Certified data sources are the foundation of a governed Tableau environment. They ensure that all dashboards use consistent, vetted data with shared field names and definitions. This guide covers how to build and maintain a certified data source programme — and what it takes to keep certification meaningful over time.
A certified data source in Tableau is a published data source that has been reviewed, validated, and marked as officially endorsed by the data team. The certification badge — a small blue checkmark that appears on data sources in Tableau Server and Tableau Cloud — signals to dashboard authors that this data source meets quality standards and is maintained by someone accountable for its reliability.
Certification is governance made visual. It answers the implicit question every analyst asks when they open the data source pane: "Can I trust this?"
What Certification Means
Certification is a commitment, not just a badge. A certified data source commits to:
**Documented field definitions**: every field a dashboard author might use has a description that explains what it means, how it is calculated, and what the valid values are. "Revenue" in a certified data source means a specific, documented calculation — not whatever the analyst thinks it might mean.
**Tested data quality**: the underlying data pipeline includes automated quality tests. The certified data source will not silently serve wrong data without the data team knowing. When data quality fails, the certification is visible evidence that someone is accountable.
**Defined refresh schedule**: the data source refreshes on a known schedule. The documentation states when data is current as of. Dashboard authors can communicate data currency to their stakeholders accurately.
**Named owner**: a specific person is responsible for this data source. When a dashboard author has a question about a field definition, they know who to ask. When data quality degrades, there is a named escalation point.
**Governance approval**: the data governance team or data steward for the relevant domain has reviewed and approved the field definitions and data model decisions in the certified source.
Without these elements, certification is cosmetic. A badge without backing commitments erodes trust faster than no badge — it creates false confidence that then gets violated.
Building the Certification Programme
**Define the certification criteria before certifying anything.** What must be true for a data source to earn the certification badge? Write it down. Common criteria: all fields have descriptions; the data source passes automated quality tests; it has a documented owner; it refreshes on a defined schedule; it has been reviewed by the governance function. Without documented criteria, certification is arbitrary.
**Start with the sources that matter most, not the most.** The highest-value certified sources are those that drive the most significant business decisions. Certifying 200 mediocre data sources dilutes the certification signal. Certifying the 10 data sources that drive executive reporting, budgeting, and operational decisions concentrates value where it matters.
**Establish the field documentation standard.** Field descriptions should be meaningful, not perfunctory. "Order ID" as a description for the order_id column is not useful. "Unique identifier for each order transaction. Corresponds to the order_id in the source ERP system. One order_id per completed transaction — returns and exchanges create new order_ids." is useful. Set a standard and enforce it before certifying.
**Create a certification review process.** Who reviews a data source before it is certified? Who signs off? A lightweight process — a checklist completed by the owner, reviewed by the data team lead or data steward — is sufficient. The process need not be bureaucratic; it needs to be consistent.
Maintaining Certification Over Time
Certification is not a one-time stamp — it is an ongoing commitment. Sources degrade. Field definitions become stale as business logic changes. Owners change roles. Quality tests that once passed start failing without anyone noticing.
**Periodic recertification**: conduct an annual or semi-annual review of all certified data sources. For each, verify: the owner is still the right person, the field definitions still reflect current business logic, the quality tests are still running and passing, the refresh schedule still meets consumer needs. Recertification catches drift before it becomes a trust problem.
**Remove or downgrade certification when it cannot be maintained**: a certified source whose owner left and has not been replaced should have its certification removed until a new owner is designated. A certified source that regularly fails quality tests or refreshes inconsistently should be downgraded. Leaving the badge on a source that no longer meets standards is worse than never having certified it.
**Make certification status visible in the data catalog**: the catalog entry for each certified source should show its certification status, the date of last recertification, the current owner, and the known quality test results. Consumers who check the catalog should be able to assess the current health of a certified source, not just its historical certification.
Published Data Source Architecture
Certified data sources are published to Tableau Server or Tableau Cloud as shared data sources — not embedded in individual workbooks. The shared data source model has several governance advantages:
**Single source of truth**: when a field definition changes (the revenue calculation is updated to include a new revenue type), the change is made in one published data source and is immediately reflected in every workbook that connects to it. Compare this to the embedded data source model, where the same field is defined independently in each workbook — a change requires modifying every workbook.
**Consistent extract schedules**: a published data source has one extract refresh schedule. All workbooks connecting to it receive the same data at the same time. There is no risk of some dashboards showing one refresh cycle's data and others showing a stale version.
**Usage analytics**: Tableau Server and Tableau Cloud provide usage data for published data sources — which workbooks connect to them, how many users query them, query performance. The owner of a certified source can see who is using it and how, enabling informed maintenance decisions.
**Access control at the source level**: permissions can be set on the published data source, controlling who can connect to and query it. Sensitive certified sources (containing financial data, HR data, PII) can be restricted to appropriate roles without requiring workbook-level permission management.
The Connection Between Certified Sources and dbt
In modern data stacks, certified Tableau data sources often connect directly to dbt-modelled tables in the analytical warehouse. The dbt model is the transformation layer; the certified Tableau data source is the presentation layer. Both benefit from shared governance:
Field descriptions written in the dbt model YAML are the same descriptions that appear in the Tableau data source. Tools like dbt-tableau-utils synchronise dbt field descriptions to Tableau field comments automatically, keeping documentation consistent between layers without manual duplication.
dbt tests on the underlying table validate the data quality that the certified Tableau source commits to. When a dbt test fails, the certified source's data quality commitment is violated — and the data team knows about it before dashboard users do.
Our Tableau consulting and managed BI services practice builds and maintains certified data source programmes — contact us to discuss governance for your Tableau environment.
A former Microsoft data architect audits your data foundation, identifies your top priorities, and sends you a written plan. Free. No pitch.
Book a Call →