Scale on a Graph

A scale on a graph is the rule that converts data values into positions along an axis. It sets the axis range (minimum to maximum), the interval between labelled divisions, and the type of spacing — linear, logarithmic or categorical — so that every distance on the chart stands for a known amount of data.

Scale is what makes "halfway up the axis" mean a specific number. If the scale runs from 0 to 100 with a division every 20, a point halfway up represents 50. Change the scale and the same data can look flat or dramatic, which is why scale is one of the most consequential — and most easily abused — choices in any chart.

The parts of a scale (and what the divisions are called)

People often ask what the divisions on each axis are called. The short answer: the small marks are tick marks, the numbers beside them are tick labels, and the fixed gap between neighbouring ticks is the interval (also called the increment or step). When ticks are extended across the plot area they become gridlines.

Types of scales in charts

Most chart scales fall into one of four families. Knowing which one you are looking at tells you how to read distances on the axis.

• Linear scale — equal distances represent equal differences. The step from 10 to 20 takes the same space as 80 to 90. This is the default and what readers assume unless told otherwise.
• Logarithmic scale — equal distances represent equal ratios, so 1, 10, 100 and 1,000 are evenly spaced. Useful when data spans several orders of magnitude or when percentage growth matters more than absolute change.
• Categorical (nominal or ordinal) scale — the axis lists separate groups, such as product names or survey answers. Spacing carries no numeric meaning; only order may matter.
• Time scale — a special linear scale for dates. Equal distances should represent equal lengths of time, so a gap in the data (a missing month) should show as a gap, not be squeezed out.

How to choose a scale: a worked example

Say you are plotting monthly sales between 12 and 87 units on a bar chart. A good scale comes from a few quick decisions:

1. Find the data range. Minimum 12, maximum 87.
2. Decide where the axis starts. Bars encode value by length, so a bar chart's value axis should start at zero (the baseline). Line charts and scatter plots can start closer to the data.
3. Round the top up to a "nice" number. 87 rounds up to 100.
4. Pick an interval from the 1–2–5 family (1, 2, 5, 10, 20, 50…). An interval of 20 gives six labelled ticks: 0, 20, 40, 60, 80, 100. An interval of 10 gives eleven, which is busier but still readable.
5. State the units in the axis title — "Units sold".

The result, 0 to 100 in steps of 20, is easy to read at a glance, and every bar's length is an honest proportion of its value.

What are equalized scales in a chart?

"Equalized" or shared scales usually means one of two things. First, when several charts sit side by side — for example small multiples showing each region's sales — they should use the same scale on the value axis. If one panel runs 0–50 and the next 0–500, identical-looking bars hide a tenfold difference. Second, on scatter plots and maps, an equal (1:1) scale means one unit takes the same physical length on both axes, so slopes and shapes are not stretched. Spreadsheet tools often scale each axis independently by default, so set shared or fixed limits deliberately when comparison matters.

How to detect misleading scale manipulation

Because scale controls how big a difference looks, it is the first place to check when a chart seems surprising. Watch for these signals:

• Truncated value axis on a bar chart. The axis starts at, say, 45 instead of 0, so bar lengths no longer match the values.
• Uneven intervals. Tick labels jump 0, 10, 20, 50, 100 while the spacing stays the same.
• Unlabelled log scale. Ticks at 1, 10, 100 with no mention of "log", making explosive growth look gentle.
• Mismatched panels or dual axes. Two series on different scales can be lined up to suggest a relationship that is just an artefact of the chosen ranges.

To measure how much a chart distorts, compare the change it shows with the change in the data — the idea behind Edward Tufte's "lie factor". Example: two bars with values 50 and 55 on an axis starting at 45. Drawn heights are 5 and 10, so the second bar looks 100% bigger, while the real difference is 10%. Distortion ratio = 100% ÷ 10% = 10. A ratio near 1 is honest; far above 1 means the scale exaggerates. More examples are in our guide to misleading charts.