The Art of Making Data Tell the Truth

Charts lie all the time. Not through fabricated data, but through design choices that distort perception: truncated axes, cherry-picked time ranges, misleading scales, and 3D effects that warp proportions. Understanding these techniques doesn't just help you spot manipulation — it makes you a better communicator when you create visualizations yourself.

The truncated Y-axis

This is the most common chart crime. When a bar chart's Y-axis starts at 95 instead of 0, a 1% difference looks like a 100% difference visually. Politicians and advertisers use this constantly to make small changes look dramatic. The fix: always check where the axis starts, and be suspicious of any chart where the bars don't start at zero.

Cherry-picked time ranges

Select the right start and end date, and almost any trend can be made to look like growth or decline. A company showing its stock performance "since our low point in March" rather than year-over-year is almost certainly obscuring an unflattering longer-term picture.

Correlation presented as causation

Spurious correlations are everywhere. Nicolas Cage movies released per year correlates almost perfectly with swimming pool drownings. GDP correlates with cheese consumption. These are meaningless — but correlation is visually compelling, which is why it gets abused constantly in data journalism.

What honest visualization looks like

Edward Tufte's principle of data-ink ratio — maximize the proportion of ink that actually represents data — is still the best single rule. Show full scales. Label axes clearly. Don't use 3D. Don't use pie charts with more than five slices. And always include sample size and confidence intervals when presenting statistics.

The psychology of visual perception

Effective data visualisation isn't just aesthetics — it's applied cognitive science. Human visual perception has specific characteristics that good charts exploit and bad charts ignore. Preattentive attributes — features like colour, size, position, and shape — are processed by the brain before conscious thought, in under 250 milliseconds. Position along a common scale is processed most accurately; colour hue least accurately for quantitative comparison.

This is why bar charts outperform pie charts for comparison tasks: comparing bar heights using a shared baseline is faster and more accurate than estimating angular areas. It's not opinion — it's reproducible in controlled studies. The science of visualisation provides principled guidance that separates decoration from communication.

Chart selection is a decision, not a default

The most common visualisation mistake is defaulting to whatever chart type the software suggests first. A line chart implies continuity and is appropriate for time series data. A bar chart is correct for categorical comparison. A scatter plot reveals correlations between continuous variables. Using the wrong chart type — a line chart for unordered categories, a pie chart for more than five segments — actively misleads the viewer.

Before selecting a chart type, ask: what relationship am I trying to communicate? Comparison, distribution, composition, or relationship? Each has chart types suited to it, and the choice should be deliberate. Edward Tufte's concept of "data-ink ratio" is useful here: maximise the proportion of ink devoted to data, and minimise decorative elements that convey no information.

Colour: the most misused tool in visualisation

Colour should do work, not decoration. Using colour to distinguish categories is appropriate only when the distinctions matter. Gratuitous colour variation in a single-category bar chart adds visual noise without information. Sequential colour scales (light to dark) are appropriate for quantitative data; diverging scales (two colours from a neutral midpoint) work well when there's a meaningful zero or boundary.

Critically, approximately 8% of men and 0.5% of women have some form of colour vision deficiency. Visualisations that rely on red/green distinctions alone are inaccessible to a significant fraction of your audience. Accessible colour palettes like ColorBrewer provide options that work for colour-blind viewers and in greyscale printing.

Interactive visualisation and when it helps

The proliferation of JavaScript libraries like D3.js, Plotly, and Vega-Lite has made interactive visualisation accessible to any web developer. But interactivity adds cognitive overhead — users must discover and understand the interaction model before benefiting from it. For most analytical communication, a well-designed static chart outperforms a complex interactive one.

Interactivity earns its complexity when users have genuinely different questions that a single static view cannot answer, or when exploration of large datasets is the goal rather than communication of a specific finding. Dashboards for operational monitoring, exploratory data analysis tools, and consumer-facing data products are appropriate homes for rich interactivity. An executive summary slide is not.