Data visualization design lives in a permanent tug-of-war between two complaints: “this chart is so flat nobody’s eye lands on it” and “it looks great, but I can’t tell what the numbers actually mean.” Good data visualization is the discipline of carrying a number without distorting it while still making someone want to keep looking.
1. Where Accuracy Breaks First
Most distortion in a chart isn’t malicious — it creeps in through choices made purely to look more finished. The rule worth internalizing is that the more visually striking an option looks, the more suspicious you should be that it’s bending the data.
The statistician who put this most clearly is Edward Tufte. His “data-ink ratio” argues that anything on a chart that isn’t representing data — decorative gridlines, ornamental backgrounds, unnecessary drop shadows — should be stripped out. Asking whether every line and every color on screen is actually carrying a value is still the most basic test anyone reviewing a chart can run.
Truncating the y-axis instead of starting at zero, exaggerating the apparent change
Adding 3D effects that make a value look larger than it is
Applying a gradient to unordered categorical data, implying an order that doesn’t exist
2. Handling Color
Color is the most powerful tool in data visualization and the easiest one to misuse. Building a semantic color system with defined brightness steps for general UI work, then extending it specifically for charts, means you’re not re-litigating color choices for every new graph — and the result stays both consistent and accurate.
3. The Chart-Type Mistake People Keep Making
Chart type gets misused almost as often as color does. The classic case is the pie chart: the moment it holds more than three or four slices, comparing wedge sizes by eye becomes genuinely hard, yet it keeps getting reached for out of habit whenever the story is “these are shares of a whole,” regardless of how many categories are involved. Switching to a bar chart alone usually makes the comparison far more accurate. The right question when picking a chart type is never “which one looks best” — it’s “which shape represents this relationship without distorting it.”
Building the habit of pairing data type with chart type up front — line charts for change over time, bar charts for comparing magnitude across categories, scatter plots for the relationship between two variables — avoids the far more common and far more dangerous sequence: picking an attractive shape first and then forcing the data to fit it. Starting from form and retrofitting the data is where distortion almost always begins.
4. Where to Put the Appeal, If Not in the Data
If you want to add appeal without compromising accuracy, the safe place to build it is in what surrounds the data, not in the data itself. A smooth entrance animation, detail that surfaces on hover, disciplined typography and generous whitespace — none of these touch the underlying values, yet all of them make the overall impression considerably more inviting. The same accessibility checklist that covers color contrast and keyboard navigation for any interface applies directly to data visualization too. If a colorblind user can’t tell the legend items apart, the most sophisticated chart in the world has failed for half its audience.
In practice, these two demands often sit with different people on a team. An analyst or engineer verifies that the data is accurate; a designer polishes the animation and tone. Splitting the responsibility this way keeps one person from chasing both goals at once and quietly dropping one of them.
A Data Visualization Checklist
Is any axis truncated in a way that exaggerates or minimizes the value?
Does the color palette match the data’s nature — categorical, sequential, or diverging?
Can a colorblind user still distinguish the legend and categories?
Do the animation and decoration leave the underlying values undistorted?
Does the same data, redrawn as a different chart type, still support the same conclusion?
💡 Pro tip — Build appeal into animation, hover interaction, typography, and whitespace instead of the data itself, and the chart stays trustworthy while still looking good.
Closing thoughts
Data visualization design ultimately comes down to refusing to trade away either trust in the numbers or interest in the screen. Before the next dashboard ships, the habit worth building is checking, before any styling happens, whether the representation distorts the underlying value at all. The examples on the official D3.js site are a solid reference for seeing accuracy and expressiveness handled together.
Design Daily Life · Notes on design, daily