Free graph maker & data converters

Which Chart Should I Use?

By Plotiq Editorial Team · Published 14 November 2026 · Updated 14 November 2026 · 5 min read

The right chart depends on the question your data answers, not on which one looks best. This guide gives a short decision process, explains what each of the six common chart types is good at, and shows the mistakes that make figures misleading.

Start with the question, not the chart

Before you open any tool, finish this sentence: “I want the reader to see how ___ compares, changes or relates.” The verb tells you most of what you need. Changes over an ordered variable such as time points to a line chart. Relates one measured variable to another points to a scatter plot. Compares separate categories points to a bar chart. Distributed describes how the values of one variable are spread, which is a histogram or box plot. Share of a whole is the only job a pie chart does well.

It also helps to ask what type of variable sits on each axis. A category (treatment A, treatment B) cannot be joined by a line because there is nothing between the categories. A continuous number (time, concentration, temperature) can be.

Line chart: change along an ordered axis

Use a line when the X values are ordered and the measurements belong to one continuous process, such as temperature while a sample cools or monthly sales across a year. The line asserts that something happens between the points. If that is not true, for example when the X axis lists five different patients, use markers only or a bar chart.

The figure below shows two samples cooling from the same starting temperature. A line chart lets the eye follow each curve and compare how quickly they approach room temperature, which a table of numbers cannot do.

0510152030405060708090100Time (min)Temperature (°C)Sample A (°C)Sample B (°C)
Two cooling curves: change over time is the natural job for a line. Made with the Scientific Graph Maker.

Scatter plot: the relationship between two variables

A scatter plot shows each observation as a point with an X and a Y value. Use it when both variables are measured on a scale and you want to know whether, and how strongly, they are related: height against weight, dose against response, concentration against absorbance. A fitted line with its equation and R² then summarises the trend (see line of best fit and R²).

The difference from a line chart is intent. In a scatter plot the points are independent observations and the order in which you collected them does not matter. Joining them with a line would imply a sequence that is not there.

Bar chart: comparing categories

Bars compare a value across separate groups: yearly totals, mean enzyme activity per treatment, votes per candidate. Bar length is read as a quantity, so the axis must start at zero; a truncated axis exaggerates small differences. Where the bars are means, add error bars so the reader can see whether the differences are larger than the spread (see SD versus SE).

Keep the category order meaningful. Sort by size when there is no natural order, and keep time-based categories in chronological order. Avoid three-dimensional bars, which distort the lengths.

Histogram and box plot: how values are distributed

When you have many raw measurements of one variable, the question is usually about the shape: where is the centre, how wide is the spread, is it skewed, are there outliers? A histogram bins the values and shows the shape directly, and works best with a few dozen values or more. A box plot summarises the same distribution in five numbers and is better for comparing several groups side by side, or when you only have 10–30 values per group.

A histogram is not a bar chart. Its bars touch because the X axis is a continuous number scale, and the bar height is a count or frequency. The guides on histogram bins and box plots go deeper into both.

Pie chart: use rarely, and only for parts of a whole

People compare angles and areas poorly, so a pie chart is hard to read once there are more than about five slices or when two slices are close in size. It is acceptable when the numbers add up to a meaningful 100% and you want to show one dominant share, for example “64% of responses were agree or strongly agree”.

If you need to compare the sizes accurately, a sorted bar chart almost always communicates better. Never use a pie chart for values that are not parts of one total, such as averages for different groups.

A quick decision table

Match your data and question to the chart in this table:

Your dataQuestionChart
Values over time or another ordered scaleHow does it change?Line
Two measured variablesAre they related?Scatter (plus trendline)
One value per categoryWhich is larger?Bar
Many raw values, one variableWhat is the shape?Histogram
Several groups of raw valuesHow do the groups differ?Box plot
Parts that add to a totalWhat share is each?Pie (few slices) or sorted bar

Common mistakes

Connecting unrelated points with a line; truncating the Y axis on a bar chart; showing means without any measure of spread; using colour as the only way to tell series apart (about one reader in twelve men has some colour vision deficiency, and printouts are often black and white); and leaving out axis titles or units. A good test is whether someone who has not read your text can tell what is plotted and in what units from the figure alone.

Frequently asked questions

Can I use a line chart for categories?
Only if the categories have a real order and you accept that the line implies a trend between them, for example months of the year. For unordered categories use bars.
What is the difference between a bar chart and a histogram?
A bar chart compares separate categories and its bars have gaps. A histogram shows the distribution of one continuous variable, so the bars touch and the X axis is a number line.
When is a box plot better than a histogram?
When you are comparing several groups or have a small sample. A box plot compresses each group into a compact summary, so five groups fit side by side, whereas five histograms become cluttered.
Should I use 3D charts?
No. Perspective makes lengths and areas harder to compare and adds no information.
How many series can one chart hold?
Around four to six before it becomes hard to read. Beyond that, split the data into several panels or highlight the series that matter.

Keep reading

Written and reviewed by the Plotiq Editorial Team. Spotted a mistake? Tell us and we will correct it. See About Plotiq for how these guides are made.