# LakeFusion Master Data Management

> Databricks-native master data management that runs in your Databricks workspace on Unity Catalog tables, matching records with rules, a Random Forest model trained on steward decisions and an LLM for uncertain pairs, then merging them into golden records.

Source: https://toolsfordata.com/tools/lakefusion/ · Updated 2026-11-20

[Website](https://lakefusion.ai) · [Docs](https://lakefusion.ai/mdm) · [Pricing](https://aws.amazon.com/marketplace/pp/prodview-hhofk6gcnpoca)

Best for: Databricks teams that need master data management for customer, supplier and product records without leaving the lakehouse.

## Overview

- Core Strength: Databricks-native MDM that matches, merges and stewards golden records in your workspace
- Matching: Rules, then a Random Forest, then an LLM for uncertain pairs
- Golden Records: Survivorship, steward review, hierarchies, PIM
- Platform Fit: Databricks and Unity Catalog on AWS or Azure
- Deployment Model: Self-hosted in your cloud account (PaaS)
- Pricing: $16,667/month on AWS Marketplace
- Notable Caveat: Docs portal is private; facts from release posts

## What it's used for

LakeFusion is used for master data management inside Databricks: it matches customer, supplier and product records in Unity Catalog tables with rules, a Random Forest trained on steward decisions and an LLM for uncertain pairs, then merges them into golden records that stewards review.

## Pricing

LakeFusion's AWS Marketplace listing is a flat $16,666.66 a month for unlimited usage, prorated and cancelable at any time, plus your own AWS infrastructure costs.

## Deployment

LakeFusion is deployed in your own AWS or Azure account as a PaaS and runs inside your Databricks workspace, against Unity Catalog tables.

## Review

Strengths:
- Runs inside your Databricks workspace on Unity Catalog tables, so golden records never leave the lakehouse.
- A full MDM hub: survivorship, steward review, hierarchies, reference data and, since 6.0, PIM.
- Three matching stages (rules, a Random Forest trained on steward decisions, an LLM for uncertain pairs) with SHAP explanations per pair.

Caveats:
- Databricks only, on AWS or Azure: Snowflake is just a source, and there is no Fabric option.
- The docs portal is private, so the facts here come from release posts and the website.
- The only public price is a flat $16,666.66 a month on AWS Marketplace, before infrastructure.

Recommendation: For ML entity resolution inside Databricks without a full hub, consider Zingg; for a Microsoft-first hub with Databricks connectors, Profisee; for MDM that matches in place on Snowflake, Semarchy; for SaaS real-time MDM, Reltio.

## Pricing

- Model: flat subscription
- Entry team price: $16,666.66/month (unlimited usage, plus AWS infrastructure)
- Source: https://aws.amazon.com/marketplace/pp/prodview-hhofk6gcnpoca (checked 2026-11-20)

## Features

### Matching

- ML Matching: Yes — A "Random Forest layer trained on an organization's own verified Ground Truth decisions" scores pairs (v6.0). ([source](https://lakefusion.ai/blog/introducing-lakefusion-6-0-17-capabilities-for-smarter-governed-master-data))
- Rule-Based Matching: Yes — The pipeline starts with "Deterministic Rules"; 5.0 adds rule reordering and a NOT operator. ([source](https://lakefusion.ai/blog/lakefusion-5-0))
- Real-Time Matching: Yes — "Search Before Create" sends a partial record "through a real-time API" and returns ranked candidates. ([source](https://lakefusion.ai/blog/introducing-lakefusion-6-0-17-capabilities-for-smarter-governed-master-data))
- Incremental Matching: Unknown — Matching only new or changed records is not described. ([source](https://lakefusion.ai/blog/introducing-lakefusion-6-0-17-capabilities-for-smarter-governed-master-data))
- Persistent Entity ID: Unknown — Golden records give "a single consistent identity", but a stable ID across merges isn't described. ([source](https://lakefusion.ai/mdm))
- Match Explainability: Yes — "SHAP feature contributions also provide visibility into why a pair received its score". ([source](https://lakefusion.ai/blog/introducing-lakefusion-6-0-17-capabilities-for-smarter-governed-master-data))

### Golden Records

- Survivorship Rules: Yes — Teams "set survivorship rules that decide which attributes land in the governed master record". ([source](https://lakefusion.ai/mdm))
- Survivorship on Read: No — Match-and-merge plus survivorship "produce trusted enterprise records" that are stored. ([source](https://lakefusion.ai/mdm))
- Stewardship Review: Yes — "Let data stewards review shaky matches, accept or reject merges". ([source](https://lakefusion.ai/mdm))
- Record History: Yes — "Lineage & audit trails" are listed under MDM. ([source](https://lakefusion.ai/lakefusion-databricks))
- Hierarchy Management: Yes — "Create parent-child and mixed-entity hierarchies". ([source](https://lakefusion.ai/mdm))

### Domains & Modeling

- Multi-Domain MDM: Yes — Customer, supplier, organization, product, location and reference data on "one governed base". ([source](https://lakefusion.ai/mdm))
- Configurable Data Model: Yes — 5.0 adds structs, arrays and 1:1, 1:N and N:N relationships to the model. ([source](https://lakefusion.ai/blog/lakefusion-5-0))
- Product Information Management: Yes — 6.0 adds PIM: "Multi-level taxonomy trees", attribute and specification libraries, media assets. ([source](https://lakefusion.ai/blog/introducing-lakefusion-6-0-17-capabilities-for-smarter-governed-master-data))
- Reference Data Management: Yes — 5.0 adds "a dedicated Reference Entity type" for code lists. ([source](https://lakefusion.ai/blog/lakefusion-5-0))
- Master Data Authoring: Yes — Stewards update records and manage golden records in the hub. ([source](https://lakefusion.ai/blog/lakefusion-mdm-the-complete-guide-for-enterprise-data-teams))

### Data Quality

- Validation at Entry: Yes — Dependent references "can be enforced during" data entry and loading. ([source](https://lakefusion.ai/blog/introducing-lakefusion-6-0-17-capabilities-for-smarter-governed-master-data))
- Standardization: Yes — It "standardizes and enriches incoming data". ([source](https://aws.amazon.com/marketplace/pp/prodview-hhofk6gcnpoca))
- Third-Party Enrichment: Yes — Integrates external data "through partners like Dun and Bradstreet". ([source](https://lakefusion.ai/partnerships))

### Data Distribution

- REST API: Yes — 6.0 documents "REST and SDK APIs". ([source](https://lakefusion.ai/blog/introducing-lakefusion-6-0-17-capabilities-for-smarter-governed-master-data))
- Change Events: Unknown — Only a webhook action on graph conditions is described. ([source](https://lakefusion.ai/blog/introducing-lakefusion-6-0-17-capabilities-for-smarter-governed-master-data))
- Warehouse Write-Back: Yes — Golden records are mastered in place in Unity Catalog tables in your Databricks workspace. ([source](https://lakefusion.ai/lakefusion-databricks))
- Data Syndication: Yes — Products "can then be published into a live catalog" across channels (6.0). ([source](https://lakefusion.ai/blog/introducing-lakefusion-6-0-17-capabilities-for-smarter-governed-master-data))

### AI & Agents

- MCP Server: Yes — 6.0 "introduces an MCP server that exposes governed LakeFusion capabilities to AI agents". ([source](https://lakefusion.ai/blog/introducing-lakefusion-6-0-17-capabilities-for-smarter-governed-master-data))
- AI Stewardship Assistance: Yes — AI column mapping and LLM review of uncertain pairs, which teams "can override". ([source](https://lakefusion.ai/blog/introducing-lakefusion-6-0-17-capabilities-for-smarter-governed-master-data))

### Platform Native

- In-Place Matching: Yes — "Master enterprise entities inside the Databricks environment you already have" instead of a separate MDM platform. ([source](https://lakefusion.ai/mdm))
- Snowflake Native: No — Runs on Databricks; Snowflake is only a source. ([source](https://lakefusion.ai/blog/introducing-lakefusion-6-0-17-capabilities-for-smarter-governed-master-data))
- Databricks Native: Yes — "LakeFusion runs inside the customer's Databricks workspace and works against Unity Catalog tables." ([source](https://lakefusion.ai/lakefusion-databricks))
- Microsoft Fabric Native: No — Only Databricks on AWS and Azure is named. ([source](https://lakefusion.ai))

### Deployment

- SaaS Deployment: No — A "Platform-as-a-Service (PaaS) solution deployed directly within your cloud infrastructure". ([source](https://lakefusion.ai/trust-and-security))
- Self-Hosted Deployment: Yes — "Deployed directly within your cloud infrastructure (AWS or Azure)". ([source](https://lakefusion.ai/trust-and-security))

### Open Source

- Open Source: No — Commercial license (vendor EULA). ([source](https://aws.amazon.com/marketplace/pp/prodview-hhofk6gcnpoca))

## Integrations

6 integrations, each sourced from the vendor docs.

- Data Enrichment: Dun & Bradstreet, Moody's
- Data Governance: Databricks Unity Catalog
- Data Platform: Databricks
- Data Warehouse: Snowflake
- Databases: Databricks Lakebase

## Alternatives

- [Zingg](https://toolsfordata.com/tools/zingg/): The other Databricks-native pick: ML entity resolution as a pipeline step in your workspace, open source, with persistent IDs in the paid editions but no golden records yet.
- [Profisee](https://toolsfordata.com/tools/profisee/): Consider Profisee for a Microsoft-first MDM hub with side-by-side steward review, native in Fabric and connected to Databricks.
- [Semarchy](https://toolsfordata.com/tools/semarchy/): Consider Semarchy for no-code MDM that also matches data in place, with a Snowflake native app if your lakehouse isn't Databricks.
- [Reltio by SAP](https://toolsfordata.com/tools/reltio/): Consider Reltio for SaaS MDM that matches in real time with pretrained LLM models and pipes golden records to Databricks, Snowflake or BigQuery.
- [Tamr](https://toolsfordata.com/tools/tamr/): Consider Tamr for ML matching with curator queues and persistent Tamr IDs, as SaaS or self-hosted, connected to Snowflake, BigQuery and OneLake.
