Data visualisation is the representation of data through graphical or visual forms so that patterns, comparisons, distributions, trends, relationships, anomalies and other characteristics can be understood more readily. Common visual forms include bar charts, line charts, scatter plots, histograms, box plots, maps, heat maps, area charts, tables and carefully designed dashboards.
Visualisation is not merely decoration added after analysis; it is an analytical and communication technique that uses spatial position, length, shape, colour, size and other visual properties to encode information. A well-designed chart can make a trend, outlier or difference immediately apparent, whereas a poorly designed chart can obscure the same information or create a misleading impression. Effective visualisation therefore depends on matching the visual representation to the data type and analytical question. A time series generally benefits from a line-based representation, categorical comparisons often suit bars, distributions can be shown with histograms or box plots, and relationships between two numerical variables can often be explored with scatter plots.
The principles of data visualisation are rooted partly in statistics, cartography and information graphics. William Playfair is widely associated with early use of graphical economic charts, including line graphs and bar charts in the late eighteenth century, while later statistical and cartographic developments expanded the systematic use of visual representations.
Florence Nightingale used statistical diagrams to communicate patterns in mortality, and Charles Joseph Minard's nineteenth-century visualisation of Napoleon's 1812 campaign is frequently discussed for its combination of geography, direction, quantity and time. Modern visualisation has been transformed by computers, allowing interactive filtering, zooming, animation and multidimensional exploration. The historical progression illustrates an enduring principle: visual encoding should make important quantitative relationships easier to perceive. Modern software makes it easy to produce complex graphics, but technical capability does not remove the need for sound statistical reasoning. A visualisation should communicate the structure of the data accurately rather than simply maximise visual complexity or novelty.
Choosing an appropriate chart begins with identifying the analytical task and the variables involved. Comparison asks which categories differ and is often served by bars or dot plots. Change over time can be shown with lines, while distributions require representations that reveal spread, skewness, concentration and outliers. Relationships between variables may be examined with scatter plots, while geographic questions may justify maps when location itself carries meaning.
Part-to-whole relationships can sometimes be represented with stacked bars or carefully used pie charts, although many comparisons are easier to make with aligned lengths than with angles. Tables remain valuable when users need precise values rather than rapid pattern recognition. Visual encodings have different perceptual accuracy: people generally compare position and length more reliably than area or angle, which is one reason a simple bar chart can often communicate quantitative differences more effectively than elaborate pictorial graphics. The selection should also account for audience, display size, accessibility, number of categories and whether the purpose is exploration, explanation, monitoring or formal reporting.
Data visualisation can support both exploratory data analysis and explanatory communication. Exploratory visualisation is used by analysts to investigate unfamiliar datasets, identify unusual observations, discover possible relationships and generate hypotheses. At this stage, interactive filtering and multiple representations can be particularly useful because the analyst does not yet know which patterns matter.
Explanatory visualisation is designed to communicate a specific finding or argument to an audience and normally requires stronger editorial choices about what to display and how to organise it. Business reports may use visualisations to explain sales performance, customer behaviour, operational bottlenecks, financial results or geographic distribution. Scientific visualisation may represent experimental measurements, simulations or spatial phenomena. Operational visualisation can support real-time monitoring by displaying current states and alert conditions. The same dataset can therefore produce very different visualisations depending on the decision or question. A chart should always have an identifiable purpose, even when that purpose is exploratory rather than communicative.
Misleading visualisation can arise through inappropriate scales, truncated axes, inconsistent intervals, distorted areas, excessive decoration, selective data ranges, inappropriate aggregation or failure to represent uncertainty. A bar chart with a vertical axis beginning far above zero can exaggerate visual differences when absolute magnitudes matter, although truncation is not inherently invalid for every chart type.
Logarithmic scales can be highly appropriate for data spanning several orders of magnitude but require clear labelling because the visual relationship differs from an ordinary linear scale. Aggregation can conceal important variation, while averages can hide distributions and subgroups. Correlation shown visually should not automatically be interpreted as causation. Missing data and measurement limitations should also be disclosed when they materially affect interpretation. Colour requires particular care because excessive colour, similar hues, cultural assumptions and colour-vision deficiencies can reduce comprehension. Accessibility may require adequate contrast, descriptive text, direct labels, patterns or alternative tabular representations. Accuracy in visualisation therefore includes not only numerical correctness but also avoiding design choices that encourage a materially incorrect interpretation.
Interactive visualisation extends static charts by allowing users to filter, drill down, sort, zoom, select time periods, compare segments or inspect individual observations. Such interaction can be powerful for large datasets because displaying every detail simultaneously would overwhelm the user. However, interaction can also hide important information when users do not know which filters to apply or when the default view is misleading. Dashboards frequently combine several visualisations with filters and key performance indicators, turning visualisation into an operational interface.
Modern business-intelligence tools can connect charts directly to data warehouses and refresh them automatically. More advanced visualisation systems can handle geospatial data, network structures, hierarchical relationships and high-dimensional datasets. The technology should remain subordinate to the analytical objective: a sophisticated interactive visualisation is not necessarily better than a simple static chart if the latter communicates the relevant relationship more clearly and reliably.
Visualisation terminology includes concepts such as axes, scales, marks, encodings, dimensions, measures, aggregation, granularity, filtering, facets and annotations. A dimension generally describes a category or grouping, while a measure is a numerical value that can be aggregated or analysed. Granularity refers to the level of detail represented, such as individual transactions versus monthly totals. Aggregation changes the information available to the viewer and can affect conclusions.
Small multiples divide a visual comparison into repeated panels using a common structure, often making patterns across categories easier to compare. Annotations can direct attention to important events or values without altering the underlying data. Data labels can improve precision but can also create clutter when every observation is labelled. Visual hierarchy helps viewers understand what should be examined first. These concepts matter because visualisation is essentially a mapping between data structures and perceptual structures, and each mapping involves choices that influence what users notice, compare and remember.
Effective data visualisation combines statistical correctness, perceptual clarity, contextual information and audience awareness. A business executive may need a concise view of revenue, margin and material changes, while an analyst may need distributions, confidence intervals and detailed filtering. A public-facing chart may require particularly clear definitions, sources, units and methodological notes. Good visualisation makes the underlying data easier to understand without pretending that uncertainty or complexity does not exist.
It should make units, time periods, denominators and relevant comparison bases clear, because percentages without denominators and growth rates without baselines can easily be misunderstood. In operational environments, visualisations should also distinguish between normal variation and genuinely actionable exceptions. The ultimate purpose is not to make data look attractive but to improve human perception and reasoning about quantitative information. A strong visualisation allows the audience to answer the intended question more accurately, quickly and consistently than they could from the unstructured data alone.
