Five best practices for data visualisation

Gustav Piater

To tell a coherent story that is easy to absorb and convincing, your visualisation must also match the data set and your objectives. It should be optimised (to promote information absorption) and it should be astutely shared (to prompt action).

These requirements are dealt with in five visualisation best practices:

1. The right chart type for the job – what must it do?

Comparisons

Column charts (vertical bars) are good for comparing values across categories (e.g. sales by product). Clustered columns/bars, in turn, are useful for comparing values across multiple category groups (e.g. sales by product by gender). And trellis charts can be used to compare values of multiple categories across multiple dimensions (e.g. sales by product by region).

For their part, bar charts (horizontal bars) are great for comparing values of more than 12 categories (e.g. sales by agent) while bullet/layered charts can be used to track when one metric overtakes another (e.g. department spend vs budget), and radar/spider charts highlight strengths and weaknesses by comparing multiple dimensions/measures (e.g. products by features).

Comparisons over time

Time-based comparisons can employ line charts (e.g. to show sales by month), multi-line charts and trellis charts (e.g. sales by product by month), step charts (e.g. balance by hour), area charts (e.g. sales volume by month) and week density displays (e.g. usage by day by hour).

Composition

To show part-to-whole relationships, we use stacked bars and percentage bars (e.g. for product sales by region), stacked areas (e.g. sales by month per product), funnels (e.g. sales by funnel stage), waterfalls (e.g. employees per team over time), tree maps (e.g. sales by product group) and pie/ring charts (e.g. revenue by gender).

Correlation

Relationships between variables can be shown in scatter plot visualisations (e.g. showing customer-age correlation), heat grids (e.g. region-product correlation), event charts (e.g. event-sales correlation), bubble charts (e.g. customer-age-revenue correlation) and circle grids (e.g. region and product by sales and profit).

Distribution

Chart types that show frequency of values in a data set include histograms (e.g. showing customers by age group), box and whisker charts (e.g. revenue by region) and scatter plots (e.g. individuals by height and weight).

Location

To show where things are happening (or aren’t, if you’re looking to plug a gap), use thematic/GIS maps (e.g. population by province), heat maps (e.g. usage by location), bubble maps (e.g. sales by store) or raster maps (e.g. sales by shop floor sections).

KPIs

The following visualisations will show if you are on target: Big number charts (e.g. total number of customers), dials or thermometers (e.g. revenue vs target), (speedo)meters (growth vs target) and bullet/layered charts (as above).

2. Format your charts to make them more understandable and appealing

Colours

Colours, when used well, distinguish between data categories or introduce a secondary metric. Used poorly, it can obscure meaning and confuse the reader. It should work for colour-blind people (8% of men).

Legends

Legends define colour schemes, so don’t use them if you only have one data category.

Grid lines

Grid lines help you compare key thresholds without your mind having to draw an imaginary line – but too many can make it hard to decipher your chart.

Labels

While charts primarily reveal patterns in your data, labels enhance visuals by adding exact values.

3. Say what your chart is about

Chart titles and descriptions

Chart titles frame the purpose and meaning of your data while descriptions provide additional context. Use both to tell a story (explain the insight, e.g. exponential growth in website traffic) or, more neutrally, to describe the query (website traffic per quarter).

Sorting

Sorting can help people make sense of what they see, if the order of the story is important. Alphabetical sorting helps with finding the right category amid a lot of information while ascending and descending orders might make sense in different circumstances.

Annotations and comments

Use this to provide additional context.

4. Direct your audience’s attention to what’s important

Conditional formatting, reference lines or trend lines and forecasts all increase the dwell time of reports, leading to a better understanding of the data when there’s a lot competing for their attention.

5. Share, share, share

Lastly, as said above, the right chart, dolled up to best effect, is only two-thirds of the battle. Real value is only created when insights are shared and acted on.

Deliver your insights to decision makers at the point (when and where) and in the manner of their customary decision-making! Personalised dashboards help people monitor what matters to them. Embedding insights into platforms that people use every day increase the likelihood of it being seen. Broadcasting insights with periodic reports and alerts help drive action. And live interactive charts tell compelling stories to live audiences.

Good luck creating actionable insights for your organisation!