Dual-axis charts: why they mislead and what to use instead
Blog
Dual-axis charts: why they mislead and what to use instead
Two y-axes let the chart maker decide where lines cross and how steep they look. Four alternatives that show the same data honestly, with when to use each.
By DataVisuals Team4 min read
Short answer
Dual-axis charts mislead because each axis can be scaled independently, so the crossing points, gaps and apparent correlation between the two lines are choices of the chart maker rather than facts in the data; indexed lines, side-by-side panels or a connected scatter plot usually show the same relationship honestly.
Key takeaways
With two y-axes, moving either scale changes where the lines cross and how related they look.
Readers tend to read meaning into lines that sit close together, even when the closeness is arbitrary.
Indexing both series to 100 at a start date puts them on one honest axis.
Side-by-side panels with a shared x-axis keep each measure in its own units.
If you must use two axes, align both at zero and say so on the chart.
A dual-axis chart puts two measures with different units on one chart: one y-axis on the left, another on the right. It's popular because it saves space and makes two trends look connected. That second part is the problem. On a chart with two independent scales, how connected the lines look is up to whoever set the axes.
This post explains what goes wrong and shows four ways to chart two measures that don't depend on that choice.
Why do dual-axis charts mislead?
Because each axis can be scaled freely, the chart maker controls where the lines cross, how far apart they sit and whether they appear to move together.
Take monthly revenue (in thousands of dollars, left axis) and ad spend (in thousands, right axis). Set the right axis from 0 to 50 and spend looks modest, tracking just under revenue. Set it from 20 to 40 and spend shoots past revenue in month four. Same data, opposite stories. Nothing on the chart tells the reader which version is "right", because there isn't one.
Fig. 1 Revenue and ad spend for six months. On two axes, the crossing point depends only on how each axis was set. Indexed to 100 on one axis, the real story shows: spend grew much faster than revenue.
Lisa Charlotte Muth's widely shared Datawrapper article on the topic makes two further points. Even with both axes starting at zero, readers tend to see a relationship in lines that happen to sit near each other. And user studies of superimposed charts with different scales found them slower and less accurate to read than the alternatives.
What should you use instead?
Four common options: an indexed chart, side-by-side panels, one series with an annotation, or a connected scatter plot.
Alternative
Shows best
Watch out for
Indexed to 100
Which series grew faster
Hides absolute values; pick a fair start date
Side-by-side panels
Each series in its own units
Takes more space; align the x-axes
One series plus labels
The main measure, with key values of the other
Only works if one series matters most
Connected scatter plot
How the two measures moved together over time
Unfamiliar to many readers; label the start and end
How does an indexed chart work?
Every value is divided by the value on a start date and multiplied by 100, so both series begin at 100 and share one axis.
If revenue went from 200 to 240, its index goes from 100 to 120. If ad spend went from 20 to 34, its index goes from 100 to 170. Now one axis is enough, and the gap between the lines means something: spend grew 70 percent while revenue grew 20 percent.
The start date matters a lot. Starting in an unusually bad month makes everything after look like growth. Pick a typical period, or show a couple of start dates to check that the story holds. Label the axis clearly, for example "Index, January = 100".
When are side-by-side panels better?
When readers need the actual values in the original units.
Two small charts stacked vertically, sharing the same x-axis and time span, let each measure keep its units. The reader compares the timing of changes by looking straight down from one panel to the other. This is the most transparent option and usually the easiest for a general audience. It's also a simple version of small multiples, which we cover in our post on splitting busy charts.
Keep the panels the same width, align the dates exactly, and give each its own clear title. A thin vertical line marking an event, like a campaign launch, across both panels makes cause and effect easier to discuss without implying more than the data shows.
What about a connected scatter plot?
It puts one measure on each axis and connects the points in time order, showing how the two moved together.
Each dot is a month, with ad spend on the x-axis and revenue on the y-axis, joined by a line from the first month to the last. A path heading up and to the right means both rose; a path going right but flat means spend rose while revenue didn't. It's an honest view of the relationship because neither axis is arbitrary relative to the other.
The catch is familiarity. Many readers haven't seen one, so label the first and last points with dates and add a sentence explaining how to read it. Our scatter plot guide covers the basics of reading and labeling scatter plots.
If you must use two axes, how do you limit the damage?
Start both axes at zero, keep the gridlines shared, color each axis like its line, and state the scales in the title or a note.
Sometimes a client or a dashboard template insists on two axes. A few rules make the result less misleading:
Start both axes at zero, so neither line's movement is exaggerated.
Use the same number of gridlines on both sides so they line up.
Color each axis's labels to match its line.
Avoid comments about the lines crossing. The crossing isn't data.
Add a note, in the style of our guide on labels and annotations, saying what each axis measures.
The one case where two axes are harmless is when both show the same quantity in different units, like degrees Celsius on the left and Fahrenheit on the right. There the second axis is just a conversion. For everything else, the line chart guide and the options above will usually tell the story more clearly.
Questions
Are dual-axis charts always bad?
Not always, but they're risky. They can work for expert audiences who know both scales well, or when both axes measure the same quantity in different units, such as Celsius and Fahrenheit. For general readers, an indexed chart or two panels is safer.
How do I index two series to 100?
Divide every value in a series by its value on the start date and multiply by 100. Both series then begin at 100, and a reading of 130 means 30 percent above the start, whatever the original units.
Why do two lines crossing on a dual-axis chart mean nothing?
Because the crossing point depends on how each axis was set. Stretch one axis and the crossing moves or disappears, while the data stays the same.