Tableau to Sigma Migration: The Practical Guide for 2026

Moving a Tableau dashboard to Sigma means putting every table it reads into a cloud data platform Sigma connects to, rebuilding its published data source as a Sigma data model, rewriting its calculated fields, LOD expressions and table calculations as Sigma formulas, and rebuilding its filters, parameters and actions as controls and actions on a workbook page. Sigma queries the data platform directly, so the order of work differs from a move between two tools that both keep their own extracts. One naming change matters before you start: Sigma retired datasets, its older modeling object, in 2026. Since June 2 nobody can create or edit one, and since September 15 a dataset can no longer serve as a data source, so every migration now lands on data models.

Sigma next to Tableau: what changes for dashboard owners

The difference that drives everything else is where the data lives. Tableau can work from a live connection or from a .hyper extract that Tableau itself stores and refreshes, and Tableau Desktop opens files and databases on a laptop. A Sigma workbook reads the data platform behind it; the workbook documentation puts it as "your data is always live". Where Sigma keeps a copy, it is a table in your own platform: materialization writes an expensive result back to the warehouse, and a CSV upload becomes a table in the schema Sigma uses for write‑back. Calculations are formulas on columns, measures that should be shared live as metrics in a data model, and row‑level security runs on user attributes or on the data platform's own rules.

The counterweight is real. A Tableau estate built on extracts of files, spreadsheets or on‑premises databases has a data project ahead of the dashboard work, because Sigma has nothing to read until those tables exist in a supported platform. Tableau Prep flows move into pipelines or views in that platform, and stories have no documented Sigma equivalent. Many teams run both tools for a while and move the dashboards whose data already sits in the warehouse first; this page is about the dashboards you decide to move.

What a Tableau dashboard becomes in Sigma

Three Sigma words first. A workbook holds one or more pages; an element is anything placed on a page, from a table, pivot table or chart to a control, text or a container; a data model holds the tables, relationships, metrics and security rules that workbooks build on. In the tables, "converts" means the construct exists on both sides, "with review" means a person decides something, and "redesign" means there is no object to convert into.

Converting LOD expressions and table calculations

A Tableau calculation is evaluated at the level of detail of the view, unless an LOD expression or a table calculation's addressing says otherwise. Sigma states the level explicitly: a table can carry several grouping levels, and an aggregate added to a grouping is calculated once per group. Most of the work is choosing the grouping that reproduces the number Tableau showed.

Two traps account for most mismatches. Filter order: a FIXED value in Tableau ignores ordinary dimension filters, while an aggregate in a Sigma grouping sees whatever filters the element applies, so a control that should not move a fixed total has to target a different element or be applied after the grouping. Addressing: a table calculation computed "across" in Tableau and the same function over a Sigma grouping agree only when the grouping and sort match the Tableau partition. The LOD expressions to DAX and LOD expressions to Databricks SQL pages work the same filter‑order logic for other targets.

Why the data moves before the dashboards

A packaged workbook can carry its own data: a .twbx is a zip of the workbook plus local files and extracts. Sigma has no equivalent container, so every extract, spreadsheet and file source becomes a table in the data platform before its dashboards can move. Sigma's own migration guide says so directly: data that exists only in an extract or an imported model has to be loaded into your data platform before you migrate. Two side effects follow. An extract that was refreshed nightly and a live query will not show the same numbers during validation unless both are compared at the same refresh point. And the refresh schedule, the gateway and the extract size limits go away, replaced by the warehouse's own cost and performance, with materialization for the results that are expensive to recompute.

Row‑level security becomes user attributes or platform rules

Tableau user filters and entitlement tables usually compare USERNAME() or group membership with a column. In Sigma the rule goes into the data model as a filter on CurrentUserAttributeText("region") or CurrentUserInTeam(…), so every workbook built on that model inherits it. With OAuth connections, Sigma respects the row‑level security already defined in the data platform, which lets a team keep one rule set for every tool. Sigma's documentation warns never to use a control for security. Validate as several real users, not as the author.

Sigma's own migration skills, and where a migration tool fits

Sigma publishes open‑source migration skills for AI coding assistants. Its help page, Migrate to Sigma with an AI assistant, labels them public beta. The tableau-to-sigma skill converts data sources and workbooks into a Sigma data model and workbook, and a tableau-assessment skill inventories a site first; both live in the sigma-migration-skills repository under the Apache 2.0 license. Running them takes an assistant that supports skills, such as Claude Code, Cursor, Cortex Code or Codex, Sigma API credentials, a Sigma connection to the data platform and credentials for the Tableau site. Sigma's QuickStart starts from a dashboard on Tableau Cloud or Tableau Server, and lists dashboard actions, drill hierarchies and reference lines as manual work after publishing. The skills check that summary values match between source and target, and Sigma notes that this check cannot catch everything and that AI assistants can make mistakes, so layout, chart types, sort order and filters are compared by hand. For a team comfortable running an AI assistant against its own Sigma organization, this is a capable free starting point.

Phases and validation

Reconcile at the grand total first, then under the filter combinations people actually use, then per user attribute. Fixed totals, percent of total and running sums are where the two tools part ways, and they are worth a check on every dashboard that has them.

What Antares does on this route

Everything above holds whichever tool you use; this section is about Antares, a BI migration tool. Tableau to Sigma is in private beta. The free Analyzer already lists Sigma as a target, marked Beta: upload a .twb or .twbx workbook, or connect Tableau Server or Tableau Cloud, choose Sigma, and it scores how hard each workbook is to move there. The Analyzer reads workbook structure, not the data in it.

Conversion on this route is not self‑serve; the Converter page takes requests for beta access, and the home page keeps the current status of every route. In the beta, one Tableau dashboard becomes a Sigma data model and a workbook page. An LLM drafts each calculated field as a Sigma formula, and code checks it against the meaning of the source calculation before it is accepted; a calculation the check cannot prove is listed for manual work instead of being shipped. Quick filters and parameters become Sigma controls, and blended sources keep their links. A conversion counts as done only when every element exports real rows from Sigma and the page renders. The beta does not yet convert Tableau relationships or several dashboards into one workbook, and it deploys through Databricks: the tables a dashboard reads land as Delta tables, and Sigma reads them through its Databricks connection. Run the free Analyzer on a workbook to see what moves.

If your source is Power BI

The shape is the same and the vocabulary shifts. A Power BI semantic model becomes a Sigma data model, DAX measures become metrics and formulas, report pages become workbook pages, and RLS roles become user‑attribute rules or the data platform's own security. Import‑mode data has to be landed in the platform first, for the same reason extracts do. Sigma's powerbi-to-sigma skill, in the same public beta, converts reports and semantic models including DAX measures. On the Antares side, the free Analyzer takes a .pbix file or a zipped PBIP project and scores how hard it is to move to Sigma, also marked Beta. Converting Power BI reports to Sigma is not available yet; the home page lists the route as coming soon, and the Power BI to Tableau and Power BI to Databricks AI/BI guides cover the routes that are further along.

Related reading: Tableau to Power BI, Tableau to Databricks AI/BI and BI migration approaches. Primary sources: Sigma on workbooks, data models, metrics, window functions, Subtotal, controls, actions, row‑level security and AI‑assisted migration; Tableau on LOD expressions and workbook and extract files.