Two Measures on One Dashboard Chart: Should They Share the Picture but Not the Scale?
Before adding a second vertical axis to a dashboard chart, ask what readers are supposed to compare. Lines with independent scales can cross or appear to move together without establishing that their values are equal or their changes are comparable. Separate, aligned charts are often a clearer starting point for two different measures.
This is a guide to commissioning a reporting view, not an analysis of your company's performance. The decision is specifically about overlaying measures with separate vertical scales, rather than choosing a logarithmic scale for one measure.
A shared date does not make the values comparable
Imagine a hypothetical service dashboard showing incoming requests and average completion time over the same reporting periods. The dates belong together, but the vertical quantities have different units. A request count and a duration are not quantities that become equal when two lines meet on the screen.
A dual-axis chart can assign one vertical scale to each measure. The resulting positions depend on both sets of axis limits. A crossing therefore cannot be read as one business quantity overtaking the other. Even two measures with the same unit need care if they use different ranges.
Write the intended reading task before selecting the chart. Do you want readers to inspect the timing of changes, compare absolute levels in one unit, or investigate a possible relationship? These are different questions. Combining the series into one compact card is not itself a reason to accept the same visual comparison for all three.
Why the second axis changes the story
The Office for National Statistics explains the problem in its article on dual-axis charts. Its examples show how independently scaled lines can suggest incorrect conclusions about which series is higher, which grew faster and when one overtook another. It also demonstrates that changing the axes can tell different visual stories using the same underlying data.
The concern is not that every stored number has become wrong. The picture invites comparisons that the independent scales do not support. A chart can be numerically faithful when read against each individual axis and still be easy to misinterpret at a glance.
For a commissioned dashboard, I recommend reviewing the actual visual reading task, not accepting a second axis merely because the chart library offers one. Labelling both sides is necessary context if that design is retained, but it does not remove the comparison the overlaid lines invite.
Start with two aligned panels
One alternative discussed by ONS is to separate the series into two charts. Each can retain a useful scale while remaining part of the same figure. That makes the separate quantities easier to recognise without implying that their vertical positions are directly interchangeable.
For the hypothetical requests-and-duration report, commission an upper panel for request counts and a lower panel for completion time. My practical recommendation is to align the reporting dates and make the period definition visible. Keep each panel's measure and unit explicit. If a period is absent from one dataset, do not shift the remaining points to make the panels look aligned.
Separate panels do not prove that one measure causes the other. Their purpose is to support inspection of timing without the false equality implied by crossing lines. An analytical conclusion still requires suitable evidence beyond the layout.
On a narrow screen, stacking panels can preserve their relationship without squeezing two sets of vertical labels into one plot. Test the real labels and intended period range rather than assuming that desktop readability will carry over.
Sometimes one scale with annotations is enough
ONS also suggests retaining a single-axis chart and using annotations to draw attention to changes that might otherwise be overlooked. This is relevant when the measures are genuinely comparable on that shared scale, but smaller movements need explanation.
Do not solve a unit mismatch by simply placing both measures on one axis. Nor should a note claim a dramatic change that the data does not establish. An annotation should explain the observed event or explicitly identified context, with its basis available to the reader.
If the business actually wants a comparison of relative change, specify that as a different derived report. Agree its calculation, reference point, exclusions and interpretation before implementation. Converting both series into an index or percentage is not a cosmetic replacement for axis labels; it changes the question being answered. This article does not validate a particular transformation for your data.
Filters and exports must preserve the reading context
A scale that changes when a filter is applied can make two saved images look more comparable than they are. Decide whether ranges remain fixed within the intended comparison or change with a visible explanation. Preserve that policy in any permitted export.
A screenshot should include enough context to identify each measure, its unit and the reporting period. Where access permissions allow it, a readable table can let users inspect the underlying values without relying solely on visual height. These are proposed product requirements, not capabilities tested in a particular app.
If a dual-axis view is still requested for specialist users, record why the alternatives do not meet the need and agree a deliberate review. Do not describe an expert audience as a guarantee against misunderstanding. The ONS discussion explicitly includes experts who may be reading quickly among those who can draw the wrong visual conclusion.
Test the interpretation, not just whether the lines appear
Include chart meaning in the acceptance criteria for your web app. Use harmless synthetic series with different units and independently variable ranges.
- Check that every plotted series has the agreed measure, unit and reporting period.
- In a split view, verify that matching periods align and that missing observations do not silently move dates.
- Ask a reviewer what a crossing or apparent shared height means. The accepted design should not encourage an unsupported equality or overtaking claim.
- Change the selected period and inspect the documented scale behaviour.
- Review the narrow-screen layout and an exported image with their actual labels.
- Compare readable values with the plotted observations, subject to the same access rules.
These are proposed acceptance cases, not a browser test performed for this article. Bring the two measure definitions and the decision you want to support to a reporting-app discussion. They provide a stronger starting point than a request to fit another line inside the existing card.
Questions to settle before implementation
If the colours match the axes, is the problem solved?
Colour matching can help readers associate a line with its axis. It does not make the vertical positions comparable or stop a crossing from looking meaningful. Review whether a split view communicates the task more clearly.
Do separate charts have to use identical vertical ranges?
Not when they describe different measures or units. The purpose of splitting is to make those independent scales explicit. Their shared time alignment and clear labels still need review.
Does a similar shape establish a business relationship?
No. A visual resemblance can motivate investigation, but this layout does not certify an association or causal explanation. Avoid turning a design choice into a performance claim.
Source and limits
Office for National Statistics: Dueling with axis, the problems with dual axis charts, published 3 July 2019 and read 26 October 2026. The source supports the risks of misleading visual comparisons and the alternatives of annotations or separated charts. The service-dashboard example, filter rules and acceptance cases are practical commissioning guidance. No business measurements, statistical assessment or implemented dashboard results are claimed.