1/13

02 Nov 2026

Tableau to Databricks Migration: Can Genie Code Import It?

A Tableau to Databricks migration is really one of three projects: repoint Tableau to governed Databricks tables so the data and logic move while the dashboards stay, import the workbooks into Databricks AI/BI dashboards and metric views, or rebuild the dashboards in another BI tool. Databricks documents an /importBI command in Genie Code, described in its release notes as generally available, that reads Tableau .twb, .twbx, .tds, and .tdsx files and creates AI/BI dashboards and metric views, and Tableau connects to Databricks SQL warehouses through its native connector. The risk sits in calculations that depend on the view, such as INCLUDE and EXCLUDE expressions and table calculations, in which identity reaches Databricks for row-level security, and in small differences like ranking and date truncation, so every number should be reconciled before a workbook goes live.

Mature Tableau estates accumulate business logic in places never meant to hold it: the same calculated field copied across dozens of workbooks, extracts refreshing on independent schedules, and Prep flows running on individual desktops. This page covers the three paths, how Tableau concepts map to Databricks, what the Genie Code import does and does not do, how to keep Tableau on top of Databricks safely, and where results drift. For the services side, see our Databricks migration services overview, and for the pipelines feeding your dashboards, our ETL migration to Databricks article.

Quick Summary

Decide first whether the dashboards move or only the data layer, because that choice sets the scope, the cost, and the user impact. Databricks can now import Tableau workbooks into AI/BI dashboards, which makes the dashboard path faster to start but not exempt from validation. Whichever path you take, view-dependent calculations, security identity, and reconciliation are where projects succeed or stall.

Key Takeaways

Pick the path before the plan. Repointing Tableau, importing into AI/BI dashboards, and rebuilding in another tool are different projects.
Databricks can import Tableau files. The /importBI command in Genie Code builds AI/BI dashboards and metric views from workbooks, and it is a starting point, not a finish line.
View-dependent logic resists SQL. INCLUDE and EXCLUDE expressions and table calculations often stay in Tableau or need redesign.
Security depends on identity. Unity Catalog row filters apply per user only when the connection carries the user's identity.
Small differences move totals. Ranking variants, date truncation, and extract timing are common culprits.
Reconcile before repointing. No workbook should go to production until its numbers match the old source.

Three Paths: Repoint, Import, or Rebuild

The phrase "Tableau to Databricks migration" covers three different projects, and conflating them is the most common cause of scope creep:

Path	What Moves	Best When	Watch For
Repoint Tableau	Data sources, extracts, and Prep logic move to governed Databricks tables, and workbooks are repointed	Users like Tableau, and duplicated logic and extract sprawl are the pain	The authentication model for row-level security, and live-query versus extract performance
Import into AI/BI dashboards	Workbooks become Databricks AI/BI dashboards and metric views through /importBI or partner tools	You want data, BI, and AI on one governed platform	Reviewing parity, because features and design may differ
Rebuild in another BI tool	Dashboards are redesigned in a different platform, with Databricks as the data layer	You are standardizing on another BI tool	A full redesign, as in our Tableau to Power BI, Tableau to Microsoft Fabric, and Tableau to Amazon QuickSight articles

A backend repoint has the lowest user impact, because people keep the interface they know, and it is the foundation for any later replacement. Many estates mix paths: heavy, widely shared dashboards stay on Tableau, while new self-service work lands in AI/BI dashboards. For help comparing the front-end options, see our MicroStrategy vs Power BI vs Tableau comparison, and for broader scoping, our enterprise BI migration services page.

Why Teams Move Tableau Logic to Databricks

The motivation is rarely that Tableau is a weak visualization tool. It is a question of where transformation logic should live:

One governed definition. A single implementation of each business rule replaces copies of the same calculation scattered across workbooks.
Modern pipelines. Extract schedules and desktop Prep flows give way to versioned, testable, orchestrated pipelines.
Scale independent of the BI layer. Databricks compute scales with data volume without touching the dashboards.
Unified governance. Access control, lineage, and auditing in Unity Catalog apply to every consumer, not only Tableau.
Reuse beyond dashboards. The same curated gold tables serve dashboards, data science, applications, and other BI tools.
Audit support. Delta Lake adds ACID transactions, time travel for audit comparisons, and schema enforcement.

If your estate also feeds from Oracle, SQL Server, or Teradata, the upstream moves are covered in our Oracle, SQL Server and SSIS, and Teradata articles.

How Tableau Concepts Map to Databricks

These are functional equivalents, not one-to-one replacements, because Tableau evaluates many objects relative to the current view while SQL evaluates them at a fixed grain:

Tableau	Databricks	Note
Published and embedded data sources	Unity Catalog tables or views in a gold schema, or metric views	One governed object replaces many copies
Joins and relationships	Explicit SQL joins in silver or gold tables	Tableau relationships pick join behavior at query time, so check for row duplication
Row-level calculated fields	Derived columns or SQL expressions	Confirm function semantics such as date truncation
LOD expressions (FIXED)	Window functions or grouped subqueries joined back	Predictable once the filter behavior is decided
LOD expressions (INCLUDE and EXCLUDE) and table calculations	Usually kept in Tableau, or redesigned as measures	They depend on the view, and /importBI covers some cases
Data source filters	WHERE clauses in views	Applied before every query
User filters and row-level security	Unity Catalog row filters and column masks	The connection's identity decides whose scope applies
Parameters	Tableau parameters, or AI/BI dashboard parameters	Keep in Tableau unless the dashboard moves
Hyper extracts (.hyper)	Delta tables served by a SQL warehouse	Extracts remain optional for concurrency or offline use
Tableau Prep flows	Notebooks or Lakeflow pipelines run by Lakeflow Jobs	Each Prep step becomes versioned, testable code
Extract refresh schedules	Lakeflow Jobs that run upstream, and then trigger refreshes	Triggering can use Tableau's REST API
Dashboards	AI/BI dashboards through /importBI, or stay in Tableau	Depends on the path you chose

Importing Tableau Workbooks with Genie Code

Databricks documents an import path for Tableau files, which makes the dashboard route much faster to start than a manual rebuild. According to the Databricks documentation and release notes:

What it reads. You upload a Tableau .twb, .twbx, .tds, or .tdsx file, or point to one stored in a Unity Catalog volume, and use the /importBI command in Genie Code.
What it builds. An AI/BI dashboard that replicates the visualizations, connected to metric views that mirror the business logic in the file.
Metric views. Genie Code creates local metric views scoped to the dashboard by default, and you promote them to Unity Catalog metric views to share them.
Limits. Direct uploads are limited to 100 MB, so use a Unity Catalog volume for larger files, or unzip a .twbx and upload the .twb.
Scale. As of October 2026, Genie Code can import up to 100 Tableau or Power BI files at a time, and later releases have added support for more parameterized and ranked table calculations.
Status. The release notes describe the import as generally available, so confirm the current status in your workspace.

Partner accelerators follow the same idea, with vendors advertising their own coverage rates for typical Tableau workbooks. Treat every figure as a vendor claim, and treat every import as a first draft: review the generated dashboard for visual parity, check each calculation against Tableau's output, and expect to redesign features that have no equivalent. An import does not prove the numbers match, so the reconciliation work in this page applies in full.

Keeping Tableau on Top of Databricks

If you repoint rather than replace, the connection details matter as much as the tables:

Native connector. Tableau Desktop, Cloud, and Server connect to a Databricks SQL warehouse through the Databricks connector or Partner Connect.
Authentication. Tableau lists Databricks login (OAuth, recommended), personal access token, and service principal, and basic username and password authentication ended in July 2024.
Tableau Cloud. Databricks enforces OAuth when publishing to Tableau Cloud, and Tableau Server needs its own OAuth configuration.
Token lifetime. By default, OAuth tokens for Tableau expire after 90 days, which belongs in your operations runbook.
Metric views. Tableau can query Unity Catalog metric views through Custom SQL with the MEASURE() pattern or through BI compatibility mode (Beta), which cannot join metric views to other tables in the BI tool and does not support materialized metric views.

For most repoint projects, curated gold tables or views are the simpler target, with metric views added where one governed set of measures is worth the extra constraints. Performance comes from sizing SQL warehouses for real concurrency, pre-aggregating in gold, and using liquid clustering on frequently filtered columns, and it should be measured, not assumed, because live queries can be faster or slower than Hyper extracts.

LOD Expressions and the Order of Operations

According to Tableau's order of operations, FIXED expressions are evaluated before dimension filters, so an ordinary dashboard filter does not change them unless it is a context filter, while INCLUDE and EXCLUDE expressions are evaluated after dimension filters. Top N filters are also applied before ordinary dimension filters unless those are context filters. That is why FIXED logic moves into SQL predictably and INCLUDE logic loses its ability to respond to the view.

A FIXED expression such as the first order date per customer, {FIXED [Customer ID] : MIN([Order Date])}, is a window function in Databricks SQL. This is an illustrative example, not a client engagement.

SELECT
  order_id,
  customer_id,
  order_date,
  MIN(order_date) OVER (PARTITION BY customer_id) AS first_order_date
FROM silver.orders;

The SQL version ignores dashboard filters entirely, which matches a FIXED expression when no context filters are set. If the Tableau workbook used context filters, decide which filters should shape the value and build them into the query, because the dashboard cannot reach into a gold table at run time. Please test this example in a Databricks workspace before adapting it.

Row-Level Security: Whose Identity Reaches Databricks?

Recreating Tableau user filters as Unity Catalog row filters and column masks applies the rules to every consumer, not only Tableau. Unity Catalog evaluates them at query time, so no extra warehouse-level enforcement is needed. What matters is the identity the connection carries:

Connection Style	Whose Scope Applies	Implication
Per-user OAuth (user-to-machine)	Each viewer's own identity	Row filters apply per person
Shared service principal or token	The service principal's identity for everyone	All viewers share one data scope unless per-user delegation exists upstream

Decide the authentication model first, test each role with real accounts, and document who sees what. Our guides to migrating row-level security and data security compliance cover the design side.

Where Tableau Results Drift on Databricks

Tableau Behavior	What Changes on Databricks	Safeguard
Four ranking functions: RANK, RANK_DENSE, RANK_UNIQUE, and RANK_MODIFIED	SQL has rank(), dense_rank(), and row_number(), with no direct modified-competition equivalent	Map each Tableau function deliberately and compare top-N lists around ties
Top N applied before dimension filters	A SQL filter order can return different rows	Rebuild the filter order explicitly, using context filters where needed
Date truncation and week start settings	Truncation defaults and week boundaries can differ	Confirm the start-of-week rule and test boundary dates
Hyper extract snapshot versus live query	Live queries show current data, while extracts show the last refresh	Align refresh timing during parallel runs
Ratios and averages	An average of row-level ratios is not a sum divided by a sum	Compute ratios from totals and check distinct-count logic
NULL handling in derived logic	Concatenation, comparisons, and defaults can diverge even when SUM and AVG agree	Test derived fields on rows with NULLs

These are the differences behind dashboards that look right and quietly are not, and the reason for column-level checks like those in our automated ETL testing with Python approach. Compare record counts, KPIs at total and dimension level, rankings around ties, filter behavior, and load times under realistic concurrency, each with an agreed tolerance and an owner.

A Phased Tableau to Databricks Migration Plan

Phase	What Happens	Exit Criterion
Inventory	Catalog workbooks, data sources, extracts, calculated fields, LOD expressions, parameters, and usage, and retire unused content	A scored inventory, because workbook files are XML and can be read programmatically
Logic documentation	Record every calculated field with its definition, owner, and expected output	A signed-off logic catalog
Path decision	Choose repoint, import, or rebuild for each subject area	A path per subject area
Data build	Land source data unchanged in bronze, and build silver and gold at the grain dashboards need	Gold tables reconciled to sources
Logic migration	Move row-level calculations and fixed-grain aggregations into gold tables, views, or metric views	Documented functional equivalents
Dashboard step	Repoint workbooks to gold objects, or import them with Genie Code, keeping field names identical where possible	Dashboards loading from the new source
Parallel run and validation	Run old and new side by side for at least one full reporting cycle and reconcile	Signed-off parity
Cutover and monitoring	Promote through environments, then watch job failures, freshness, and warehouse performance	Stable operations

Migrate in waves by subject area, and see our key checklist for BI modernization and data warehouse to Databricks article for the wider sequencing. If your Tableau version itself is aging, our note on Tableau's release cycle explains why upgrade planning should run alongside.

Best Practices for a Low-Risk Migration

Document business logic first. A catalog of calculated fields with definitions and owners is the most valuable artifact of the project.
Retire before you migrate. Never move workbooks or data sources nobody uses.
Modularize transformations into small, single-purpose SQL or PySpark steps that can be tested independently.
Use liquid clustering on new Delta tables, which Databricks recommends over traditional partitioning and ZORDER.
Apply Unity Catalog governance from day one, with least-privilege grants and row filters or column masks for sensitive data.
Choose the authentication model early, because it decides whether row-level security works per person.
Optimize for Tableau consumption by filtering early, avoiding unneeded columns, and exposing pre-aggregated gold views.
Reconcile before repointing and keep old and new pipelines running in parallel for a full cycle.
Keep the platform picture in view, including legacy systems to Databricks, data platform migration to Databricks, and cloud migration to Databricks.

Frequently Asked Questions

Can Databricks import Tableau workbooks?

Yes. The /importBI command in Genie Code reads Tableau .twb, .twbx, .tds, and .tdsx files and builds an AI/BI dashboard and metric views, which Databricks' release notes describe as generally available. The result is a starting point that still needs review and reconciliation.

Do Tableau dashboards have to be rebuilt?

Not if you repoint them. Workbooks can be pointed at curated Databricks tables, and keeping column names identical to the original field names minimizes broken references, while importing into AI/BI dashboards or rebuilding in another tool are separate, larger projects.

How do you connect Tableau to Databricks?

Use Tableau's native Databricks connector or Partner Connect against a SQL warehouse, with Databricks login (OAuth) recommended and OAuth required when publishing to Tableau Cloud. Personal access tokens and service principals are also supported, and basic username and password authentication ended in July 2024.

How is Tableau row-level security handled in Databricks?

Tableau user filters are recreated as Unity Catalog row filters and column masks, which apply at query time. They filter per person only when the connection carries each user's identity, such as per-user OAuth, because a shared service principal gives every viewer the same data scope.

What happens to Tableau extracts and Prep flows?

Extracts are replaced by Delta tables served through a Databricks SQL warehouse, though extracts stay optional where concurrency or offline use requires them. Prep flows are rebuilt as notebooks or Lakeflow pipelines run by Lakeflow Jobs, turning each step into versioned, testable code.

Scope Your Tableau Migration with DataTerrain

Moving Tableau logic onto Databricks is easy to start and hard to finish, because the effort hides in view-dependent calculations, security identity, and the numbers that must still match. DataTerrain brings 17+ years and 400+ clients in BI and data modernization, with automated discovery of workbook logic, conversion support, and figure-for-figure validation. Start with our Reports Conversion service or the wider Databricks migration services page.

Talk to our migration team