Choosing the right chart for the question

Why this matters

A chart is an argument made with shapes. The right chart makes a pattern obvious in a second; the wrong one hides it, or suggests one that is not there. Most poor charts are not badly drawn. They are the wrong type for the question, chosen because it was the default, or because it looked impressive.

The exam asks you to select appropriate visuals for data and purpose. This lesson covers the core chart types and what each answers, the question families that point to them, more specialised charts, what to do when a chart has too much in it, and when a table is the better answer.

The lesson

Bar, line, pie, scatter and histogram, and what each answers

Five chart types cover most analytical needs.

A bar chart compares values across categories: sales by region, tickets by team. Bars can be vertical (often called a column chart) or horizontal, and horizontal bars suit long category names. Bars should start at zero, because the eye compares their lengths, and sorting them by value, rather than alphabetically, usually makes the comparison clearer.

A line chart shows change over a continuous axis, almost always time: monthly revenue, daily visitors. The line joins the points because the values in between are meaningful, which is why lines should not be used for unordered categories.

A pie chart shows the parts of a whole. It works only when there are few parts, perhaps two to five, that add up to a meaningful total, and when the point is roughly how the whole divides. People judge angles and areas less accurately than lengths, so comparing similar slices is hard, and a bar chart is usually a clearer alternative.

A scatter plot shows the relationship between two numeric variables, one point per record: satisfaction against recommendation, price against quantity. It reveals direction, strength, clusters and outliers.

A histogram shows the distribution of one numeric variable, as bars counting how many values fall in each range. Unlike a bar chart, its bars represent adjacent ranges of a continuous value and touch.

Comparison, trend, composition, distribution and relationship

Choosing a chart is easier when the question is named first.

• Comparison between categories: bar chart.
• Trend over time: line chart, or a column chart for a few periods.
• Composition of a whole: stacked bar, or a pie for very few parts, or a treemap for many.
• Distribution of values: histogram or box plot.
• Relationship between variables: scatter plot, or a bubble chart when a third variable sets the point size.

Many real questions combine two families. "How has each region's sales changed over the year?" is a trend and a comparison, which suggests a line chart with a line per region, or a small separate chart per region. Naming both parts of the question makes the design choice explicit.

Stacked charts, heat maps, box plots and waterfalls

Several more specialised charts answer particular questions well.

A stacked bar or column chart shows totals and their composition together. The bottom segment is easy to compare across bars; the others are not, because they start at different heights. A 100% stacked chart shows proportions only, dropping the totals.

A heat map shows values in a grid, such as day of week against hour of day, with colour representing size. It reveals patterns in two dimensions at once, such as the busy hours of each weekday, and colour is imprecise for reading exact values.

A box plot (box-and-whisker) shows a distribution's five-number summary: the median, the quartiles as the box, the whiskers, and outliers as points. Several box plots side by side compare distributions across groups compactly.

A waterfall chart shows how a starting value becomes an ending value through a series of increases and decreases, such as last year's profit to this year's through each contributing change. It is the visual form of a reconciliation bridge.

Other common types include the area chart, a line chart filled below; the treemap, nested rectangles sized by value; and the gauge and KPI card, single numbers against a target on dashboards.

Too many series, and when to split a chart

A chart has a limited capacity. A line chart with twelve lines in twelve colours, a pie with fifteen slices, or a bar chart with forty categories asks the reader to do work the chart should have done.

Several remedies help:

• Highlight the series that matters in a strong colour and show the rest in grey for context.
• Group small categories into "other", stating what it contains.
• Filter to the top few categories, with the rest available in a table.
• Split the chart into small multiples: a grid of small charts with the same axes and scale, one per category, which lets the eye compare shapes without untangling overlapping lines.

The test is whether a reader can answer the chart's intended question within a few seconds. If they have to consult the legend repeatedly, the chart is too busy.

Tables when exact values matter more than shape

A chart is best at showing shape: trends, comparisons, outliers, patterns. It is poor at communicating exact values, which have to be read off an axis or a label.

A table is better when the reader needs to look up precise figures, compare individual values, or see several measures for each item at once: a price list, a set of regional results with sales, margin and growth side by side, or figures that will be copied into another document.

Good tables follow a few conventions: align numbers to the right so digits line up, use consistent decimal places, put units in headers rather than in every cell, sort rows meaningfully, and use light formatting, such as subtle shading for emphasis, rather than heavy gridlines. A highlight table or conditional formatting adds a little of a chart's quick visual reading to a table's precision.

Choosing between a chart and a table is itself part of the question: "roughly how do the regions compare?" wants a chart; "what exactly did each region sell?" wants a table.