Databricks Dashboards

A dashboard is a single page of charts, numbers, and filters that answers business questions at a glance. Databricks AI/BI dashboards run on your lakehouse tables, so viewers see fresh data without exporting files. Analysts build a dashboard once, share it with a team, and refresh it on a schedule.

The Car Dashboard Analogy

A car dashboard shows speed, fuel, and warnings in one glance. The driver never opens the engine to learn the fuel level. A data dashboard works the same way for a business. A manager reads revenue, orders, and delays without reading raw tables.

ANATOMY OF A DASHBOARD
──────────────────────────────────────────────────────────
┌────────────────────────────────────────────────────────┐
│  FILTERS:  [Region ▼]  [Date range ▼]                  │
├──────────────┬──────────────┬──────────────────────────┤
│ Total Sales  │ Orders       │ Average Order            │
│  (counter)   │  (counter)   │  (counter)               │
├──────────────┴──────────────┴──────────────────────────┤
│  Sales by Month  (line chart)                          │
├─────────────────────────────┬──────────────────────────┤
│  Sales by Region (bar)      │  Top Products (table)    │
└─────────────────────────────┴──────────────────────────┘

Building Blocks

Datasets

A dataset is a saved SQL query that feeds the dashboard. Each chart reads from a dataset. Write one dataset per theme, such as sales, returns, or customers, and reuse it across many charts.

Visualizations

A visualization turns dataset columns into a visual. Common types include counters, bar charts, line charts, pie charts, maps, and tables. Pick the type that matches the question: trends over time suit lines, and comparisons across groups suit bars.

Filters and Parameters

A filter lets viewers narrow the data without touching SQL. A date filter, a region filter, and a product filter cover most needs. A parameter inserts a viewer's choice directly into a dataset query.

Pages

A dashboard holds several pages. One page can show sales, and another page can show delivery performance.

Create Your First Dashboard

1. Open the Dashboards page from the sidebar and choose Create dashboard.
2. Open the Data tab and add a dataset with a SQL query.
3. Open the Canvas tab and add a visualization.
4. Pick the dataset, the chart type, and the columns for each axis.
5. Add a filter widget and connect it to the fields it should control.
6. Publish the dashboard and share it with your team.

Writing Good Datasets

A dashboard runs only as fast as its queries. Summarize the data in the dataset and avoid pulling millions of raw rows into a chart.

SELECT DATE_TRUNC('month', order_time) AS month,
       region,
       SUM(amount)                      AS revenue,
       COUNT(*)                         AS orders
FROM main.shop.silver_orders
GROUP BY DATE_TRUNC('month', order_time), region;

Point datasets at gold tables whenever possible. Gold tables already hold clean summaries, so queries finish in seconds.

Choosing the Right Chart

Question	Best Chart
How did sales change over time?	Line chart
Which region sells the most?	Bar chart
What is the current total?	Counter
Where do customers live?	Map
Which exact orders need review?	Table

Sharing and Refreshing

Publishing a dashboard creates a version that viewers can open. You can embed credentials so viewers see data without direct table access, or you can let each viewer use their own permissions. Embedded credentials suit broad audiences, and viewer permissions suit sensitive data.

A schedule refreshes the dashboard automatically, such as every morning at seven. Subscriptions send a snapshot to chosen email addresses on that schedule.

PUBLISH AND REFRESH CYCLE
──────────────────────────────────────────────────────────
Draft (author edits) ──► Publish ──► Viewers open the page
                                           ▲
                      Schedule refreshes data every morning
                                           │
                      Email snapshot goes to subscribers

Asking Questions in Plain Language

Databricks includes Genie, a feature that answers data questions typed in everyday words. A business user asks for last week's top products, and Genie writes the SQL and shows the result. Genie works best on well-named tables with clear column descriptions, so documenting your tables improves the answers.

Example: Store Manager Dashboard

A retail company builds a dashboard for store managers. One dataset summarizes daily sales by store. A second dataset lists low-stock products. The store filter controls every chart, so each manager sees only the numbers for the chosen shop.

STORE MANAGER DASHBOARD
──────────────────────────────────────────────────────────
Filter: Store = Pune Central
   │
   ├──► Counter: Today's sales
   ├──► Line: Sales for the last 30 days
   └──► Table: Products with fewer than 10 units left

Design Habits That Help Viewers

• Place the most important number at the top left.
• Limit each page to six or eight visuals.
• Label axes and units clearly.
• Use the same color for the same meaning across the whole dashboard.

Key Points

• A dashboard presents charts, counters, and filters on one page.
• Datasets hold the SQL that feeds each visualization.
• Filters let viewers explore without writing code.
• Gold tables give dashboards fast and trusted data.
• Scheduled refreshes and subscriptions keep viewers current.
• Genie answers questions typed in plain language.