What you will be able to do
- Match a time-based trend to a line, area, bar or combo visualization
- Tell when a pie chart fits and when it does not
- Choose a counter for a single headline value and a scatter or bubble chart for relationships between numbers
Key concept
The insight decides the chart — Each Databricks visualization type is built for one kind of question: change over time, share of a whole, a single headline number, or the relationship between variables. To pick the effective type, first name the question the viewer is asking, then choose the chart made for it.
1.Start from the question, not the chart
AI/BI dashboards come with a large set of visualization types. The dashboard concepts page groups them into chart visualizations (bar, line, area, scatter, pie, combo, funnel and more), table visualizations, statistical visualizations, map visualizations, and specialized ones such as gauges and single-value indicators. Exam questions on this objective almost never ask how to configure a chart. They describe what a stakeholder wants to see and ask which type shows it most clearly.
So the useful skill is translating. "How has revenue moved month by month?" is a question about change over time. "What share of orders is high priority?" is about proportion. "What is today's total?" asks for one number. "Do longer trips cost more?" asks how two numeric variables relate. The Databricks visualization-types documentation gives a short purpose statement for each chart type, and those statements are what this page is built on.
Checkpoint 1 of 6· Check yourself
A stakeholder asks, "Do longer trips cost more?" Which visualization type matches that question?
The question asks how two numeric variables relate, which is what a scatter visualization is commonly used to show.
“Scatter visualizations are commonly used to show the relationship between two numerical variables.”Source: docs.databricks.com
Sources1
2.Change over time: line, area, bar and combo
Four chart types are described in terms of time. The line visualization is the most direct: line visualizations "present the change in one or more metrics over time." In the documentation's example, the x-axis is the order date with a yearly transform, the y-axis is the average total price, and order priority is mapped to color so that each priority gets its own line.
Area visualizations "combine the line and bar visualizations to show how one or more groups' numeric values change" across a second variable, usually time. You can stack them, or use 100% Stack, so they also show how the groups add up to a total. The documentation notes that they are often used for sales funnel changes over time.
Bar charts are the most flexible. They cover change over time, comparison across categories, and proportionality. Their layout can be set to Stack, 100% Stack or Group. Combo charts put the two shapes together: one series appears as bars and another as a line, and each can have its own y-axis. The notebook and SQL editor reference describes them as presenting "the changes over time with proportionality." The AI/BI combo example plots average trip distance as bars on the left axis and average fare as a line on the right axis, with both bucketed weekly.
| Type | Described purpose | Layout options named in the docs |
|---|---|---|
| Line | Change in one or more metrics over time | Line shape: Linear, Smooth, Step |
| Area | Groups' numeric values changing over time; sales funnel changes | Stack or 100% Stack; Line shape |
| Bar | Change over time or across categories; proportionality | Stack, 100% Stack, or Group |
| Combo | Bars plus a line for two measures on left and right y-axes | Left Y-axis (Bar), Right Y-axis (Line) |
Checkpoint 2 of 6· Check yourself
An analyst needs one chart that shows weekly average trip distance and weekly average fare together, using a separate y-axis for each measure. Which type fits?
Combo charts draw one series as bars and another as a line, each on its own y-axis, so two measures can share the same time axis.
“Combo charts combine line and bar charts to present the changes over time with proportionality.”Source: docs.databricks.com
3.Share of a whole: pie or bar
When the question is about proportion (what fraction of revenue each priority contributes), a pie works. Its angle encodes a summed measure, such as total price, and its color encodes the category, such as order priority. The documentation also sets a clear limit: pie visualizations "are not meant for conveying time series data." If the stakeholder wants proportions over time, use a bar chart with a 100% Stack layout. Bars show proportionality "similar to a pie visualization," and they still have a time axis.
Checkpoint 3 of 6· Exam question
A data analyst builds an AI/BI Dashboard widget that plots a single metric, daily gross merchandise revenue, across every day of the past 12 months, and stakeholders want to spot seasonal upward and downward trends at a glance. Which visualization type should the analyst choose for this widget?
Correct answer: A — Configure a line chart, since it presents the change in a single metric over a continuous time axis and makes seasonal trend direction easy to read.
- A. A line visualization is built to present how one or more metrics change over a continuous axis, which is exactly what a twelve-month daily revenue series needs. It makes rising and falling seasonal trends visually obvious because consecutive days are connected into a single traceable path.
- B. A pie chart shows proportionality between a small set of categories, not change over a continuous axis, so it is explicitly not meant for time series data. Turning 365 days into pie slices would produce an unreadable widget rather than a trend view.
- C. A funnel chart is designed to analyze how a metric changes across a small number of defined sequential stages, such as steps in a signup process, not a daily time series. There are no discrete stages here, only a continuous run of calendar days.
- D. A choropleth map colors geographic localities by an aggregated value and is meant for regional comparisons, not for tracking a single metric across time. Recasting the request as a country-by-country comparison would drop the day-by-day trend the stakeholders asked for.
Checkpoint 4 of 6· Check yourself
A stakeholder asks for each order priority's share of total price for every month of the year. Which choice follows the Databricks guidance?
Pies are not meant for time series. A bar chart can show proportionality, and 100% Stack keeps a time axis.
“In the Layout section, choose Stack or 100% Stack, or Group.”Source: docs.databricks.com
Sources2
4.One headline number, or how two numbers relate
Sometimes the insight is a single figure, such as today's revenue. For that, use a counter. Counters "display a single value prominently, with an option to compare them against an offset value." You fill in a Value column and a Comparison column. You can also pick a date column and an aggregation to draw a sparkline in the background. Conditional formatting on the value can color it based on the data. A counter is a deliberate choice: it gives up detail so the one number stands out.
When the question is whether two numeric measures move together, use a scatter. Scatter visualizations "are commonly used to show the relationship between two numerical variables," and you can map a third dimension to color to compare groups. The notebook EDA tutorial adds that scatter plots can highlight outliers. For a fourth variable, use a bubble chart. Bubble is not a separate type in the picker: you select Scatter and set the Size encoding. The documentation's bubble example plots trip distance against fare, colored by pickup zip and sized by minutes in the taxi. It computes that last value in the dataset query:
SELECT
*,
TIMESTAMPDIFF(MINUTE, tpep_pickup_datetime, tpep_dropoff_datetime) AS minutes_in_taxi
FROM samples.nyctaxi.trips
LIMIT 500;Checkpoint 5 of 6· Exam question
A dashboard widget needs to show what percentage of total quarterly subscription revenue came from each of four pricing tiers (Basic, Standard, Premium, Enterprise), with the emphasis on the part-to-whole split rather than any change over time. Which visualization type best fits this widget?
Correct answer: A — Configure a pie chart, since it shows proportionality between a small number of categories and clearly conveys how each tier's slice compares to the whole.
- A. A pie visualization exists specifically to show proportionality between metrics, and four pricing tiers is a small enough set of slices for viewers to compare shares of the whole at a glance. This matches the stated goal of a part-to-whole split rather than a trend.
- B. A line chart presents change over a continuous axis such as time, but the request explicitly says the emphasis is the split among tiers, not a trend. Forcing four static quarterly shares into a line would misrepresent them as evolving values.
- C. A scatter visualization is built to reveal the relationship between two numerical variables, such as spend versus revenue, not to show how one total divides among four named categories. There is no second continuous variable being compared here.
- D. A box chart summarizes the distribution, quartiles, and outliers of numerical data within categories, which answers a spread question, not a part-to-whole share question. It would not communicate what percentage of the total each tier represents.
Checkpoint 6 of 6· Check yourself
In an AI/BI dashboard, how do you create a bubble chart where marker size shows trip duration?
In AI/BI dashboards, a bubble chart is a scatter with a Size encoding. It is not a separate type.
“To make a bubble chart, select Scatter as your visualization type.”Source: docs.databricks.com
Exam traps
Each one states something that sounds right. Open it to see what is actually true.
1.A pie chart is a good way to show how category shares change month by month.Why is that wrong?
Pies show proportionality but are explicitly not meant for time series. Use a bar chart, for example with 100% Stack, so the time axis stays.
Covered in Share of a whole: pie or bar
2.AI/BI dashboards have a separate Bubble visualization type.Why is that wrong?
A bubble chart is a scatter whose marker size reflects a metric. You select Scatter and set Size.
3.To show one KPI against last period's value, you need a table or a two-bar chart.Why is that wrong?
A counter shows a single value prominently and can compare it with an offset value, optionally with a sparkline.
Sources
Every claim above is drawn from one of these pages, quoted as it was written on the date shown.
- 1.
“AI/BI dashboards provide a rich library of visualization types to display your data effectively.”
↩︎ Start from the question, not the chart - 2.
“Line visualizations present the change in one or more metrics over time.”
↩︎ Change over time: line, area, bar and combo“combine the line and bar visualizations to show how one or more groups' numeric values change”
↩︎ Change over time: line, area, bar and combo“They are often used to show sales funnel changes through time.”
↩︎ Change over time: line, area, bar and combo“Pie visualizations show proportionality between metrics. They are not meant for conveying time series data.”
↩︎ Share of a whole: pie or bar“Counters display a single value prominently, with an option to compare them against an offset value.”
↩︎ One headline number, or how two numbers relate“Scatter visualizations are commonly used to show the relationship between two numerical variables.”
↩︎ One headline number, or how two numbers relate“Bar charts represent the change in metrics over time or across categories and show proportionality, similar to a pie visualization.”
↩︎ Key concept“They are not meant for conveying time series data.”
↩︎ Exam trap 1“Bubble charts are scatter charts where the size of each point marker reflects a relevant metric.”
↩︎ Exam trap 2“Counters display a single value prominently, with an option to compare them against an offset value.”
↩︎ Exam trap 3“Scatter visualizations are commonly used to show the relationship between two numerical variables.”
↩︎ Checkpoint“Pie visualizations show proportionality between metrics. They are not meant for conveying time series data.”
↩︎ Prediction“In the Layout section, choose Stack or 100% Stack, or Group.”
↩︎ Checkpoint“To make a bubble chart, select Scatter as your visualization type.”
↩︎ Checkpoint - 3.
“Combo charts combine line and bar charts to present the changes over time with proportionality.”
↩︎ Change over time: line, area, bar and combo - 4.
“For instance, scatter plots can highlight outliers, while time series plots can reveal trends and seasonality.”
↩︎ One headline number, or how two numbers relate