Data Visualisation is the representation of data through graphical, spatial, interactive or other visual forms so that patterns, distributions, comparisons, relationships, trends and anomalies can be examined more effectively. Common forms include bar charts, line charts, scatter plots, histograms, box plots, heat maps, maps, network diagrams and statistical graphics designed for specific analytical tasks. Visualisation can serve exploratory purposes during analysis, explanatory purposes when communicating findings and operational purposes when monitoring changing information. Effective visualisation is not merely decorative presentation; it encodes quantitative or categorical information using visual variables such as position, length, area, colour, shape and orientation. The choice of encoding affects how accurately viewers can perceive and compare the underlying values.
Data Visualisation has a long history predating digital computing. Statistical graphics developed through the nineteenth and twentieth centuries as governments, scientists and researchers sought to communicate demographic, economic, geographic and scientific measurements. The development of mechanical and electronic computing later enabled rapid generation of graphics from large datasets. Modern software supports interactive filtering, zooming, linked views, animation and real-time dashboards. Digital visualisation also permits users to explore multidimensional datasets that would be difficult to represent in static form. However, additional technical capability does not guarantee better communication; interactive complexity can obscure important patterns if controls, scales or defaults are poorly designed.
A chart should be selected according to the analytical relationship that needs to be communicated. Bar charts are generally suited to comparing discrete categories, while line charts commonly show ordered measurements over time or another continuous sequence. Scatter plots reveal relationships between two quantitative variables and can expose clusters, nonlinear patterns and outliers. Histograms display the distribution of a quantitative variable by grouping observations into bins, making bin width an important design decision. Box plots summarise selected distributional characteristics and can facilitate comparison among groups, while density plots provide a smoothed representation of distributions. Pie and area encodings can be harder to compare accurately because humans generally judge lengths and positions more precisely than angles or areas.
Visual integrity depends on accurate scales, appropriate transformations and explicit context. Truncating an axis can exaggerate visual differences in some charts, although there are specialised situations where non-zero baselines are appropriate if the encoding remains clear and the design does not imply a misleading magnitude relationship. Logarithmic scales can be appropriate for quantities spanning several orders of magnitude but require clear labelling because equal distances represent multiplicative rather than additive changes.
Dual axes can create misleading apparent relationships when independent scales are combined. Missing data, censored values and uncertainty should not be silently presented as ordinary observations. Confidence intervals, prediction intervals, error bars and uncertainty bands can communicate statistical uncertainty when their definitions are clearly stated.
Colour is an important visual encoding but must be used carefully. Sequential colour scales can represent ordered magnitudes, diverging scales can represent deviations around a meaningful midpoint and categorical palettes can distinguish nominal groups. Excessive colour variation can make charts difficult to interpret, while red-green combinations can create accessibility problems for some viewers. Colour should not be the sole carrier of essential information when an alternative encoding can be added. Labels, shapes, line styles and annotations can improve accessibility and reduce ambiguity. Data Visualisation should also consider screen size, printing conditions, contrast, text readability and assistive technologies when intended for broad audiences.
Interactive Data Visualisation introduces additional concepts such as brushing, filtering, linked views, drill-down, tooltips and dynamic aggregation. Dashboards often combine several visualisations with controls that allow users to examine operational or analytical data from different perspectives. Real-time systems may display streaming observations, requiring decisions about update frequency, buffering, aggregation and treatment of missing or delayed data. Interactivity can support discovery but can also introduce reproducibility problems if users can manipulate filters without recording their state. Good analytical systems therefore provide clear defaults, explicit units, documented definitions and mechanisms for preserving or communicating the selected view.
Data Visualisation is closely related to statistical analysis and exploratory data analysis because visual displays can reveal assumptions or anomalies before formal modelling. A scatter plot may reveal heteroscedasticity or nonlinear structure that would be obscured by a single correlation coefficient. A distribution plot can identify skewness or multimodality that affects model selection. Geographic visualisations can expose spatial clustering, but apparent patterns can also arise from differences in population density, map projection, classification thresholds or denominator choice. Visual correlation does not establish causation, and attractive graphics can create persuasive but unsupported impressions if scales or comparisons are inappropriate. Visualisation therefore complements quantitative analysis rather than replacing statistical reasoning.
Good Data Visualisation requires an explicit audience, analytical purpose, accurate data transformation and a visual encoding suited to the task. Designers should preserve the relationship between displayed geometry and underlying quantities, state units and definitions, distinguish observed values from estimates, and make uncertainty visible where it matters. Common problems include excessive decoration, unnecessary three-dimensional effects, overloaded dashboards, inconsistent scales, hidden denominators, selective time windows and inappropriate colour schemes.
Reproducible visualisation pipelines can generate graphics directly from validated datasets using code, reducing manual transcription errors and allowing changes to be tracked. In scientific and technical contexts, a visualisation is most useful when it makes relevant structure easier to inspect while retaining enough context for the viewer to understand exactly what the displayed data represent.
