London Tableau User Group – May 2016
Li Kim Lee of Mezzo Labs attended the third London Tableau User Group of 2016. Here is a summary of the presentations at the event.
Reza Rokni, Google
Data @ Google
Reza Rokni explained how Google BigQuery is one of the newest database tools on the data scene today.
Based on the premise of “data centre as computer”, BigQuery can used as part of your ETL process, to separate data processing from data storage. It also enables data sharing with others, giving them actual access to data.
Reza performed a quick-and-dirty real-time SQL query to demonstrate the speed of the tool. His query of 4 TB in 100 billion rows was completed in just 31 seconds.
Reza then went on to talk about Stream Processing, the real-time discovery of insights via Dataflow, enabled by Google Cloud Dataflow, an ETL system. It is powered by MillWheel, an Apache-based tool. Stream Processing forms the basis for Machine Learning via Cloud ML.
Reza then demoed his speech being translated instantaneously into text via the web speech API. The TUG audience preferred the text rendered in a code “aesthetic” i.e. green text on black screen over a more user-friendly white screen.
Andy Cotgreave, Technical Evangelist Director, Tableau
MakeOverMonday started a year ago when data whiz, Andy Kriebel, posted charts ripe for improvement online, which Andy Cotgreave would then revamp via Tableau to keep his hand in.
Since then MakeOverMonday has grown and now has 35 people per week challenging themselves to improve the vizzes posted by Kriebel.
Almost a social data experiment, MakeOverMonday is now a great place for breaking apart data visualisations and experimenting. Cotgreave then encouraged the audience to try and have go themselves.
In the second part of his presentation, Andy asked the audience “how do you visualise time?”. The general consensus (via the beer-fuelled crowd) was “line chart”.
Andy stated the correct answer was “it depends”.
To prove his point, he looked back into the mists of time and showed various examples from the 15th century and onwards. We especially admired the 54-foot long chart by Jacques Barbeu-Dubourg from 1753, which is somewhat analogous to the scrolling function on a tablet.
Also notable was the first instance of time shown on the x-axis by Joseph Priestley in 1765, via what Andy dubbed a Gantt chart “invented before Gantt did”.
Andy went on to walk us through six examples of line chart alternatives in Tableau, including:
- Slope charts showing first and last points to show change
- “Cycle” plots, for example, putting smaller timescales before larger ones, where the reverse hierarchy shows seasonality trends
- Highlight tables with trends called out in colours AND numbers
- Small multiples, utilising the available space within cells of a highlight table.
The possibilities of depicting time could endless but having several such charts in your arsenal can aid data exploration.
Andy then ended with a lovely Joseph Priestley quote, all too familiar to us all data geeks: “Laborious and tedious as the compilation of this work has been… a variety of views were continually opening upon me during the execution of it.”
Andy Kirk, data visualisation specialist
Visualisation is a game of decisions
Andy Kirk asked what we consider evil in visualization. The audience shouted out: pie charts, Qlik, 3D.
Kirk stated evilest evil is “arbitrary”. We should aim to eliminate it. We need a framework for thinking – a workflow for data visualisation – and illustrated this by showing Liverpool’s season match by match.
He walked through workflow steps:
- Questions on context, formulate brief
- Framing of data + familiarisation, versions
- Exploration via Tableau – seeing – step back, emphasise editorial thinking
- Developing your design solution – representation + presentation of data
Andy ended by noting that
Good Visualisation = Trustworthiness + Accessibility + Elegance
…with Elegance being the hardest of all.