Axis

An axis is a scale made visible. Somewhere behind every chart is a function turning numbers into pixels, and the axis is the part of that function the reader can actually see: a ruled edge, a handful of ticks sampling the scale, and a label on each tick saying what value that position stands for. Without it a bar chart is a row of shapes with a shared bottom, readable as bigger and smaller but not as amounts. The positional accuracy of a chart is inherited entirely from its axes, which is why they are drawn first and thrown away last.

The two names most people reach for describe geography rather than work. An x-axis and a y-axis say where the edge sits on the page; a value axis and a category axis say what it carries, and only the second distinction changes the design. A value axis holds a continuous quantity, so the space between ticks means something and the reader may interpolate across it. A category axis holds names with no distance between them, so its ticks are slots rather than measurements and there is nothing to interpolate. Confusing the two produces the classic mistake of a line drawn through unordered categories, implying a trend where only an alphabet exists.

Most axis craft is subtraction. Fewer ticks than feel safe, spaced at round numbers a reader can do arithmetic with, because a label of 2,150 costs more attention than 2,500 saves. Where the data’s own extent gives ugly numbers, extend the domain to a rounded one and accept the headroom, or switch to a scale whose structure suits the data, such as the log scale that makes three orders of magnitude readable at once. Rotated tick labels are almost always a message about layout rather than typography: if the category names do not fit horizontally, the bars want to be horizontal too. And the choice of domain is a separate decision from the component itself, which is where a truncated axis comes in: the axis is the apparatus, truncation is one claim you can make with it, and a bar chart cut off above zero misreports every length on it.

An axis also earns its space, which is the argument for sometimes going without one. A sparkline drops both axes deliberately: at text size there is no room for a scale, and the chart is trading precise reading for the shape of a trend beside the number it belongs to, so the reader is expected to get the exact figure from the text and only the direction from the picture. Note also that tick marks is a live word in two places at once, on a chart’s axis and on a slider track, where the notches are printed stops rather than a scale being sampled. Carbon’s guidance on axes and labels is the most practical published version of the tick-density rules, down to when to skip labels rather than turn them.