An independent study reference written by Dr Phuc V. Nguyen. It is not official subject material — for assessment requirements always follow your subject outline and vUWS.

Visual Communicationintermediate

Common visualisation mistakes

Q: A dashboard flags accounts as at risk with a red fill and healthy with a green fill, with no other difference. Which fix best addresses the accessibility problem without losing the visual cue?

Keep the two hues, and add a distinct icon and a text label to each status. The failure is that meaning rides on hue alone. Adding a second, non-hue channel such as an icon plus a text label makes the status readable for a colour-vision-deficient reader, in greyscale, and on a poor projector, while the colour still helps everyone else scan quickly. A brighter red does not help a reader who cannot separate red from green, and the other options either discard a useful cue or introduce a new contrast problem.

Q: A chart shows monthly advertising spend and monthly revenue on two separate vertical axes, and the lines move together. What should a careful analyst say?

Scale choice sets how closely the lines appear to track, so the strength of any relationship has to come from a scatter plot or a calculated measure rather than from this chart. Two independent scales let a designer stretch or shift either series, and that changes how close the lines look, their relative slopes, and how large each movement appears. It does not invent turning points or move when a peak occurs, so shared timing is real and can be read off the chart. What cannot be read off it is how strongly the two move together, because that is exactly the part the scales control. Plot spend against revenue as a scatter, or compute the relationship, and note that either one still measures association rather than establishing that advertising drives revenue. Two series can legitimately appear together, on a shared scale or in stacked panels, and starting both at zero does not remove the free choice of range.

A short list of failures accounts for most unreadable business charts. Decoration that competes with the data, including three-dimensional effects and heavy backgrounds. Dual axes, where the two chosen scales set how closely the series appear to track. Overplotting, where thousands of points collapse into a solid mass. Category order left alphabetical when the reader needs rank. Encoding by colour alone, which fails for readers with colour vision deficiency and in greyscale printing. Low contrast and small type, which fails everyone in a meeting room. Accessibility is not a polish step at the end, it decides whether the chart works.

Why it matters

Most bad charts are not the product of deception. They are the product of defaults. The software offers a depth effect, an alphabetical sort and a red-green palette, and nobody changes them. Treat every default as a decision somebody else made without knowing your reader. Sorting by value, dropping the depth effect and checking the palette takes about a minute, and it fixes most of what goes wrong.

Before you read on — recall

A dashboard flags accounts as at risk with a red fill and healthy with a green fill, with no other difference. Which fix best addresses the accessibility problem without losing the visual cue?

Formulas

Data-ink ratio

\text{data-ink ratio} = \dfrac{\text{ink used to display data}}{\text{total ink in the graphic}}

Tufte proposes raising this ratio by removing marks that carry no information, such as heavy gridlines, borders, background fills and duplicated labels. Treat it as a direction rather than a target. Pushed to the limit it also strips gridlines and reference lines that genuinely help a reader read values, so the useful version of the rule is to delete what carries nothing, not to delete everything that is not a data point.

Contrast ratio, as defined for accessibility

C = \dfrac{L_{1} + 0.05}{L_{2} + 0.05}

Here

L_1

is the relative luminance of the lighter colour and

L_2

that of the darker, each on a scale from zero to one. The published thresholds are a ratio of at least four point five to one for ordinary text, and at least three to one for large text and for graphical elements such as chart lines and icons. A grey of hex value 767676 on white gives about four and a half to one and passes, while the lighter 999999 that many tools use by default for axis labels gives about two point eight to one and fails.

Worked examples

Scenario

A project status report marks every workstream red, amber or green using colour fill only. One senior stakeholder consistently misreads which workstreams are in trouble.

Solution

Red-green colour vision deficiency is common, and red and green fills of similar lightness become nearly indistinguishable. Encoding by colour alone also fails when the report is printed in greyscale or projected on a washed-out screen. The fix is redundant encoding. Keep the colour, and add a second channel that does not depend on hue, such as a distinct icon per status, a short text label in the cell, and a real difference in lightness rather than only in hue. The report then works for every reader in the room.

Scenario

A marketing chart plots spend on the left axis and sign-ups on the right axis. The two lines track each other closely, and the team concludes that spend drives sign-ups.

Solution

With two independent scales, the designer chooses where each line sits. Shifting either axis range changes how closely the lines appear to track, so the visual agreement is partly an artefact of that choice. Better forms are two stacked panels sharing a time axis, so each series is read on its own scale without the implied alignment, or a scatter plot of spend against sign-ups, which shows the relationship directly. Establishing that spend causes sign-ups needs a design capable of supporting a causal claim, not a chart.

Common mistakes

✗A chart with correct numbers cannot be a bad chart. Correctness is about the values, readability is about whether the reader extracts them. A chart can be arithmetically perfect and still leave an audience with the wrong impression through ordering, scaling or colour.
✗Accessibility only matters for public-sector or regulated work. Colour vision deficiency, low-contrast projectors and greyscale printing are present in ordinary meetings. Designing for them improves legibility for everyone, which is why contrast and redundant encoding count as basic quality rather than compliance.
✗Removing everything except the data always improves a chart. Reducing non-data marks helps up to a point. Gridlines, reference lines, direct labels and units are not data, and removing them makes values harder to read. The rule is to remove what carries nothing.
✗A dual-axis chart is fine as long as both axes are labelled. Labels do not remove the problem, because how closely the two lines appear to align is still set by two freely chosen scales. Separate panels or a scatter plot avoid the implication entirely.

Revision bullets

•Most bad charts come from unchanged software defaults
•Sort categories by value unless the reader needs another order
•Dual axes let the designer set how closely two series appear to track
•Never encode meaning by colour alone, add shape, label or lightness
•Contrast floors: four and a half to one for text, three to one for graphics
•Cut non-data ink, but keep the gridlines and labels that aid reading

Quick check

A dashboard flags accounts as at risk with a red fill and healthy with a green fill, with no other difference. Which fix best addresses the accessibility problem without losing the visual cue?

A chart shows monthly advertising spend and monthly revenue on two separate vertical axes, and the lines move together. What should a careful analyst say?

Connected topics

Correlation vs Cause Measures & Dimensions Visual Variables Choosing a Chart Dashboards Truncated Axes Chart Ethics Ethics by Design

More in Visual Communication

Purpose of Visuals Data Storytelling

Sources

Tufte (2001)
Tufte, E. R. The Visual Display of Quantitative Information. 2nd ed. Graphics Press, 2001.
Source of the data-ink ratio and of the treatment of chartjunk and decorative effects.
W3C, Web Content Accessibility Guidelines 2.2
World Wide Web Consortium. Web Content Accessibility Guidelines (WCAG) 2.2. W3C Recommendation, 2023.
Defines the contrast-ratio calculation and the thresholds, and the use-of-colour criterion requiring a second cue.
Few (2012)
Few, S. Show Me the Numbers: Designing Tables and Graphs to Enlighten. 2nd ed. Analytics Press, 2012.
Catalogue of recurring business-chart failures with worked before-and-after redesigns.

How to cite this page

Dr. Phil's Quant Lab. (2026). Common visualisation mistakes. Derivatives Atlas. https://phucnguyenvan.com/concept/ba-visualisation-mistakes