Skip to content

Data Analyst Associate · Domain 6: Working with Dashboards and Visualizations in Databricks · 16% of the exam

Task 6.1: Build dashboards using AI/BI Dashboards, including multi-tabs/page layouts, multiple data sources/datasets, and widgets (visualizations, text, images); Create visualizations in notebooks and the SQL editor; Identify the effective visualization type to communicate insights clearly

AI/BI dashboard pages, datasets and widgets, notebook and SQL editor charts, and the visualization types the documentation describes for each kind of insight.

Study it

  • AI/BI dashboards: pages, datasets and widgets

    Lesson coming

  • Visualizations in notebooks and the SQL editor, and choosing a visualization type

    Lesson coming

Sample questions

Try each one before opening the answer. Every option is explained, with the Databricks documentation page that proves it.

Question 1 · choose 1

A dashboard author wants a table widget on an AI/BI dashboard that lists the column names and data types of main.sales.orders, so business users can see what the table holds. Which dataset definition works?

  1. ADESCRIBE TABLE main.sales.orders
  2. BSHOW COLUMNS IN main.sales.orders
  3. CCREATE TEMPORARY VIEW cols AS SELECT * FROM main.sales.orders, followed by a SELECT of the column names from cols
  4. DSELECT column_name, data_type FROM main.information_schema.columns, filtered to the table
Show the answer and why
  • ADESCRIBE TABLE main.sales.orders

    Incorrect

    Dashboard dataset queries are read-only, and commands such as DESCRIBE are not supported for defining datasets.

  • BSHOW COLUMNS IN main.sales.orders

    Incorrect

    Only queries that can be wrapped in additional SQL are valid as dataset definitions, and a SHOW command cannot be wrapped that way.

  • CCREATE TEMPORARY VIEW cols AS SELECT * FROM main.sales.orders, followed by a SELECT of the column names from cols

    Incorrect

    CREATE statements are not supported for defining datasets.

  • DSELECT column_name, data_type FROM main.information_schema.columns, filtered to the table

    Correct

    Every catalog has a read-only INFORMATION_SCHEMA of metadata views, and a SELECT on it is a read-only query that a dataset can use.

AI/BI dashboard datasets accept only read-only queries that can be wrapped in additional SQL; DESCRIBE, EXPLAIN, CREATE TABLE, INSERT and DELETE are not supported. Metadata such as column names and types can still be shown by selecting from the catalog's INFORMATION_SCHEMA, which holds read-only metadata views.

Question 2 · choose 1

In a notebook, an analyst runs a SQL cell that returns about 400,000 order rows (region, amount) and adds a bar chart of the sum of amount by region. To look at one region, she applies the filter region = 'APAC' on the results table. The chart now shows only APAC, but its APAC total is far below the one that a separate GROUP BY query returns. What should she do?

  1. AClear Limit 1000 in the Run (1000) menu, then run the cell again
  2. BChange the visualization from a bar chart to a pivot table of amount by region
  3. CClick Truncated data on the results table and select Full dataset
  4. DAdd LIMIT 400000 to the cell's query so that the results table holds every row
Show the answer and why
  • AClear Limit 1000 in the Run (1000) menu, then run the cell again

    Incorrect

    Run (1000) and Limit 1000 belong to the SQL editor. In a notebook the returned data is truncated above 10,000 rows or 2 MB, and the filter then sees only the returned rows.

  • BChange the visualization from a bar chart to a pivot table of amount by region

    Incorrect

    Filters on the results table also apply to the cell's visualizations, and on truncated data they act on the returned rows whatever the chart type.

  • CClick Truncated data on the results table and select Full dataset

    Correct

    On truncated data the filter applies only to the returned rows by default; Full dataset applies it to the whole result, and the chart follows the filter.

  • DAdd LIMIT 400000 to the cell's query so that the results table holds every row

    Incorrect

    A result of 400,000 rows is still above the 10,000 rows at which the returned data is truncated, so the filter would again see only part of it.

In a notebook, filters on a results table also affect its visualizations. By default they apply only to the results shown: when the data returned is truncated, for example above 10,000 rows or 2 MB, the filter covers only the returned rows and a note says so. Clicking Truncated data and selecting Full dataset filters the full result instead, which can take longer on large data.

Question 3 · choose 1

A delivery dataset has one row per van with latitude and longitude of its last stop, and a second dataset has total deliveries per US state with the state name. On an AI/BI dashboard, the team wants a marker for each van's position and the states shaded by delivery volume. Which visualizations fit?

  1. AA point map from the latitude and longitude, and a choropleth map from the state names
  2. BA choropleth map from the latitude and longitude, and a point map from the state names
  3. CA path map from the latitude and longitude, and a choropleth map from the state names
  4. DA heatmap of latitude against longitude, and a counter of deliveries per state
Show the answer and why
  • AA point map from the latitude and longitude, and a choropleth map from the state names

    Correct

    A point map places markers at latitude and longitude pairs, and a choropleth colors localities returned by name according to aggregate values.

  • BA choropleth map from the latitude and longitude, and a point map from the state names

    Incorrect

    A choropleth needs geographic locations by name (or a geometry column), and a point map needs coordinates, so the inputs are swapped.

  • CA path map from the latitude and longitude, and a choropleth map from the state names

    Incorrect

    A path map draws lines from a line column or a sequence of points; it does not place one marker per van.

  • DA heatmap of latitude against longitude, and a counter of deliveries per state

    Incorrect

    Neither a heatmap of coordinates nor this other widget places markers on a map or shades states by volume.

AI/BI map visualizations need matching geographic data: a choropleth colors countries or states returned by name (or as GEOMETRY or GEOGRAPHY) by an aggregate value; a point map places a marker at latitude and longitude pairs; a path map draws lines such as routes.

Question 4 · choose 1

An analyst plans an AI/BI dashboard with one page per region. What is the maximum number of pages a dashboard can have?

  1. A15
  2. B5
  3. C100
  4. DNo limit
Show the answer and why
  • A15

    Correct

    A dashboard canvas can contain up to fifteen pages.

  • B5

    Incorrect

    The limit is higher: up to 15 pages per dashboard.

  • C100

    Incorrect

    100 is the limit on datasets per dashboard, not pages.

  • DNo limit

    Incorrect

    Dashboards have a documented page limit of 15.

AI/BI dashboards have structure limits: up to 15 pages per dashboard and up to 100 datasets per dashboard. New dashboards start with a single page.

Question 5 · choose 1

A dashboard author wants a block of formatted text with a heading, a link and a small logo on an AI/BI dashboard page. Which widget fits, and which syntax can she use to edit it?

  1. AA table visualization, with HTML in its column headers
  2. BA text widget, which supports Markdown and can include images
  3. CA filter widget, with the text as its title
  4. DA counter visualization with a format template
Show the answer and why
  • AA table visualization, with HTML in its column headers

    Incorrect

    Table visualizations display data; formatted text goes in a text widget.

  • BA text widget, which supports Markdown and can include images

    Correct

    Text widgets add formatted text, links and images, and can be edited in Markdown syntax.

  • CA filter widget, with the text as its title

    Incorrect

    Filter widgets filter data; they are not for formatted text.

  • DA counter visualization with a format template

    Incorrect

    A counter is a data visualization; its format template only adds prefix or suffix text around the value, not formatted text with links and images.

Text widgets let you add formatted text, links and images to dashboards, and you can show and edit their content as Markdown. Image widgets show a standalone image such as a logo.

Question 6 · choose 1

An analyst has query results in the SQL editor and wants a bar chart of them. How does she start one?

  1. AClick + above the results and select Visualization
  2. BExport the results to CSV and open them in a spreadsheet
  3. CCreate a materialized view first, then chart the view
  4. DAsk a workspace admin to enable charts for the warehouse
Show the answer and why
  • AClick + above the results and select Visualization

    Correct

    To create a visualization, click + above a result and select Visualization to open the visualization editor.

  • BExport the results to CSV and open them in a spreadsheet

    Incorrect

    Visualizations are built in place from the results.

  • CCreate a materialized view first, then chart the view

    Incorrect

    Charts can be built from query results directly.

  • DAsk a workspace admin to enable charts for the warehouse

    Incorrect

    Visualizations are built into notebooks and the SQL editor.

In notebooks and the SQL editor, click + above a result and select Visualization to open the visualization editor, configure it, and click Save.

Question 7 · choose 1

A team builds charts in notebooks and wants to share dashboards across the organization. Which kind of dashboard does Databricks recommend?

  1. AA notebook dashboard
  2. BA legacy Databricks SQL dashboard
  3. CAn AI/BI dashboard
  4. DA cohort visualization
Show the answer and why
  • AA notebook dashboard

    Incorrect

    Notebook dashboards present notebook cell outputs; Databricks recommends AI/BI dashboards for sharing across an organization.

  • BA legacy Databricks SQL dashboard

    Incorrect

    The type Databricks recommends for creating and sharing dashboards across an organization is the AI/BI dashboard, not a legacy one.

  • CAn AI/BI dashboard

    Correct

    Databricks recommends AI/BI dashboards for creating and sharing dashboards across an organization.

  • DA cohort visualization

    Incorrect

    A cohort is a visualization type, not a kind of dashboard.

Notebook dashboards reformat a notebook into a presentation tied to cell outputs; Databricks recommends AI/BI dashboards for creating and sharing dashboards across an organization.

Question 8 · choose 1

A sales dashboard needs one prominent number, this month's revenue, shown next to last month's revenue for comparison. Which AI/BI visualization type is built for this?

  1. ABox chart
  2. BCounter
  3. CCohort chart
  4. DChoropleth map
Show the answer and why
  • ABox chart

    Incorrect

    A box chart shows the distribution of values, optionally by category.

  • BCounter

    Correct

    Counters display a single value prominently, with an option to compare it against another value.

  • CCohort chart

    Incorrect

    Cohort charts track groups over time periods.

  • DChoropleth map

    Incorrect

    A choropleth is a map, not a single headline number.

In AI/BI dashboards, counters display a single value prominently with an option to compare it against a comparison value; the notebook counter likewise shows one value against a target.

Practice domain 6 →Practice all domains →