Building AI/BI Dashboards and Visualizations

What it is

An AI/BI Dashboard is the Databricks dashboard for sharing insights. It has a Data tab, where you define datasets from Unity Catalog tables, views or SQL queries, and a canvas of one or more pages, where widgets display those datasets.

Quick charts can also be made straight from a result in a notebook or the SQL editor. Parameters let readers change the values a query uses, such as a date range or region, without editing SQL.

How it works

Datasets

Each dataset is a table, a view or a custom SQL query, added on the Data tab, for example with Add SQL dataset. A dashboard can hold several datasets from different tables, and each visualization reads from one dataset.

Put joins and heavy logic in the dataset query or in a view, then point several widgets at the same dataset. That keeps the canvas simple and lets widgets share results.

Pages and widgets

A dashboard can have multiple pages, shown as tabs; add one with the plus icon next to the current page title. Use pages to separate audiences or subjects, such as a summary page and a detail page.

Add widgets from the toolbar at the bottom of the canvas. Filters can apply to the whole dashboard, one page or one widget, and clicking a value in one chart can cross filter the others.

Visualizations in notebooks and the SQL editor

Run a query, click + above the result and choose Visualization. Pick the Visualization Type, choose the columns to plot and click Save; the chart appears as a tab beside the results.

In a notebook, any SQL cell result, or display(df) in Python, gives the same option. From the tab menu, Add to dashboard sends a chart from the SQL editor or a notebook SQL cell to an existing or new AI/BI dashboard.

Parameters in queries

A named parameter marker is a colon and a name. It works in the SQL editor, notebooks, the dashboard dataset editor and Genie Agents.

SELECT region, SUM(amount) AS sales
FROM main.sales.orders
WHERE order_date BETWEEN :dates.min AND :dates.max
  AND region = :region
GROUP BY region;

Typing the marker adds a parameter widget where you set the type and a value. Types include String, numeric, Date and Date and Time (timestamp); the date types support a Range read with .min and .max. String values are quoted and escaped for you, so do not add quotation marks around the marker.

Use IDENTIFIER(:tbl) when the parameter names a table or column rather than a value.

Parameters in dashboards

On the Data tab, click Add parameter or type the marker. The gear icon opens Parameter details to set the display name and type, and the field under the name holds the default value.

Test by running the dataset query on the Data tab: the run uses the default value and confirms the query works. Then add a filter widget on the canvas and add the parameter under Parameters in its settings.

For a multi value pick list, select Allow multiple selections and change the query so it handles a list and an empty selection.

WHERE ARRAY_CONTAINS(:region, region) OR :region IS NULL

A default set in the filter widget overrides the default set in the dataset editor.

Choosing a visualization type

Pie charts become hard to read with many slices, and a bar chart sorted by value compares categories more clearly.

How the exam asks about this

• "Show monthly revenue across two years so the trend is clear." A line chart with month on the x axis.
• "Let readers pick a date range that is used inside a subquery." A date range parameter with .min and .max, connected to a filter widget.
• "Add a company logo and a short explanation to the dashboard." A text box widget with Markdown, including an image.