Data visualization mistakes that hide the insight

A chart should make one comparison easy. Bars suit category comparison, lines suit change over time, histograms and box plots suit distribution, and scatter plots suit relationships. A pie chart can show a few parts of one whole, but it struggles with many slices or small differences.

The common failure is choosing a visually interesting chart before deciding the question. A stacked area chart may look impressive while making it impossible to compare individual series. Start with the sentence the chart should support, then choose the simplest form that supports it.

Bar length encodes magnitude, so a bar chart should generally start at zero. A truncated axis exaggerates small differences and is one of the quickest ways to lose trust. Line charts can use a narrower range when the change is small, but the axis must be labelled clearly and the choice explained.

Dual axes create a visual relationship that may not exist. If two series must share a page, consider indexing them to a common base or using two aligned charts. In small multiples, keep the scales consistent unless the variation itself is the point.

More series, labels, and colours do not make a chart more complete. They make the important signal harder to find. A dashboard with twenty visuals often has no clear entry point. Decide what the viewer should notice first, then remove anything that competes with it.

Progressive disclosure helps: summary first, filters and drill-downs second, detail last. Small multiples can replace a crowded multi-line chart when the comparison is between segments. If a visual needs a paragraph of explanation, it is probably doing too much. A useful test is to cover the title and see whether the main message is still obvious from the chart alone.

Colour should carry information, not decoration. Use a consistent palette for the same categories across pages, and reserve a strong accent for the point that matters. A rainbow of series colours forces the viewer to consult the legend repeatedly.

Check contrast and avoid relying on red and green alone, since colour vision varies. Labels, patterns, or direct annotations can carry the same meaning. When a chart highlights an exception, grey the rest of the data rather than adding another bright colour.

A number without a comparison is difficult to interpret. Show a target, a prior period, a benchmark, or a relevant range. State whether the period is complete, whether the definition changed, and how many records sit behind the point. A spike based on three transactions deserves a warning, not a headline.

Context also means honest uncertainty. Forecasts, survey results, and small samples should show a range or a caveat where it changes the reading. The aim is not to clutter the chart but to prevent a confident conclusion that the data cannot support. Where a comparison is missing, a small table may communicate context better than another chart.

A title such as 'Revenue by month' tells the viewer what the axes contain. A title such as 'Revenue fell after the April price change' tells them what to notice. The second is more useful, provided it is supported by the chart and the analysis.

Use the subtitle for method, period, and definitions. Annotate the turning point rather than forcing the viewer to infer it. Good visualisation reduces the work needed to reach the insight while keeping the evidence visible. Avoid titles that assert a cause the chart cannot establish; 'revenue fell after the price change' is safer than 'the price change caused revenue to fall' unless the analysis supports the stronger claim.