# Coalesce

> Column-aware transformation built from reusable patterns, not one-off SQL

Coalesce is a commercial transformation platform that builds warehouse pipelines from reusable metadata-driven nodes and templates rather than hand-written SQL, with column-level lineage, cataloging, and quality checks in one managed service.

Website: https://coalesce.io

## Use it when

- Warehouse pipelines repeat the same shapes; metadata-driven nodes and templates enforce standards and propagate a pattern change everywhere it is used.
- Column-level lineage should come automatically from the tool's column-awareness, not from parsing SQL.
- The team prefers a guided visual builder over a repository of hand-written models.
- You run Snowflake, Databricks, Microsoft Fabric, or BigQuery.

## Think twice when

- Engine coverage is a constraint: those four, plus Redshift in preview, are the whole list, where dbt and SQLMesh reach further.
- The team wants plain versioned SQL files; dbt is the direct alternative with the larger ecosystem.
- Per-seat plus per-action pricing from $150 a user each month does not fit the budget.
- Raw SQL is common in your models; unqualified names break the lineage.

## How it runs

SaaS only: development happens in the browser against your warehouse and is always free. A deploy promotes a git commit's metadata into an environment and applies DDL; scheduled refreshes then run the DML, from the built-in scheduler or externally through the CLI and REST API from Airflow or Azure Data Factory.

## Details

### General

| Attribute | Value |
| --- | --- |
| License | Commercial |
| Deployment | Managed |
| Workload | Batch |
| Operational complexity | Low |
| Pricing | Free, Subscription |
| Documentation | https://docs.coalesce.io |

### Transformation

| Attribute | Value |
| --- | --- |
| Engines | Snowflake, Databricks, Microsoft Fabric, BigQuery; Redshift in preview |
| Model language | Visual nodes generating SQL, with Jinja templates underneath and raw SQL where you need it |
| Model kinds | Stage, View and Work ship built in; Source, Persistent Stage, Dimension and Fact come from platform packages, and node types are YAML plus Jinja templates you can author yourself |
| Change detection | A deployment plan diffs the metadata against the target and lists every create, alter and drop before it runs, structural, not a judgement about whether downstream logic breaks |
| Environments | Deploy promotes a git commit's metadata into an environment and applies DDL; refresh runs the DML separately, and you must deploy before you refresh |
| Testing | Built-in column tests for uniqueness and nulls, plus node tests as custom SQL that fails if it returns rows; each runs before or after the refresh and can halt the pipeline |
| Lineage | Column-level and structural: lineage comes from the column metadata rather than parsing SQL, which is why raw SQL has to use qualified names or the lineage breaks |
| Semantic layer | None |
| Scheduling | A built-in scheduler in the app, or the CLI and REST API from Airflow, Azure Data Factory or any DevOps tool |
| Billing unit | Per seat and per action, $150 a user each month on Starter, with production node executions and refreshes counted as actions; development is free and a one-user tier stays free |

## Capabilities

- [Transformation](https://matca.io/capabilities/transformation)

---

Source: https://matca.io/coalesce
Last updated: 2026-08-31T11:40:48.708Z