In 2020, there will be nearly 44 zettabytes (i.e. 44 trillion GBs) of user-generated data through a plethora of sources: from streaming platforms to smartphones, from e-commerce websites and apps to CCTV. A study by IBM showed that by 2017, 90% of the existing data had been produced in just the previous 2 years.
Given this ubiquity of data across all types and sizes of organisations, data visualisation brings four clear benefits:
Standardisation of information processing
Optimisation of investment in Business Intelligence (B.I.) stacks
Fact-based interactive and predictive business narratives
Actionable insight to support decision-making.
This article will show you how to apply best practice through a variety of frameworks and enable you to get the most from your data sets and articulate powerful insights with clarity to your colleagues.
You may be familiar with the S.M.A.R.T. approach, first conceptualised by George T Doran and subsequently updated by several organisations to include E.R.
Specific: clearly establish purpose, importance, outcomes and roles
Measurable: define quantifiable KPIs and /or behaviour criteria for qualitative goals
Achievable: identify any limitations, constraints, and their impact
Relevant: focus on business priorities
Time-Bound: set deadlines
Evaluated: assess progress and evaluate outcome
Reviewed: feedback to support iterative processes
Although initially conceptualised for business contexts, the SMARTER approach is also helpful in defining clear objectives and assessing achievements across all the stages of the CRISP-DM methodology of which visualisation is the tip of the iceberg.
It is essential that any data visualisation practitioner understands the business questions, needs and goals, along with the scope of the project. In particular the type of analysis, key stakeholders, KPIs, deliverables and target audience.
It is also crucial to identify the type (one, two or multi-dimensional; text, etc.) and assess quality of the data extracted, map and resolve any issues (e.g. tracking errors, missing entries or duplicates) or liaise with stakeholders who can. At this stage it is important to do some exploratory analyses both to identify key variables and early insights, but also to formulate hypotheses to expand on those findings.
At this stage, the extract, transform, and load (ETL) process enters the transformation phase: the data gathered from multiple sources needs to be cleansed, formatted and often blended in order to be usable. The structure, content, relationships and derivation rules should be clear and consistent.
Consists in determining the most relevant variables, developing the analytical methodology and best models. This can range from simple timeseries and scatter plots analyses to more sophisticated techniques such as linear regressions and decision trees, among others.
This stage involves the evaluation of the results and their validity, in order to determine if further cleaning is required and if the outcome answers the initial brief.
The focus should be on presenting the results using the most appropriate methods and charts according to the insights found and the target audience. It is important that dashboards showcase findings communicated through a narrative that ultimately supports actionable recommendations and empowers the relevant stakeholders, whilst striking a balance between complexity and clarity. This process should be iterative and fine-tuned based on end-user feedback.
There are three main types of dashboards with different purposes:
Operational dashboards: focusing on monitoring dimensions and metrics that are relevant from an ops point of view which either need to be reviewed regularly or in response to specific events.
Strategic or Executive dashboards: top-level BAU (Business-as-Usual) visualisations of KPIs usually with minimal interaction
Analytical dashboards: provide a high-degree of interactivity in order to support further data exploration in relation to a main subject.
There are three types of data that require different types of dashboarding visual features:
Categorical: the type that belongs together such as geographical data
Ordinal: belongs together and has a logical order (1st place, second and third)
Quantitative: defines values of a metric such as sales, temperature, or income.
Use visual features to enhance comprehension of the information presented.
Form: shape, size, orientation, length, width, enclosure, density (e.g. patterns are ideal to display similar sets of information)
Colour: intensity and hue to emphasize key data points and minimise others; one colour for continuous data; contrasting colours for comparisons; branded palettes; colour-blind friendly; minimise non-data-ink or redundant data-ink.
Position: 2D or 3D shapes (the latter particularly for interactive visualisations)
Motion: flicker, direction and velocity (particularly for real-time data sources)
Interactivity: facilitate exploration by drilling down or up to probe the data
Structural Elements: such as clear, but inconspicuous, tick marks and axes
Labelling: directly label data points on the chart rather than relying on legends
Narrative flow: use text boxes and visual hierarchies
Contextual Information: overlay directly onto the chart
Mobile friendly: create responsive visualisations with libraries such as D3.js or Highcharts as an alternative to desktop
Implementing these tips and working towards a S.M.A.R.T.E.R. data visualisation practice will enable your target audience to comprehend your insights quickly. Increasing key stakeholder engagement with data viz can improve collaboration across your organisation and establish a clear correlation between business operations and activities versus goals. May the data viz force be with you.
We believe that moving too slowly in digital is the biggest risk your business faces. If you are ready to move faster in digital, we are here to help.