Why this matters
A table of numbers can hide the exact thing you need to see. The right picture makes it obvious. But a chart is not automatically honest or clear, and choosing badly can bury a finding as easily as reveal it.
The idea
Data visualization has two jobs, and they are not the same. It deepens the analyst's own understanding, letting you see distribution, trends and outliers that a table conceals. And it communicates results to other people, so someone seeing the data for the first time can grasp it at a glance.
Bar, line and scatter plots handle one or two variables well. When you need three or more, or many category combinations at once, two more tools:
A heatmap puts items on two axes and uses colour to show a third value for each combination: customers by day of week and time slot, or average temperature by city and month. It also solves a problem scatter plots have. With many points they overlap and the density becomes unreadable, which is overplotting; a heatmap shows density directly. Scatter plots also cannot handle categorical combinations at all.
A mosaic plot shows the relationship between two categorical variables as areas of rectangles. Row and column widths follow the proportions of the frequencies, so each rectangle's area is that combination's share. Where a cross-tabulation is hard to read, differences in proportion become visible as differences in area.
Data storytelling is what turns a chart into a message. Three elements:
| Element | What to do |
|---|---|
| Clarify the message | The one sentence you want understood. Put the claim in the title: "The class scored highly overall on the maths test" |
| Select elements | Remove anything unrelated to the message; highlight what matters in colour and fade the rest |
| Provide supporting information | Title, axis labels, units and notes, properly set out |
The second is the counterintuitive one. Adding more to a chart usually makes the message harder to find, not easier.
Picture it
flowchart TD
Q{What are you showing?} --> A[One or two variables]
Q --> B[Three or more, or many combinations]
A --> A1[Bar, line or scatter plot]
B --> B1["Heatmap: colour shows a third value"]
B --> B2["Mosaic plot: area shows category proportions"]
A1 --> S["Then: clear message, only what serves it, honest labels"]
B1 --> S
B2 --> SWorked example
A shop has two years of visit records and wants to know when it is busiest.
A line graph of daily visitors over 730 days is technically correct and practically useless: the weekly rhythm is buried in the noise, and no amount of staring recovers it. A scatter plot of hour against day is worse, because thousands of points land on the same spots and overplotting hides exactly the density you came to see.
A heatmap of day of week against hour of day, coloured by visitor count, answers the question immediately. Seven rows, twenty-four columns, and the dark band is your answer. The pattern was always in the data; the first two charts just could not show it.
Then storytelling does the rest. Title it with the finding, "Busiest between 5pm and 7pm on Thursdays and Fridays", rather than the neutral "Visitors by day and hour". Fade the quiet cells so the busy block carries the eye. Label the axes and say what the colour means.
Same data, same analysis, completely different chance of anyone acting on it.
Your turn
Try the practice questions for this lesson. Judge statements about visualization and storytelling, match each graph to what it represents, and pick the right chart for a given question.
Take it further
Your class has data on students' home governorate and favourite sport, and monthly rainfall for several Egyptian cities. Should you use a bar graph for everything, or match each dataset to the right graph?
- Choose a graph for each. Name the graph you would use for governorate against favourite sport, and for city against month rainfall. Justify each choice from the number and type of variables.
- Make it honest. For one of them, write the title as a claim rather than a label, and list the supporting information a reader needs to interpret it correctly.
- Decide. Present one chart and say what a reader should conclude, and what they should not.
Two categorical variables and a value across two axes are different problems, and the graph that suits one hides the other.
Recap
- Data visualization serves both the analyst's understanding and the audience's.
- Bar, line and scatter plots suit one or two variables.
- A heatmap shows a third value by colour and avoids overplotting; a mosaic plot shows category proportions by area.
- Data storytelling means one clear message, only the elements that serve it, and honest labels.