Section 6: Working with Dashboards and Visualizations in Databricks¶

Weighted 16%, about seven questions. Seven objectives: build AI/BI dashboards; make visualizations in notebooks and the SQL editor; work with parameters; configure sharing and embedding; schedule refresh; configure alerts; and choose the right visualization type.

Table of Contents¶

1. Building AI/BI dashboards
2. Visualizations in notebooks and the SQL editor
3. Parameters
4. Sharing, embedding and refresh
5. Alerts
6. Choosing a visualization
7. How the exam tests this section
8. Key terms

Building AI/BI dashboards¶

An AI/BI dashboard is built from datasets (a table, a view or a custom query), pages (tabs for multi-page layouts) and widgets (visualizations, text and images). Because one dashboard can have many datasets, a single page can mix data sources. Filters work at global, page and widget level with cross-filtering and drill-through, and you can add custom calculations without changing the dataset. AI assistance helps author visualizations.

📖 AI/BI dashboards - datasets, filters, sharing, embedding and refresh

Visualizations in notebooks and the SQL editor¶

You do not need a dashboard to see a chart. In a notebook, results from a SQL or Python cell can be turned into a chart with the + button on the result. In the SQL editor, add a visualization to a query result the same way. These are exploration tools; when the audience is a team, publish a dashboard.

Parameters¶

A parameter substitutes a value into a dataset query at run time, using the :keyword marker. A date-range parameter uses .min and .max:

-- Syntax per the Databricks docs; not run (needs a dashboard dataset).
SELECT order_date, SUM(amount) AS revenue
FROM main.gold.orders
WHERE order_date BETWEEN :date_param.min AND :date_param.max
GROUP BY order_date;

To define one: type the marker (or Add parameter), use the gear icon to set the keyword, display name and type (String, Date, Date and Time, Numeric, Date Range, or multiple selection), set a default and run the query; the default is saved. Then connect it to a filter widget on the canvas so viewers can change it. Parameters not exposed on the dashboard use their defaults. Parameters change the query itself; field filters are more efficient for large data and cascade across filters, so prefer filters unless you need to alter the query logic.

📖 Dashboard parameters - syntax, types, widgets, filters versus parameters

Sharing, embedding and refresh¶

• Publish the dashboard with either shared data permissions (viewers see data through the publisher's credentials) or individual data permissions (each viewer's own access applies).
• Share with workspace users and groups, or at the account level; published dashboards can be shared with anyone in the Databricks account, even without workspace access.
• Embed a published dashboard in an external site with an iframe.
• Refresh and delivery: set an automated refresh schedule, and send email or Slack subscriptions.

📖 AI/BI dashboards - publishing, sharing, embedding, schedules and subscriptions

Alerts¶

A SQL alert runs a query on a schedule and notifies you when a condition is met. Each evaluation ends as OK, TRIGGERED or ERROR. When you create one you choose a column (the first value or an aggregation across rows such as SUM or AVERAGE), an operator and a threshold, and Test condition shows whether it would trigger now. Set the schedule (Quartz cron syntax is available), then choose the users or notification destinations; you can repeat notifications until the status returns to OK, and optionally notify on OK. Workspace admins create the notification destinations, such as webhooks. An alert can also run as a task in Lakeflow Jobs so downstream tasks branch on the result.

📖 SQL alerts - states, query ownership and webhook payloads 📖 Create an alert - condition, schedule and notifications

Use alerts for "tell me when" (average temperature above a limit in the last 15 minutes); use dashboards for "let me look". Refreshing a dashboard by hand, reading job logs or resizing a cluster are not alerting.

Choosing a visualization¶

Pick the chart from the question, not from habit:

How the exam tests this section¶

• Parameter mechanics: the date widget's min and max values flow into the query's WHERE; they are not just axis settings.
• Sharing: shared versus individual data permissions; embedding through iframes; account-level sharing.
• Alert versus dashboard versus job: a threshold with immediate notification is an alert.
• Chart choice: trends are lines, comparisons are bars.

Key terms¶

• AI/BI dashboard - a Databricks dashboard built from datasets, pages and widgets, with AI-assisted authoring.
• Dataset - a table, view or custom query that supplies data to a dashboard.
• Widget - a dashboard element such as a visualization, filter, text box or image.
• Parameter - a named placeholder, written as :keyword, that substitutes a value into a query at run time.
• Field filter - a dashboard filter on a dataset column that cascades across filters and performs well on large data.
• Shared data permissions - a publishing mode where viewers see data using the publisher's credentials.
• SQL alert - a scheduled query with a condition that notifies users or destinations when the condition is met.
• Notification destination - a place alerts are sent, such as email or a webhook, created by workspace admins.