Tableau to Microsoft Fabric: The Technical Migration Playbook for Reporting
Tableau extracts, calculations and filters move through OneLake into a semantic model with DAX and row-level security on Microsoft Fabric.
Tableau to Microsoft Fabric is a redesign, not a copy. A workbook bundles extracts, calculated fields, filters and dashboards, and each part has a different home in Fabric and Power BI. This playbook maps that technical work for reporting, from LOD expressions to DAX, extracts to a lakehouse and user filters to row-level security.
It is written for BI leaders and architects on Microsoft 365 or Azure who have decided to move and need to know what the work involves. If you are still building the case, start with our legacy BI total cost comparison.
Triage Your Tableau Estate by Usage First
Do not migrate what nobody opens. Pull usage data from Tableau Server or Cloud, then rate each workbook by views, owner, data source and calculation complexity. Retire stale content, merge near duplicates and rebuild only what earns its place. Microsoft's Power BI migration guidance makes the same point: focus on the business questions a report answers, not on reproducing every visual.
Group what remains into data domains such as sales or finance, and give each domain one shared semantic model. Many single-report files are a sign of poor design.
Tableau to Microsoft Fabric: Extracts to a Lakehouse and Direct Lake
A Tableau extract is a stored snapshot, often with joins and filters hidden inside it. In Fabric, land the source data in OneLake as Delta tables in a lakehouse, shape a star schema there and connect a semantic model. In Direct Lake mode the model reads Delta tables from OneLake, and a refresh only updates metadata, so large tables avoid a full import copy.
Use Import mode where you need sources outside OneLake, and check Direct Lake limits for your tables before you commit. Move extract filters and joins upstream so every report shares one version of the logic.
Calculated Fields and LOD Expressions to DAX
Most calculated fields translate directly. Row-level logic becomes a calculated column or an upstream transformation, and aggregations become measures. LOD expressions need more care. Tableau's documentation says FIXED expressions ignore view filters other than context, data source and extract filters, while INCLUDE and EXCLUDE are evaluated after dimension filters. A DAX measure responds to filter context instead, so translation is about controlling that context with CALCULATE, REMOVEFILTERS and VALUES.
| Tableau | DAX pattern (illustrative) | Note |
|---|---|---|
| FIXED first order date | CALCULATE ( MIN ( Sales[Order Date] ), REMOVEFILTERS ( 'Date' ) ) | Customer stays in context, date filters lift |
| INCLUDE customer, average | AVERAGEX ( VALUES ( Customer[Customer] ), [Total Sales] ) | Iterates a finer level |
| EXCLUDE region | CALCULATE ( [Total Sales], REMOVEFILTERS ( Geography[Region] ) ) | Removes one dimension |
| Percent of total | DIVIDE ( [Total Sales], CALCULATE ( [Total Sales], REMOVEFILTERS ( Product ) ) ) | Denominator ignores Product |
| Rank | RANKX ( ALLSELECTED ( Product[Product] ), [Total Sales] ) | Respects slicers |
| Running total | RUNNINGSUM ( [Total Sales] ) | Visual calculation, on the visual only |
These are starting patterns, not a converter. Test every translation against real Tableau output.
A worked example, illustrative only. A Tableau view groups sales by customer cohort, defined as the year of each customer's first order with FIXED [Customer] and MIN([Order Date]). A measure cannot be a chart axis, so the cleaner Fabric build computes first order date once in the lakehouse, adds it to the customer table and groups by its year. The pattern is the same either way: the customer stays in context and the date filter is lifted.
Parameters, Sets, Groups and Tableau Prep
Tableau parameters usually become what-if or field parameters. Sets and groups become calculated columns, a mapping table or a measure flag. Microsoft documents that calculation groups, what-if parameters and field parameters work in Direct Lake models.
Tableau Prep flows move to Dataflows Gen2 for low-code shaping, with pipelines, notebooks or SQL for heavier work. Dataflow Gen2 can write straight to Lakehouse tables, which keeps one governed layer under every report.
Tableau User Filters to Row-Level Security
Tableau separates permissions, which control content, from row-level security, which controls the rows each person sees. User filters built on functions such as USERNAME() map to Power BI roles. The strongest pattern is a security table of users and entitlements, related to the data, with a role filter on USERPRINCIPALNAME().
Microsoft states that RLS applies to workspace Viewers, not Admins, Members or Contributors. The Test as role feature uses your own identity, so test dynamic RLS by signing in as real Viewer accounts. Map roles to Microsoft Entra security groups where you can.
- Tableau to Microsoft Fabric is a redesign: triage by usage and retire before you rebuild.
- Move extract logic into a lakehouse, then use Direct Lake or Import for the model.
- Translate LOD expressions by controlling filter context, and test each one.
- Rebuild Tableau user filters as dynamic RLS and test with real Viewer accounts.
- Run both platforms in parallel and prove parity before cutover.
Redesign the Dashboards, Then Run in Parallel
Tableau dashboards lean on actions, filters and dense layouts. Rebuild them with Power BI slicers, cross-filtering, drill-through and bookmarks, and aim for the business question, not pixel parity. For print-ready output, use paginated reports, as in our guide to Crystal Reports to Power BI paginated reporting.
Run Tableau and Power BI side by side through at least one full reporting cycle. Compare totals, row counts and role-filtered views, and have each business owner sign off. Plan waves by estate size.
| Estate size | Typical signals | Approach |
|---|---|---|
| Small | A few dozen active workbooks, one or two sources | One wave, one semantic model |
| Mid-size | Hundreds of workbooks across several domains | Waves by domain, shared models, RLS built once |
| Large | Multiple sites, heavy LOD use, many data sources | Platform first, then domain waves with inventory automation |
This is a planning guide only. Confirm waves after the inventory.
Tableau to Microsoft Fabric FAQ and Next Steps
Can Tableau workbooks be converted automatically?
Not reliably. Treat any converter output as a draft, because calculations, layouts and security still need rebuilding and testing against the original.
Do I have to use Direct Lake?
No. Import mode stays valid. Direct Lake suits large lakehouse tables where a full import copy is the problem.
What happens to Tableau Prep flows?
Rebuild them as Dataflows Gen2 for low-code shaping, or as pipelines, notebooks or SQL where the work is heavier.
How do we prove parity?
Compare row counts, totals and role-filtered views against Tableau for each report, then have the business owner sign off.
Our Tableau to Fabric migration service follows this playbook, backed by Tableau consulting, Power BI consulting and Microsoft Fabric migration experience. The same approach applies to SAP BusinessObjects reporting, and our guide to migrating to Microsoft Fabric covers the wider program.
A short Tableau migration assessment confirms your estate, your waves and your first model. Speak with a certified consultant to turn this playbook into a plan for reporting you can trust.