Chartjunk and the Data-Ink Ratio
ARTICLEWhy stripping the decoration out of a chart almost always makes it clearer — and the simple idea that tells you what to keep.
"Chartjunk" is any part of a chart that does not help the reader understand the data, and the "data-ink ratio" is the proportion of a chart's ink that actually encodes the numbers rather than decorating them. Both terms come from the statistician Edward Tufte, who introduced them in his 1983 book The Visual Display of Quantitative Information. Together they give you a practical test for almost any design choice: if an element does not change what the reader can see in the data, it is probably chartjunk, and removing it raises the data-ink ratio.
What chartjunk is
Chartjunk is decoration masquerading as design. It is the part of a chart that draws the eye, takes effort to read past, and contributes nothing to the message. The usual offenders are easy to recognise once you start looking:
- 3-D effects — extruded bars, tilted pies, and "depth" that distorts the very lengths and angles the reader is trying to compare.
- Heavy gridlines — dark, thick lines that compete with the data instead of quietly supporting it.
- Background images and fills — photos, gradients, or textures behind the plot that reduce contrast and add visual noise.
- Redundant labels — the same value shown as a bar, a number, and a legend entry all at once.
- Decorative borders, shadows, and bevels — frames around frames, drop shadows on every bar, glossy highlights.
None of these help anyone read a value or spot a trend. They are pure overhead — and worse, some of them (3-D especially) actively mislead by warping the encoding. That overlap with deception is why chartjunk shows up so often in our roundup of chart mistakes to avoid and in the gallery of misleading charts.
The data-ink ratio
Tufte's data-ink ratio puts a number on the same intuition. Imagine adding up all the ink (or, on screen, all the pixels) used to draw a chart, then asking how much of it is data ink — the marks that would change if the underlying numbers changed. The line of a line chart is data ink; the bars of a bar chart are data ink. A filled gray background, a 3-D shadow, and a double border are not — they look identical no matter what the data says.
The data-ink ratio is data ink divided by total ink. The guiding principle is to push that ratio higher: erase non-data ink, and erase redundant data ink, as far as you can without harming readability. You are not trying to make the chart sparse for its own sake — you are removing whatever the reader would not miss.
For any element on a chart, ask: "If I deleted this, would the reader lose information?" If the answer is no, it is chartjunk. This single question resolves most design debates faster than any style guide.
Cluttered vs. clean
The contrast below shows the same five values twice. On the left, the chart is buried under a dark fill, heavy gridlines, a thick frame, and shadowed bars. On the right, the same data is drawn with nothing but the bars and a quiet baseline. The clean version is faster to read and easier to compare — and it shows exactly the same numbers.
How to cut chartjunk
Reducing chartjunk is mostly subtraction, and it follows a reliable order:
- Remove 3-D entirely. There is almost no case where a flat chart is harder to read than its 3-D version, and the 3-D one distorts comparisons.
- Mute or drop gridlines. If you keep them, make them thin and light so the data sits clearly in front.
- Delete backgrounds, frames, and shadows. A chart rarely needs a border or a fill behind the plot area.
- Cut redundancy. Choose one way to show each value. If bars are labelled directly, you may not need a value axis and data labels and a legend.
- Label directly. Where it fits, placing a category or series name next to its line or bar beats a separate legend the eye has to keep cross-referencing.
These habits sit at the centre of clear chart design generally — see our notes on storytelling with data and dashboard design basics, where uncluttered charts matter most because many of them share a screen.
Sensible limits
Maximising the data-ink ratio is a principle, not a law. Some non-data ink earns its place: axis labels, units, a light reference line, or a single annotation that explains an outlier all help the reader even though they are not data. The goal is not the barest possible chart — it is a chart with nothing the reader would miss and nothing that gets in the way. When in doubt, add the element back only if removing it cost the reader something real.
Deleting axis labels, units, or scale information in the name of "less ink" does not raise the data-ink ratio in any useful sense — it removes context the reader needs to interpret the data correctly. Cut decoration, not meaning.
Ready to apply this to your own data? Build a chart with one of the free, no-signup makers at make a chart, start from the right type with how to choose a chart, or browse every option in the chart types guide — then strip out everything the reader would not miss.