This content is part of the comprehensive Governance in Tableau documentation.
Data source management includes processes related to selection and distribution of data within your organization. Tableau connects to your enterprise data platforms and leverages the governance you already have applied to those systems. In a self-service environment, content authors and data stewards have the ability to connect to various data sources, build and publish data sources, workbooks, and other content. Without these processes, there will be a proliferation of duplicate data sources, which will cause confusion among users, increase likelihood of errors, and consume system resources.
Tableau’s hybrid data architecture provides two modes for interacting with data, using a live query or an in-memory extract. Switching between the two is as easy as selecting the right option for your use case. In both live and extract use cases, users may connect to your existing data warehouse tables, views, and stored procedures to leverage those with no additional work.
Live queries are appropriate if you have invested in a fast database, need up-to-the-minute data, or use Initial SQL. In-memory extracts should be used if your database or network is too slow for interactive queries, to take load off transactional databases, or when offline data access is required.
With support for a new multi-table logical layer and relationships in Tableau 2020.2, users aren’t limited to using data from a single, flat, denormalized table in a Tableau Data Source. They can now build multi-table data sources with flexible, LOD-aware relationships between tables, without having to specify join types in anticipation of what questions can be asked of the data. With multi-table support, Tableau data sources can now directly represent common enterprise data models such as star and snowflake schemas, as well as more complex, multi-fact models. Multiple levels of detail are supported in a single data source, so fewer data sources are needed to represent the same data. Relationships are more flexible than database joins and can support additional use-cases as they arise, reducing the need to build new data models to answer new questions. Using relationships in well-modeled schemas can reduce both the time to create a data model as well as the number of data sources to answer business questions. For more information, see Metadata Management later in this section and The Tableau Data Model.
When publishing a workbook to Tableau Server or Tableau Online, the author will have a choice to publish the data source or leave it embedded in the workbook. The data source management processes you define will govern this decision. With Tableau Data Server, which is a built-in component of the Tableau platform, you can share and reuse data models, secure how your users access data, and manage and consolidate extracts with Published Data Sources. Further, Published Data Sources allow Tableau Creator- and Explorer-licensed users to have access to secure, trusted data in Tableau for web authoring and Ask Data. For more information, see Best Practices for Published Data Sources, Edit Views on the Web, and Optimize Data for Ask Data.
With increased data discovery capabilities, Tableau Catalog indexes all content, including workbooks, data sources, and flows to allow authors to search for fields, columns, databases, and tables in workbooks and published data sources. For more information, see Data Management Add-on.
When Tableau Catalog is enabled, content authors can Search for Data by selecting from Data Sources, Databases and Files, or Tables to see if it exists in Tableau Server and Tableau Online and minimize duplication of data sources.
In addition, the Data Details tab on a view published to Tableau Server and Tableau Online will provide consumers with relevant information about the data used in it. Details include information about the workbook (name, author, date modified), the data sources used in the view, and a list of the fields in use.
For data stewards who create new Published Data Sources, the workflow below shows the two major decision points that impact data source management—live or extract and embedded or shared data model. This is not to imply that a formal modeling process must always occur before analysis begins.