--- title: Convert from Tableau or Power BI description: Turn an existing Tableau workbook or Power BI report into a Dashies dashboard, and read the report that says exactly what did and did not come across. updated: 2026-08-04 tier: free --- Dashies can read a Tableau workbook or a Power BI report and rebuild it as a dashboard that refreshes from your connected warehouse on a schedule. The file is parsed **in your browser** and never uploaded. ## Convert a file :::steps ### 1. Open the converter Go to [dashies.xyz/app/convert](https://dashies.xyz/app/convert) and pick the tool your dashboard is in. If you start from a data source card instead, the converter builds against that warehouse automatically. ### 2. Choose your file ::::tabs{sync=bitool} :::tab{label="Tableau"} Accepts `.twb` or `.twbx`, up to **64 MB**. Refusals you may see: ```bash That is not a Tableau workbook. Choose a .twb or .twbx file (got report.pdf). That workbook is 91 MB - the limit is 64 MB. ``` ::: :::tab{label="Power BI"} Accepts `.pbix`, or a zipped PBIP project, up to **64 MB**. Refusals you may see: ```bash That is not a Power BI report. Choose a .pbix or a zipped PBIP project (got report.pdf). That report is 91 MB - the limit is 64 MB. ``` ::: :::: Qlik is listed but not available yet. ### 3. Read the conversion report Every element of the source is classified as **converted**, **approximated**, or **unsupported**, with a reason. **Nothing is ever silently dropped.** An unsupported element is reported with its original preserved, so the report is a complete account of the difference between the two files. ### 4. Publish, and verify the numbers Then [set a refresh schedule](/guides/refresh-schedule) and run [the number cross-check](/guides/verify-your-numbers). A converted dashboard needs that check as much as an authored one, more so where a measure was approximated. ::: ## What comes across The target chart vocabulary is deliberately small: bar, horizontal bar, line, and area, plus metric, table, text, and a single-select filter. **Converted, meaning a faithful equivalent:** - Tableau: bar, line, and area marks; text marks become a table; `Automatic` with one measure over a dimension becomes a line for a date and a bar for a category. - Power BI: bar, column, and clustered charts; line; area; cards and multi-row cards become one metric per value; a table visual becomes a table; a single-select slicer becomes a filter. **Approximated, meaning it renders but differently:** - Pie, donut, treemap, funnel, ribbon, and waterfall all become bar charts. - Tableau Gantt becomes a bar. - KPI and gauge become a metric, dropping the goal, trend, and target. - Pivot tables and matrices are flattened to a table. - Dual-axis, stacked, and multi-measure charts collapse to one series. - Multi-select filters are downgraded to single-select. - Layout comes across as relative weights, not absolute geometry. - Multiple Tableau dashboards, or a story, each become one Dashies page. **Unsupported, meaning reported rather than rendered:** - Scatter plots comparing two measures. - All maps, in both tools. - Range, relative-date, and top-N filters. There is no range control. - Parameter controls. - Colour-by-dimension encoding. The runtime is single-accent by design. - Conditional formatting. - Custom marketplace visuals. - `StdDev`, `Var`, percentile, and `Mode` aggregations. ## Two things worth knowing before you start :::warning{title="Formulas are never executed and never become cube SQL"} A simple aggregate such as `SUM([Sales])` maps to a measure. An arithmetic ratio such as `SUM(a)/SUM(b)` becomes a ratio metric, which is an approximation. Everything else is preserved as an annotation only, not computed: Tableau level-of-detail expressions (`{FIXED}`, `{INCLUDE}`, `{EXCLUDE}`), table calculations (`WINDOW_*`, `RUNNING_*`, `INDEX()`, `RANK()`), and named DAX measures. If your report's headline number comes out of one of those, you will need to express it in cube SQL yourself. See [Author a dashboard with your AI](/guides/author-a-dashboard). ::: :::danger{title="Power BI cannot do averages, distinct counts, or medians"} This is the sharpest difference between the two converters, and it looks identical in the UI. A `.pbix` keeps its data in a compressed model the browser-side converter never unpacks, so there is no readable extract to fall back on. A **non-additive** Power BI aggregate, meaning `Average`, `DistinctCount`, or `Median`, is therefore reported as **unsupported**. The Tableau path can handle the same aggregates, because a `.twbx` can carry a readable data extract. Same word, two behaviours. If someone tells you "averages are supported", check which tool they meant. ::: ## Where the data comes from afterwards | Source | Result | |---|---| | Tableau, bound to a matching Dashies data source | Live. It refreshes from your warehouse. | | Tableau `.twbx` with a bundled extract, no matching data source | A snapshot of the bundled data, frozen. | | Power BI, bound to a matching data source | Live. | | Power BI, no matching data source | Unresolved. There is no snapshot fallback. | | Either tool, no match found | Unresolved. | To get a refreshing dashboard rather than a frozen one, [connect the warehouse first](/guides/connect-warehouse), then convert. ## What is deliberately stripped Web-page zones and image zones are removed rather than rendered. A published dashboard that embedded a remote frame or a remote image would be a way to reach back into networks or to track viewers, so those never make it into the output. Spreadsheets bundled inside a `.twbx` are not parsed at all, and Tableau's `.hyper` and `.tde` extract formats are unsupported. ## A known limitation A converted single-value filter still opens **unfiltered**, because the runtime has no pre-applied filter default yet. The filter works once you use it; it just does not start in the state the source file had. ## Check it worked 1. **Read the conversion report before publishing.** Every approximated and unsupported item is listed. If a headline number is in that list, the dashboard is not yet equivalent to the original. 2. **Compare one figure against the original tool.** Open the source report, pick the top-line metric, and check it matches. 3. **Run [the number cross-check](/guides/verify-your-numbers)** on every measure a tile displays. An approximated ratio is exactly the shape that check exists to catch. 4. **Confirm it refreshes.** Open [dashies.xyz/app/schedules](https://dashies.xyz/app/schedules): the dashboard should show a cadence and a next run. If it does not, the data binding came out unresolved and the page is a snapshot, not a live dashboard.