By Michela Tjan
To learn more about the Open Visualization Academy (OVA), visit https://openvisualizationacademy.org
You can change a chart without changing a single number.
Move where the axis starts. Stretch the chart a little taller. Compress it a little wider.
Suddenly a small difference looks dramatic. A sharp trend looks almost flat. The data is exactly the same, but the story feels different.
We often think of axes as technical scaffolding, something every chart needs but nobody really notices. In reality, they're one of the most powerful storytelling tools in visualization. Where an axis begins, how much of the range it shows, and how the chart is proportioned all shape how readers interpret what they're seeing.
The data provides the values. The axis provides the context.
Two charts cross your feed in the same afternoon. Same metric, same company, same quarter. One looks like a gentle increase. The other looks like a dramatic surge. Nothing in the data changed between them—someone just moved where the axis starts.
We often treat axes as a technical necessity, something that's there because charts need one. But a surprising amount of the story lives in the axis itself. Where it starts, how far it extends, whether the chart is stretched vertically or compressed horizontally—these aren't neutral defaults. They're design decisions, and they shape how readers interpret the data long before they read a single number.
Zero Is the Famous One
The baseline is probably the most discussed axis decision. Most people have heard some version of the rule: bar charts should start at zero.
For bars, that's generally true. But understanding why matters more than memorizing the rule.
A bar communicates value through its length. The distance from the baseline to the top of the bar is the quantity being encoded. If you remove part of that length by truncating the axis, you change the visual relationship between values without changing the numbers themselves.
A small difference can suddenly look enormous.
Same five numbers. One axis tells the truth about the size of the change; the other manufactures one.
But the Rule Has Limits
The problem is that good advice often turns into blanket advice.
"Always start at zero" gets applied to line charts too, even though lines work differently.
A line doesn't encode value through length. It encodes change through position and slope. Often, the entire purpose of a line chart is to show relatively small movements that matter: a rise in temperature, a change in unemployment, a shift in exchange rates, or fluctuations in website traffic.
Force every line chart to include zero and those meaningful variations can disappear into a nearly flat line.
The point isn't that the rule is wrong. It's that the rule depends on the visual encoding.
Truncating a bar chart changes the length that represents the value. Truncating a line chart often just changes the zoom level.
Shape Matters Too
Baselines get most of the attention, but there's another axis decision that can be just as influential: proportion.
Take a time series and make the chart tall and narrow. Suddenly every fluctuation feels dramatic. Stretch the same chart wide and short, and those identical fluctuations start to feel much less important.
Nothing changed except the dimensions of the plotting area.
Readers rarely notice this consciously, but they feel it. The steepness of a slope influences how significant a trend appears, even when the underlying numbers remain identical.
Identical numbers. Only the rectangle changed.
So What Should You Do?
The reality is that there's no neutral axis waiting underneath the chart. Even the software's default settings are simply design choices made by someone else.
The goal isn't to avoid making choices. It's to make them deliberately.
If value is represented by length—bars, columns, areas—starting at zero is usually the most honest choice.
If value is represented by position—lines, scatterplots, dots—choose a range that reveals meaningful variation, and label it clearly.
Pay attention to aspect ratio. A chart shouldn't be stretched to maximize drama or compressed to minimize it.
None of this is about following rules for the sake of rules.
It's about recognizing that every axis communicates something about the importance of the change being shown.
The axis is never just a frame around the data. It's part of the message.
The only question is whether you intentionally shaped that message, or let the software do it for you.
