OperationsFramework / modelModelAccessible

Visual analytics

How can visual analytics improve people, teams, or organisational effectiveness?

AccessibleTacticalTeam2 min read

Contents

Data can be analysed in different ways and the most simple method is to create a visual or graph and look at it to spot patterns.

Visual analytics combines computational analysis, interactive visualisation and human judgement. It helps people explore complex data, notice patterns and exceptions, refine analytical models and communicate findings in a form that supports a decision.

When to use it

Use visual analytics when data volume, variety or interdependence makes a static report insufficient, yet the question still requires contextual human reasoning. It is especially useful for exploratory analysis, geographic patterns, networks, changing trends and model diagnosis.

Computers can integrate records, calculate quickly and search many possibilities. People contribute domain knowledge, flexible questions and interpretation. Interactive views connect the two: an analyst can filter, zoom, compare, alter assumptions and send the result back into the model.

Visual analytics can address questions such as:

Where are the most valuable customers located?

Which attributes distinguish them?

How is market share changing across time or segment?

Does the apparent relationship between factor X and factor Y survive closer analysis?

Origins

Visual analytics emerged from scientific visualisation, information visualisation, statistics, human–computer interaction and data mining. Its growth reflects a genuine problem: collection and processing capacity can outpace people’s ability to interpret results.

A widely repeated “knowledge doubling” story attributed to Buckminster Fuller in 1981 claimed that knowledge once doubled over a century, later every 25 years, then every 13 months, with an IBM prediction of every 11 hours. These figures are difficult to define and should not be treated as a validated law. Their useful point is narrower: more available information does not automatically create understanding. Interactive analysis is intended to close that gap.

What it is

The field is a feedback loop rather than a final chart. Data are transformed and modelled; visual representations expose structure and uncertainty; a person interrogates the result; and those interactions guide new transformations or models.

The same historical claim associated with 1981 used intervals of 25 years, 13 months and 11 hours to dramatise information growth. Whether or not those estimates are credible, the practical challenge remains: analysts need methods that scale computation without removing human scrutiny.

A useful visual analytical system therefore supports overview and detail, comparisons, filtering, provenance and revision. It should also reveal missingness and uncertainty rather than presenting a clean picture that disguises weak data.

How to use it

The VisMaster approach can be understood as an iterative movement among data, visualisation, knowledge and models.

First, define the decision and integrate relevant data. Clean, normalise, group and document it while preserving enough lineage to trace a surprising result back to its source.

Next, apply suitable automatic methods to generate summaries, relationships or models. Evaluate assumptions and performance before visualising the output. A sophisticated model does not become trustworthy merely because it looks compelling.

Then interact with the visual representation. Filter segments, alter parameters, compare views and inspect outliers. Use domain knowledge to formulate new questions, but guard against searching until a pleasing pattern appears. Alternate computation and visual inspection until the finding is stable enough for the decision.

Finally, communicate the conclusion with the evidence, uncertainty and limitations needed by the audience. Remove interactions and visual encodings that do not help answer the stated question.

Practical example

Hans Rosling demonstrated the method by challenging a simple split between a developed world with long lives and small families and a developing world with short lives and large families. Answering the question required birth and mortality data across countries and decades—far too much for useful inspection as raw tables.

Rosling mapped children per woman against life expectancy and animated countries through time. In 1950, the proposed split described much of the data; by 2007, many countries had moved toward smaller families and longer lives, although exceptions such as Afghanistan remained. The animation made both the historical movement and the inadequacy of a fixed label visible.

See the original talk: Hans Rosling

Top practical tip

State the decision question beside the visual and design every interaction around it. Show units, definitions, uncertainty and source lineage so users can interpret the pattern rather than merely admire the display.

Top pitfall

Interactivity can encourage endless slicing until a dramatic pattern appears. Do not confuse exploration with confirmation: validate important findings, account for multiple comparisons and present only the views that clarify the decision.