What you will be able to do
- Tell apart the Data tab and the Canvas of an AI/BI dashboard, and explain how a draft relates to the published version
- Add, rename, clone, delete, hide and copy dashboard pages to build a multi-page (multi-tab) layout
- Define several datasets from different data sources, and manage them safely
Key concept
Data tab + multi-page Canvas — An AI/BI dashboard has two parts. Datasets are defined once on the Data tab. Widgets are laid out on a Canvas that can be split into several pages, and any page can draw on any dataset. You build a dashboard by deciding which datasets feed which widgets on which pages.
1.Two halves: the Data tab and the Canvas
Every AI/BI dashboard has two parts. The Data tab is where you define the datasets the dashboard queries. The Canvas is where you place widgets. The documentation lists four kinds of widget: visualizations, filters, text and images. This lesson covers visualizations, text and images. Filters belong to a separate objective.
Datasets are part of the dashboard itself, not loose queries kept somewhere else. When you share, import or export a dashboard, its datasets go with it. One dashboard can hold many datasets, and each one can come from a different source. A single report can therefore combine a sales table, a customer view and an uploaded file.
Checkpoint 1 of 6· Check yourself
You export an AI/BI dashboard and import it into another workspace. What happens to the datasets defined on its Data tab?
Datasets are bundled with the dashboard when it is shared, imported or exported, whether through the UI or the API.
“Datasets are bundled with dashboards when sharing, importing, or exporting them using the UI or API.”Source: docs.databricks.com
Building happens in a draft. To create a dashboard, click New in the sidebar and select Dashboard, or click Create on the Dashboards listing page. Edits to the draft save automatically, but they do not reach the published version until you publish again. If you want to throw away your edits, open the kebab menu and choose Discard changes. This resets the draft to the most recently published version. You can also share a draft with other users in your workspace so they can edit it with you, but not with anyone outside the workspace.
No. Draft changes save automatically, but they do not sync with the published version. Viewers see the fix only after the dashboard is published again.
2.Building a multi-page (multi-tab) layout
When the exam guide says "multi-tab", the documentation says pages. Each page appears as a tab in the *page tab bar* along the top of the canvas. A new dashboard starts with one page called *Untitled page*. To work with pages:
- Add: click the plus icon to the right of the current page title. - Rename: double-click the title. If the name is already taken, a number is added automatically. - Delete: open the kebab menu next to the page title and click Delete. This also deletes every widget on that page. - Clone: select the page, then choose Clone. You get an exact copy, widgets included.
Cloning a page and copying it to another dashboard treat datasets differently. A clone stays in the same dashboard, so it reuses the same datasets and they remain unchanged. A copy pasted into another dashboard brings its own copies: every dataset the page uses is created again in the target dashboard. To copy a page, use the normal copy and paste shortcuts (Command-C/V on Mac, Ctrl-C/V on Windows/Linux). The target dashboard must be in draft mode.
Checkpoint 2 of 6· Put it in order
Put the steps for copying a page from one dashboard to another in order.
- 1.Press Command-C (Mac) or Ctrl-C (Windows/Linux) to copy the page
- 2.In the source dashboard, click the page name in the page tab bar to select it
- 3.Navigate to the target dashboard and make sure it is in draft mode
- 4.Press Command-V (Mac) or Ctrl-V (Windows/Linux) to paste the page
You select and copy the page in the source dashboard, then paste it into a target dashboard that is in draft mode.
“Navigate to the target dashboard and ensure it's in draft mode.”Source: docs.databricks.com
Sometimes authors want to keep a page without showing it to viewers, for example draft analysis or notes. Open the page's kebab menu and choose Hide from published dashboard. The page stays in the draft for authors and is left out of the published dashboard. To bring it back, choose Show in published dashboard. The documentation sets a maximum number of pages but leaves the actual figure to its Dashboard limits page.
Checkpoint 3 of 6· Exam question
In an AI/BI dashboard, which widget type is required to embed a static image, such as a company logo, directly onto the canvas?
Correct answer: A — A text widget
- A. A text widget accepts markdown content and is the widget type that supports embedding images on the dashboard canvas.
- B. A visualization widget renders chart output from a dataset and does not provide a way to embed a standalone image file.
- C. There is no dataset widget type on the canvas; datasets are data definitions that power widgets rather than a visible widget themselves.
- D. A filter widget lets viewers control parameter or column values and has no mechanism for displaying an embedded image.
Sources1
3.Adding datasets from multiple data sources
Datasets are added on the Data tab, and there are two ways to do it. Add data lets you pick a Unity Catalog object, create a local metric view, or upload a file. Add SQL dataset opens an editor where you write your own SQL against any data source.
| Option | What it gives you |
|---|---|
| Add data → Select a Unity Catalog object | A table, view or metric view. Most tables and views start with a default SELECT * query that you can edit |
| Add data → Create a local metric view | Dimensions, measures and joins that belong to this dashboard only and are not published to Unity Catalog |
| Add data → Upload file | Uploads a file to a chosen Catalog, Schema and Table name, either creating a new table or overwriting an existing one (overwriting needs MANAGE on the schema) |
| Add SQL dataset | A new SQL query you write yourself, against any data source |
Each SQL dataset finds its tables in one of two ways. You can write a fully-qualified table name in the query, or you can pick a catalog and schema from the drop-down selectors and write just the table name. If you do both, the fully-qualified name in the query wins. Put backticks around table or column names that contain spaces. Dataset queries are read-only. DESCRIBE, EXPLAIN, CREATE TABLE, INSERT and DELETE are not allowed, because a dataset query must be something that can be wrapped in additional SQL.
Because viewers can reach every column in a dataset, Databricks recommends avoiding SELECT *. Select only the columns you need, filter rows with WHERE clauses, and check the Result Table schema to confirm that only the intended fields are there.
Once a dataset exists, its kebab menu gives you Rename, Suggest name, Clone, Delete and Download (as CSV, TSV or Excel). Datasets that are in use on the canvas appear in bold with a blue icon. You cannot delete a dataset while a widget on the canvas is using it.
Checkpoint 4 of 6· Check yourself
You try to delete a dataset, but the Delete option is greyed out. What is the most likely reason?
Delete is not available while the dataset is in use on the canvas.
“Delete: Delete a dataset. This option is unavailable if the dataset is being used on the canvas.”Source: docs.databricks.com
Checkpoint 5 of 6· Check yourself
A new SQL dataset starts with CREATE TABLE tmp AS SELECT .... What happens?
Commands such as CREATE TABLE, INSERT and DELETE cannot define a dataset. Only queries that can be wrapped in additional SQL are valid.
“Only queries that can be wrapped in additional SQL are valid.”Source: docs.databricks.com
Checkpoint 6 of 6· Exam question
A data analyst is building an AI/BI dashboard that needs a high-level executive summary view and a separate, more detailed regional breakdown view, each with its own set of visualizations and filters that should not affect the other view. What should the analyst do?
Correct answer: A — Add a second page to the dashboard and place the regional breakdown widgets and filters there, separate from the summary page.
- A. AI/BI dashboards support multiple pages, and moving the regional widgets and their filters to a second page keeps each view's layout and filter scope independent.
- B. Duplicating into a separate dashboard file breaks the single-dashboard requirement and forces the analyst to maintain two disconnected objects instead of one cohesive dashboard.
- C. A widget-level filter only changes what one widget displays; it cannot hide an entire set of summary widgets to simulate a second view on one page.
- D. Dashboards do not auto-organize widgets into views based on dataset choice; page placement is a layout decision the analyst makes explicitly.
Sources3
Exam traps
Each one states something that sounds right. Open it to see what is actually true.
1.Copying a page into another dashboard keeps it linked to the original dashboard's datasets.Why is that wrong?
Every dataset the pasted page uses is created again as a new dataset in the target dashboard, so the page works on its own. Only a clone within the same dashboard reuses the existing datasets.
Covered in Building a multi-page (multi-tab) layout
2.Hiding a page deletes it, or hides it from authors as well.Why is that wrong?
A hidden page is only left out of the published dashboard. Authors still see it in the draft.
Covered in Building a multi-page (multi-tab) layout
3.Columns that no chart uses are invisible to viewers, so SELECT * is harmless.Why is that wrong?
Viewers can access all data in a dataset, so you should limit the columns in the query itself.
Covered in Adding datasets from multiple data sources
Sources
Every claim above is drawn from one of these pages, quoted as it was written on the date shown.
- 1.https://docs.databricks.com/aws/en/dashboards/manageOfficial docs
“Dashboard editors can build and configure their dashboards by adding widgets such as visualizations, filters, text, and images.”
↩︎ Two halves: the Data tab and the Canvas“Changes to a draft dashboard are saved automatically, but they do not automatically sync with the published version if it exists.”
↩︎ Two halves: the Data tab and the Canvas“New dashboards start with a single page named Untitled page.”
↩︎ Building a multi-page (multi-tab) layout“Naming conflicts resolve automatically by appending a number to the title.”
↩︎ Building a multi-page (multi-tab) layout“Deleting a page also deletes all of the widgets on that page.”
↩︎ Building a multi-page (multi-tab) layout“The new page is an exact copy of the original, including all widgets. The underlying datasets remain unchanged.”
↩︎ Building a multi-page (multi-tab) layout“The Canvas tab can be organized into multi-page reports.”
↩︎ Key concept“All datasets used by the page are created as new datasets in the target dashboard, ensuring the page functions independently.”
↩︎ Exam trap 1“Hidden pages remain visible to authors in the draft dashboard and are excluded when you publish.”
↩︎ Exam trap 2“Datasets are bundled with dashboards when sharing, importing, or exporting them using the UI or API.”
↩︎ Checkpoint“All datasets used by the page are created as new datasets in the target dashboard, ensuring the page functions independently.”
↩︎ Prediction“Navigate to the target dashboard and ensure it's in draft mode.”
↩︎ Checkpoint - 2.
“Click New in the sidebar and select Dashboard.”
↩︎ Two halves: the Data tab and the Canvas - 3.
“Add SQL dataset: Use this option to write a new SQL query on any data source.”
↩︎ Adding datasets from multiple data sources“A fully-qualified table name in the query overrides the catalog and schema selectors in the editor.”
↩︎ Adding datasets from multiple data sources“Dashboard queries are read-only.”
↩︎ Adding datasets from multiple data sources“Datasets currently in use on the canvas appear in bold text with a blue icon”
↩︎ Adding datasets from multiple data sources“Dashboard viewers can access all data in a dataset, even if it's not shown in visualizations.”
↩︎ Exam trap 3“Dashboard viewers can access all data in a dataset, even if it's not shown in visualizations.”
↩︎ Prediction“Delete: Delete a dataset. This option is unavailable if the dataset is being used on the canvas.”
↩︎ Checkpoint“Only queries that can be wrapped in additional SQL are valid.”
↩︎ Checkpoint