Data School Week 2: All Things Tableau | DS23

by Harry Osborne

This week, DS23 reconvened on more familiar territory, trading in Carl and Chris’ expertise on Alteryx for a week focused on Tableau with Andy and Lorna.

First off, the aim was to delineate the main ways that graphs and data should be presented, looking at the way shapes, colours and patterns are best interpreted by humans. Further to this end, we spent some time critiquing and redesigning visualisations made by organisations not privy to such information (plain to see, given some of the glaring errors on display).

This also included sections spent on the ins and outs of formatting, sizing and chart selection, particularly when trying to get across particular points (e.g. a line chart to show change over time). Some of the tools we employed included:

  • Excalidraw - a freehand design web page that lets you mock up outlines for dashboards and charts, allowing you to test out layouts and structure
  • - a colour-palette generating site for creating your own unique colour schemes, including hex codes and customisable randomisation functions
  • The Data Visualisation Catalogue - a full breakdown of (nearly) every conceivable chart and graph type, sortable by function

An example of how Excalidraw can be used to set out and trial dashboards

Later in the week, we moved more into constructing our own charts, plotting nearly all of the standard visualisation methods you would try to use when working with Tableau. This further expanded into the functionalities of Tableau, talking about basic terminology, key elements when building charts, and even some slightly more in-depth case uses of more advanced techniques like combined sets and parameters.

This saw us attempt a Workout Wednesday challenge from 2018, looking at building a parameter-directed dashboard for discovery purposes. After a full day of parameter-based challenges and experimentation, I managed to actually produce something relatively similar.

Workout Wednesday: Work in Progress (parameters on the left)

Friday heralded the arrival of a new task - create a visualisation in three hours on a dataset selected for you, but work primarily with the focus of the design, not the analysis. I am normally someone who tries to work the other way around, so this was a fair change in direction for me; nevertheless, I went to work, using a dataset built on CDC information about STD cases in the USA between 1996-2014.

By around 10:30 I was happy with the individual sheets I had made, so spent the remainder of my time trying to master containers, spacing and the distribution of information on the dashboard, particularly to convey a certain viewpoint and message. I was happy with what I had managed to produce (particularly under slightly restricted time limits and with no external help), employing a hex map as both a filter and a nationwide “graph” of where cases were the highest and lowest.

Having presented my work and made some minor tweaks around colours/padding, I was finally satisfied with my viz, and felt it showed some definite progress in my work with Tableau.

Part of my dashboard, including filtered BANs with sparklines and a histogram, plus a heatmap for every state