Are vertical bars allowed?

Yes. Vertical bar charts are often called column charts; horizontal bars can be easier when category labels are long.

Must percentages add to 100%?

Only when they represent mutually exclusive parts of the same whole. Independent category rates need not sum to 100%.

Can a bar show a negative value?

Yes, with a clearly visible zero baseline and direction indicating the sign.

AI Glossary / Data science and analytics / Advanced

Bar chart

Definition

A bar chart displays values using rectangular bars whose lengths represent magnitude on a shared numerical scale. Bars commonly compare categories, such as article topics or service channels, and may be horizontal or vertical. The chart makes differences easy to see, provided the values, baseline and category definitions support a fair comparison.

What it means

Each bar answers a question about one category: how many, how much, what proportion or what average? The bars do not have to represent counts, but the measure must be stated. A chart of total visits tells a different story from a chart of visits per published article.
Categories can have a natural order, such as satisfaction levels, or an order chosen to help comparison. The order affects how readers scan the display. Keep meaningful sequences intact and explain any grouping into an “Other” category. Keep important distinctions and missing observations visible.

How it works

Aggregate the relevant data to one value per category, or to clearly identified subgroups. Draw bars of consistent width against a common scale, with lengths measured from zero. Label the categories, units and period. For negative values, bars extend in the opposite direction from the zero baseline.
Horizontal bars often accommodate long category names. Grouped bars allow comparisons between series within categories, while stacked bars show components of a total. These variations answer different questions; too many series impair readability.
Check the data behind the display and provide an accessible text explanation or table. A chart's visual simplicity does not remove the need to disclose denominators, definitions and relevant uncertainty.

Three practical examples

A glossary editor compares the number of completed articles across topics. The chart identifies where more drafting remains, provided the editor also knows how many terms each topic requires.
A service manager compares the proportion of requests resolved during the first contact across channels. The chart states the denominator and period so a channel with few requests is not mistaken for a large body of evidence.
A learning team compares survey response categories from strongly disagree to strongly agree. It preserves that order rather than sorting by bar length, because the sequence carries meaning.

Think of it this way

A bar chart is like placing measuring sticks side by side, all starting at the same mark. Their lengths make comparisons quick. If one stick begins partway up the ruler, its visible length no longer represents the whole quantity. Clear labels help, but they cannot fully repair a visual design that encourages the wrong comparison.

Worked example

Suppose three fictional glossary categories contain 40, 30 and 20 completed articles. A count-based bar chart should show lengths in the ratio 4:3:2. The first category has twice as many completed articles as the third, and there are 90 completed articles in total.
Now suppose the required totals are 100, 50 and 25 articles respectively. Completion rates are 40 / 100 = 40%, 30 / 50 = 60% and 20 / 25 = 80%. A rate-based chart ranks the categories in the opposite order: the smallest completed count has the highest completion percentage.
Both charts can be correct because they answer different questions. Label one “Completed articles” and the other “Percentage of required articles completed,” and retain the required totals. If the count chart started its visible bars at 10, apparent lengths would be 30, 20 and 10, misleadingly making the first appear three times the third rather than twice.

Illustrative case study

In a fictional project, Alder Learning prepares a progress dashboard for its glossary launch. Its first chart shows completed article counts by topic. Managers interpret the longest bar as the topic closest to readiness, overlooking that it also has the largest remaining workload.
The team adds a clearly labeled completion-rate view and provides counts of completed and required articles in a table. It keeps editorial review separate, since a drafted article is not automatically ready to publish. Categories remain consistent across updates so readers can compare progress without relearning the labels.
The revised dashboard supports more specific planning discussions: where output is largest, where completion is furthest along and where review remains. It does not reduce those different questions to one ambiguous bar. This case is fictional.

Common misconceptions and mistakes

A bar chart is not the same as a histogram. Bar charts compare discrete categories or values, while histograms group numerical observations into intervals to show a distribution. The interpretation of width and area therefore differs.
Truncating a magnitude bar's baseline can exaggerate differences. Three-dimensional effects can also make lengths harder to judge. Color should explain a grouping or highlight, not create distinctions that have no meaning.
Comparing raw counts across groups of very different sizes may answer the wrong question. Likewise, replacing missing values with zero falsely asserts an observed absence. State whether a bar represents a count, rate, total or average before interpreting its height.

Business use

Use bar charts for clear comparisons across a manageable set of categories. Choose the measure based on the decision: staffing may require workload counts, while service comparisons may require rates with denominators.
Review labels, category coverage and the baseline before publication. For estimates, explain relevant uncertainty instead of presenting small visual differences as definitive rankings. Provide the underlying values when precision matters, and use consistent scales across comparable panels. If the main question concerns change over time, consider whether a line chart would communicate the pattern more directly.

Check your understanding

Why did the rankings reverse in the worked example?

Answer: Counts and completion rates use different denominators and answer different questions.

Why include zero for magnitude bars?

Answer: Their lengths encode the full values, so a truncated baseline distorts that visual comparison.

What does a missing bar value mean?

Answer: It needs explanation; missing data should not automatically be treated as zero.

Related terms

Categorical data
Histogram
Baseline
Data visualization
frequency
column chart
grouped bar chart
stacked bar chart
axis
denominator

Learn next

Create count and percentage summaries from the same small table and write the question each answers. Then check whether category order, labels and scales make those questions clear without extra explanation.

Sources

Australian Bureau of Statistics, Statistical terms and concepts glossary, Bar chart: definition and category context. https://www.abs.gov.au/statistics/understanding-statistics/statistical-terms-and-concepts/statistical-terms-and-concepts-glossary Office for National Statistics, Axes and gridlines: zero baselines and consistent scales. https://service-manual.ons.gov.uk/data-visualisation/guidance/axes-and-gridlines The counts, percentages and Alder Learning case are original illustrations and do not report the glossary project's actual category totals.

Previous: Bang-bang control Next: Barber-Agakov lower bound