Types of Charts, and Which One Answers Each Data Question

Which types of charts compare values across categories?

Bar charts are the types of charts that answer a comparison data question best, because readers judge bar length against a shared zero baseline.

A column chart is the same comparison with vertical bars; turn bars horizontal when labels are long. The zero baseline is not optional: a value axis that starts at 40 makes a 45 look more than twice the size of a 42. Comparison charts have four close relatives:

• Grouped bars for 2–3 series per category, such as this year against last year.
• Dot plot when values sit far from zero, because dots need no baseline.
• Bullet chart for one value against a target, the key performance indicator (KPI) case.
• Radar chart for one profile across 5–8 measures, because overlapping shapes turn hard to read.

As a rule of thumb for category count, past 10–12 bars, show the top 10 plus "Other", or use a table; for planners facing thousands of stock keeping units (SKUs), the planning tools we design for supply chain products can show, for example, an exception list ranking SKUs by forecast error and stockout risk. In the TradeZella journal screens we designed, an example of our financial dashboard design work, a Net Daily P&L (profit and loss) bar chart shows green profit and red loss days, so traders see at a glance whether each day was a winner or a loser. Digno, a performance evaluation SaaS we designed, uses an employees bar chart that compares four employees' scores, with the top performer highlighted at 88%. Sort your bars by value unless the categories carry their own order, such as months.

Which charts show change over time?

For a change-over-time data question, the best chart types are a line chart for a continuous time interval, such as days or weeks, and columns for a few discrete periods.

Time runs on the horizontal axis, left to right. Time series charts plot values recorded at regular points, and the line shows the trend. Five relatives cover other time questions:

• Area chart, a line filled to the axis, which Microsoft describes as drawing attention to the total value across a trend.
• Sparkline, a word-sized line with no axes inside a KPI card or table row.
• Slope chart, two points in time joined by lines, for before and after.
• Calendar heat map, one colored cell per day, for weekly patterns.
• Candlestick chart, open, high, low and close per period, for prices.

Label the end of each line instead of using a legend; direct labeling saves the eye a trip to the key. Past about five lines, split the chart into small multiples, a grid of small charts on one shared axis scale, such as small multiples for comparisons across 6 regions, an example rule from our chart selection matrix.

In the TradeZella screens, a candlestick chart sits among the analytics widgets, and Mentor mode lets a coach follow each mentee's cumulative P&L side by side. Count the lines on your busiest trend chart and label each one at its end.

Which types of charts show parts of a whole?

Stacked bars and pie or donut charts are the types of charts that show parts of a whole in a report, with the pie kept to a few parts.

Part-to-whole charts show proportions, so the parts must add up to 100% of one total. As a rule of thumb, keep a pie chart to about 5 slices, because past that readers compare angles they cannot judge. A stacked bar chart works for 2–4 parts when the part that matters sits on the baseline; NN/g counts stacked bars among the charts with the highest error rates, so past 4 parts, or when middle segments must be compared, use grouped bars or small multiples. A 100% stacked bar compares one split across groups, such as plan mix by region. Three more charts fit here:

• Treemap, nested rectangles sized by value, for many parts.
• Waterfall chart, for how a starting value moves to an ending value through gains and losses, such as revenue down to profit.
• Waffle chart, a 10×10 grid of squares, for a share of a count, such as 3 in 10 users.

Count your slices; past five, rebuild the pie as sorted bars.

What are the different types of pie charts?

The different types of pie charts are the standard pie, the donut chart, the exploded pie, pie of pie or bar of pie, and the 3D pie to avoid.

A donut chart is a pie with a hole whose center holds the total; the Mass Media Co. task dashboard we designed uses a task flow donut. An exploded pie pulls one slice out, and Excel's pie of pie and bar of pie move small slices into a second chart. A 3D pie tilts the circle, so front slices look larger than their values. Draw the same data as a flat pie or a donut instead.

Which charts show distribution and relationships?

Histograms and box plots are the chart types for a distribution data question, and scatter plots show whether one variable moves with another.

Use them when the question is "how spread out are the values?" or "does one number rise when another does?"

Distribution charts

Distribution charts show how values spread, so a product report shows the typical case and the outliers instead of one average.

• Histogram: counts the frequency of values in bins, such as sessions by length in 5-minute bins.
• Box plot: shows the median, quartiles and outliers, and compares the spread of many groups side by side.
• Violin plot: a box plot with the full shape of the distribution mirrored around it.
• Strip or dot plot: every value as a dot, for small samples.

What is the difference between a bar chart and a histogram?

A bar chart compares categories, while a histogram shows the frequency of continuous values grouped into bins.

Category bars have gaps and can sit in any order; in a frequency chart the bars touch, because the bins form one continuous axis, and their order is fixed. Reordering those bars would break the scale, which is the quickest test of which chart you are looking at.

Relationship charts

Relationship charts plot two or more variables against each other, so a product team sees whether usage and retention rise together.

• Scatter plot: data points on two numeric axes, showing correlation and the outliers that break it.
• Bubble chart: a scatter plot whose dot area encodes a third variable, read less accurately than position.
• Heat map: a grid of colored cells on two categorical axes; as a correlation matrix, it shows every variable pair at once.
• Connected scatter: dots joined in time order, for two variables over time.

A scatter plot shows correlation, not cause: two numbers can rise together because a third one drives both. Before you average a metric, plot its distribution; a skewed spread changes which number you report.

Which charts show flow, hierarchy and geography?

For flow, hierarchy and geography data questions, use a Sankey diagram, a treemap or sunburst chart, and a choropleth map, chart types the five basics cannot replace.

These specialist charts are harder to read, so use them when the flow, the hierarchy or the place is the point of the question:

• Sankey diagram: band widths show volumes moving between stages, such as traffic source to signup to plan.
• Funnel chart: stage-to-stage drop-off along one path, such as a checkout funnel.
• Network graph: nodes and edges, for connections between people or systems.
• Treemap and sunburst chart: nested rectangles or rings sized by value, for a hierarchy such as budget by department and team.
• Choropleth map: geographic regions shaded by a value; normalize it per capita, or large states look important only because they are large.
• Symbol map: sized circles at locations, for counts by city or store.

Charts like these are where custom data visualization starts: Sankey, network and small-multiple charts are designed and handed to developers as D3.js or Vega-Lite specs, not picked from a menu. In data storytelling, an annotation points readers to the one flow or region that matters. Two charts share the name but answer no data question: a Gantt chart schedules project tasks, and a control chart tracks a process against limits in quality control. Ask whether sorted bars answer the same question; if they do, use them.

Which types of charts fit product dashboards and apps?

In product dashboards and apps, the types of charts that work best are the simple ones (KPI cards with sparklines, bars, lines, bullet charts, tables), read at a glance.

Specialist data visualization charts belong in reports and drill-down views, where readers have time and a wide screen. Four product constraints change the choice:

• Screen width. A phone is far narrower than a desktop chart area, so the chart changes form rather than shrinking, as the image after this list shows.
• Interaction pattern. A filter, or a drill-down from a summary chart to a detail chart, keeps the first view simple, and a tooltip carries exact values.
• Live data. Fix the axis range or animate rescaling, and show when the data last updated.
• One data point. A chart with a single value is a number, so show it as a KPI card.

Tools name chart types differently: Tableau suggests charts for the selected fields in its Show Me pane, and Power BI lists them in its Visualizations pane. We design charts for Tableau and Power BI as well, with color scales and chart rules applied in each tool's formatting settings.

A gauge chart fits one value against fixed named ranges: the Digno screens we designed show an organization health score of 670.2 on a gauge with score ranges from Poor to Pro, and TradeZella pairs a profit factor gauge with win-rate ring widgets. To compare several values or show progress to a target, use bars or a bullet chart; the Anyday App business dashboard we designed, for example, shows its financial metrics with a transaction details bar chart. When your charts live inside an analytics or admin dashboard, our dashboard design services cover the screen around them. Open your busiest chart on a phone; anything that needs hover to read needs a redesign.

What chart is the most used chart?

No usage census exists, but the bar chart is the best chart for comparisons in the chooser above, and NN/g advises bars, lines or scatter plots in most cases.

The best chart to show anything is the one that answers your data question with position or length: bars for comparisons, lines for trends, scatter plots for relationships. Pick a specialist chart only when none of those three shows the point.

How do you check a chart type before it ships?

Check a chart type before it ships in a product or report with three passes: honest encoding, accessible marks and labels, and a test where users answer 5 questions.

Each pass catches a different failure: a chart that misleads, a chart some users cannot perceive, and a chart that looks right but gets misread.

Pass 1, honest encoding. Encoding accuracy is how precisely people read a value from a visual property: position beats length and angle, and both beat area. That ordering comes from Cleveland and McGill's "Graphical Perception" (1984), which the authors offered as a guideline for graph construction, and Jeffrey Heer and Michael Bostock largely reproduced it with crowdsourced participants in 2010. The table pairs common misleading choices with the replacement that fixes each one.

Pass 2, chart accessibility. Under the WCAG 2.2 Non-text Contrast criterion, graphical objects needed to understand the content, which covers chart marks such as lines, bars and slices, need a contrast ratio of at least 3:1 against adjacent colors. The Use of Color criterion rules out color as the only cue, so label lines directly, add patterns to stacked segments, and pair red and green status with icons. Every chart needs a text alternative: a one-sentence takeaway plus a data table for screen reader users. Match the color scale to the data: categorical for groups, sequential for ordered values, diverging around a midpoint such as zero change.

Pass 3, the 5-question test. In our data visualization process, users from each audience answer 5 questions with the charts, and designers check every chart against WCAG 2.2. For the intro's made-up plan data, ask how many weekly users Team has, whether Business is larger than Free, whether usage grew last quarter, which week looks unusual, and what the one takeaway is. Run the 5-question test on your most-used chart this week; one wrong answer tells you more than a design review.

When is a chart selection matrix worth building?

A chart selection matrix is worth building when several people publish charts, users misread the current types of charts, or a customer dashboard or annual report is planned.

Our data visualization work delivers a chart selection matrix beside a data audit, a visualization style guide, interactive chart prototypes, a chart accessibility spec and a developer handoff; it is a set of rules that match each data question to a chart type. Five signals call for one:

• Teams publish charts in different styles, because each report then teaches readers a new color code.
• Support tickets or user tests show misread charts, because the encoding does not match the data type.
• Customers see charts inside your product, because a branded, interactive chart becomes part of the product experience.
• A contract names WCAG 2.2, because accessibility then needs written rules per chart type.
• An annual report needs more than standard bars, because flows, hierarchies and maps need custom chart design.

The color scale and the accessibility rule for each question live beside the chart selection matrix, in the visualization style guide and the chart accessibility spec. More charts and deeper interactivity raise the price, and how chart count and interactivity set the price gives the ranges.

The question column is the part only your team can fill. A list of 3–10 questions, each marked with the types of charts in this guide that answer it, already works as a brief, and you can send us the questions your users ask of your data as they stand. Every project begins with a free consultation and a fixed-scope proposal.

FAQs