Packages related datasets, dashboards, and pipelines into a logical unit representing a business capability
Lives inside a Domain; has its own owners and documentation
The catalog entry for "what the data team offers to consumers" -- not just a raw table
Think: an internal API contract for data, not just an asset record
Data Products vs. Datasets
Dataset: a single table or view -- the physical layer
Data Product: a bundle (fact table, summary view, dashboard, pipeline) -- the business capability layer
Datasets are the unit of technical accountability; Data Products are the unit of business accountability
One Data Product can contain many datasets; by default, a dataset can also belong to more than one Data Product at once
Creating a Data Product
1. Navigate to the Domain that will contain the product
2. Click the Data Products tab (only appears on Domain pages -- not on global nav)
3. Click "+ New Data Product," enter a name and description, then create it
4. From the new Data Product's page, click "Add Assets" to bundle in datasets, dashboards, and pipelines
5. Add owners, tags, and glossary terms from the Data Product's sidebar
Brief: Screenshot of a Domain page in DataHub, showing the tab bar (e.g. Assets, Documentation, Data Products) with the "Data Products" tab visible, and the "+ New Data Product" button in the panel beneath it.
Name Data Products After Business Capabilities, Not Tables
The name should describe what the consumer gets, not how the engineer built it
Name it the way a business analyst would search for it, not the way a SQL query would reference it
What is an Application? (DataHub Cloud)
Groups related data assets around a single business purpose -- not a technical entity for software services
Sits between Domains and Data Products: Domain (business area) > Application (one purpose within it) > Data Product (a specific data offering)
A DataHub Cloud feature, disabled by default -- an admin turns it on under Settings > Appearance > Show Applications
Carries the same governance metadata as other entities: owners, domain, tags, structured properties
Most teams don't need this extra layer -- start with Domains and Data Products, and add Applications only if you need more structure between them
Applications in Practice (DataHub Cloud)
Payments Domain: everything related to processing payments across the business
Refund Processing Application: groups the assets used to handle customer refunds -- pipelines, processing datasets, and two Data Products ("Daily Refund Payments," "Daily Refund Failures")
Fraud Detection Application: a second Application in the same Domain, grouping the datasets and models used to catch fraudulent transactions
Each Application serves one specific purpose; a Domain can contain several
Data Product vs. Application -- The Relationship (DataHub Cloud)
Both are organizational containers that live inside a Domain -- not a producer/consumer pair joined by lineage
Application: the middle tier, one business purpose within a Domain (DataHub Cloud only)
Data Product: the smallest tier, a specific data offering -- usually nested inside an Application, but can also sit directly in a Domain
An Application can group multiple Data Products; a Data Product can be linked to more than one Application if your organization needs that
Create a Data Product for the Finance Domain
Assets to bundle: FACT_REVENUE_DAILY, the Weekly Revenue Summary dashboard, and daily_revenue_pipeline
Name: "Revenue Reporting"
Write a description answering: what does this represent, who should use it, what SLA does it carry?
Assign an owner (use a group, not an individual)
Link at least one glossary term
Check for Understanding
Q1: FACT_REVENUE_DAILY is a table in Snowflake. The Finance team packages it with an owner, an SLA, and documentation. Is that package a Data Product or a Dataset, and what is the difference?
Q2: Your organization has enabled Applications (DataHub Cloud) and created a "Refund Processing" Application inside the Payments domain with two Data Products nested in it. A teammate on a self-hosted OSS DataHub instance says they can't find the Applications tab anywhere. What's going on?
Q3: A critical Data Product has no linked source datasets and no SLA defined. What does a potential consumer learn from that, and what should the data owner do first?