In a profession where hours are spent working with black and white data, analysts relish the opportunity to create data visualisation with colour, shapes, maps, and actions. However, these formatting features can occasionally lead us to get carried away. The resulting visualisations are often poorly designed, noisy and don’t communicate the intended information to the user.
If you’re completely new to the subject, you may be asking ‘What is Data Visualisation?’. In his book ‘Data Visualisation – A Handbook for Data Driven Design’ Andy Kirk succinctly defines the topic as ‘the communication and presentation of data to facilitate understanding’.
In this blog series, I will focus on key data-visualisation best practices and Tableau tips to design and create informative and meaningful dashboards. The following recommendations cover chart type and colour selection and are mainly targeted at budding data analysts.
Choosing the most appropriate chart type is essential to building a clear, informative dashboard. The type of data relationships you are communicating will determine the most appropriate chart to use. I have used the familiar ‘Superstore’ dataset Tableau offers to display these charts.
Analysing trends in data over time is common practice – utilise line, bar and area charts, and ensure the time field is on the x-axis and measure on the y-axis.
Experiment with the granularity of your time field – you may spot different trends depending on whether you’re analysing your data by day, month or quarter.
When looking for relationships between two or more variables, use scatter plots, x-y heatmaps or bubble plots. If using a scatter plot, utilise the size or colour mark detail in Tableau to display a correlation between three measures. Checking different measure combinations will allow you to identify any correlations present in your data.
When ranking attributes, counties, people etc., stick with a horizontally stacked bar graph – this chart clearly shows both ranking and values to the end user and allows rapid comparison between different categories.
For part-to-whole visualisations, avoid pie charts (especially 3D or exploded pies) at all costs. Our visual system struggles when calculating and differentiating area, making a comparison between different values difficult. A better chart type is the percentage total bar chart, which allows easier comparison of contribution to the whole from each value.
Tip – percentage total is available as a quick table calculation in Tableau. Right click on the measure to access.
When analysing the distribution of values within your data, box and whisker plots or histograms are best used. Box and whisker plots offer an easy comparison of multiple distributions and include minimum, maximum, 25th and 75th percentile and median values. If you’re more interested in the distribution of data in a single field, a histogram may be better suited.
A key differentiating feature between Tableau and entry-level data processing programs is the ability to map spatial data. Spatial data can be accompanied by additional measures and dimensions to create powerful visualisations that highlight key geographical relationships between your data and the location.
With spatial data, Tableau’s tooltips will really come into their own. When a user hovers over a location, additional data can be displayed. Tooltips can be designed to present a range of data and can even include graphs and charts (more on that in subsequent blogs!).
Useful tip – for visualisations where location shape or position isn’t critical, try using a hex map. There are many tutorials on how to construct these within the tableau community, utilising pre-constructed coordinates (thanks Tableau community!), for various popular locations.
I have only briefly touched on a few chart types for each data relationship. For additional chart types and when to use them.
Be sure to check out Andy Kriebel’s ‘Visual Vocabulary’ viz on Tableau public: https://public.tableau.com/profile/andy.kriebel#!/vishome/VisualVocabulary/VisualVocabulary
In addition, the Data Viz Project offers a useful tool for matching input data types to chart types: https://datavizproject.com/
When talking about data-vis best practice, one name that regularly crops up is Stephen Few. His books and website ‘Perceptual Edge’ are comprehensive sources of covering data-vis and dashboard design best practice and optimization: http://www.perceptualedge.com/examples.php
Finally, check out Tableau’s whitepaper on the topic: https://www.tableau.com/learn/whitepapers/tableau-visual-guidebook
Colour is effective at distinguishing different dimensions, attributes or measures in your visualisation. Keep colour schemes simple and utilize different colours and shading to demonstrate differences or similarities in values or rankings. Be careful not to include too many values on colour, as this makes it difficult to determine different values. Incorporating the colour scheme of an organisation or cooperation is often appreciated, but be sure to not let them distract from key information.
Tableau’s default colour pallet is specifically designed to cater for colour blindness, so using this is good practice if you’re unsure on your colour scheme. My fellow DSAU3’er Grace recommended a google chrome extension that lets you check your vis for colour blind appropriateness: https://chrome.google.com/webstore/detail/colorblinding/dgbgleaofjainknadoffbjkclicbbgaa?hl=en
Quick trivia: Colour blindness affects approximately 1 in 12 men and 1 in 200 women
Finally, include an appropriately sized and positioned colour legend to convey the significance of each colour.