Dimension and Dimension Advanced Deploy on Snowflake Base Node Types and Snowflake Base Node Types Advanced Deploy. Dimension on Databricks Base Node Types. Dimension and Dimension Advanced Deploy on BigQuery Base Node Types and BigQuery Base Node Types Advanced Deploy. Enable SCD Type 2 and per-column change tracking on the BigQuery Advanced Deploy listing where you need them. SCD edge cases: SCD Type 2 change detection.

When You Store Measures in Fact Tables​

You model numeric measures or additive facts where each row is one business event or line item.

Use: Fact from Base Node Types, or Fact Advanced Deploy for documented MERGE, INSERT, last modified comparison, and unmatched-row handling on the package matrix.

With a business key at that row level, templates usually MERGE; without one, many Fact templates INSERT. Confirm in your package version and generated SQL.

Order header versus line usually means different tables: header attributes in Dimensions with one row per order, line measures in Facts with one row per line. Degenerate dimensions on the Fact are fine when they are not conformed across many Facts.

Package Listings for Facts​

Fact and Fact Advanced Deploy on Snowflake Base Node Types and Snowflake Base Node Types Advanced Deploy. Fact on Databricks Base Node Types. Fact on BigQuery Base Node Types and BigQuery Base Node Types Advanced Deploy.

When You Record Events or Relationships Without Measures​

You track coverage, eligibility, milestones, or many-to-many relationships with keys and optionally time, without numeric measures.

Use: Factless Fact on Snowflake from the base package, or Factless Fact Advanced Deploy from Snowflake or BigQuery Advanced Deploy packages where the listing provides them. Otherwise use Fact or Persistent Stage from Base Node Types so each row is one event or relationship.

When You Want a Simple View Object in the Graph​

You want a VIEW for clarity, security boundaries, or a thin semantic layer, with a dedicated Node in the DAG.

Use: Built-in View from Nodes and Node Types, or the View Node Type from Base Node Types in Coalesce Marketplace for your data platform. Enable built-in View in Build Settings > Node Types if it is off. Dimension, Fact, and Work can also use Create As View on the listing when a view materialization fits those layers.

When Several Sources Feed One Conformed Node​

You combine extracts or shards before a single Dimension or Fact.

Use: Multi Source on Work, Persistent Stage, Fact, or other types your Base Node Types or Advanced Deploy package supports, or chain Work Nodes that UNION inputs.

Patterns and ordering: Multi Source Nodes. Check MultiSource support on Snowflake and BigQuery package matrices for the Node Type you chose.

Advanced Deploy Options by Node Family​

This table summarizes Snowflake Base Node Types Advanced Deploy. Base options for the same families are on Snowflake Base Node Types. Change tracking and last modified comparison apply to Dimension and Persistent Stage in this matrix, not to Fact or Factless Fact. Business key applies to Dimension, Fact, and Persistent Stage. Exact flags and methods can change with package version.

Node familyChange trackingLast modified comparisonBusiness keyTypical methodsUse case
DimensionYesYesYesMERGE, INSERT, UPDATEConformed customer or product rows on a natural key, optional Type 2 on selected attributes.
FactNoYesYesMERGE, INSERTOrder lines or shipments with MERGE when a stable business key exists, or insert-only when it does not.
Factless FactNoNoNoMERGE, INSERTCoverage or eligibility without measures, for example promotions by store and week.
Persistent StageYesYesYesMERGE, INSERTHandoff between ingest and conformed layers, rows retained across runs.
WorkNoNoNoINSERTPer-batch landing or scratch results replaced or appended each run.

On BigQuery, see BigQuery Base Node Types and BigQuery Base Node Types Advanced Deploy for similar matrices that cover Dimension, Fact, Persistent Stage, Work, and Factless Fact Advanced Deploy.

When You Own the Create and Run Templates​

You ship or extend package behavior with workspace-owned Jinja and SQL instead of only UI-driven options.

Use: Code Node Types from the package that matches your platform and layer. Listings include Code on Snowflake Base Node Types, Snowflake Base Node Types Advanced Deploy, Databricks Base Node Types, BigQuery Base Node Types, BigQuery Base Node Types Advanced Deploy, Incremental Loading package, and Incremental Nodes.

When You Transform Entirely in Hand-Authored SQL (Snowflake)​

You skip the column-mapping grid and keep transformation logic in a single SQL body.

Use: SQL Stage on Snowflake Base Node Types for table materialization and stage options on the listing.

Validate Changes in a Development Environment​

Switching Node Types or moving between base and Advanced Deploy after objects exist can change materialization, templates, and options. Test Deploy and refresh in a development environment before production.

Macros and the Node editor

Invalid Workspace macros for your platform's SQL engine can prevent options from rendering or behaving as expected. Use supported functions so the Node UI and generated SQL stay aligned.

Sample Node YAML and graphs in the public Coalesce GitHub organization illustrate SCD and persistent patterns. For example, browse repositories such as Coalesce-Base-Node-Types and Coalesce-Base-Node-Types---Advanced-Deploy.

What's Next?​

Package listings remain the source of truth for exact Node labels. For example, Grouped Incremental load, Test Passed Records, Test Failed Records, and other types on the incremental and functional packages.