Tableau: Picturing Data on a Virtual Canvas

Recently I have been playing with the Personal edition of Tableau Software, and decided to add it to my suite of analytical tools.  Tableau, which sounds like [TAB – low],  is the French word for “picture” or “drawing” and the Tableau software does just that: it provides an easy way to explore data visually and suggests best practices for displaying the data at hand. For us data geeks out there, it’s also noteworthy that the phrase “tableau de bord” means dashboard in French.

On June 2nd, Tableau released Version 5.0 of its desktop and server software products, which include numerous new features to better draw out important information, to explore large data sets and to analyze data simultaneously across multiple sources.  This new release is just one more example that visual analytics no longer is the new frontier, it is the new *reality*.
(See: Visual Analytics: Breaching the New Frontier in Business Analysis)

As with all powerful analysis tools, using Tableau from scratch can be daunting for the uninitiated.  However, it provides so many ways of guiding users to success, that a novice can quickly gain proficiency without first attending days worth of user training classes.  To stay with the metaphor: Tableau software provides the equivalent of paint by numbers. There are some rules to follow and it’s possible to slip up and draw outside the lines, but the guidance is there if one chooses to use it.

To serve business leaders who prefer to see “just the facts” and let someone else go data diving, Tableau provides powerful authoring tools that allow an analyst to guide the reader to the important insights while still giving the reader the ability to explore data from different angles. Depending on the need, results can be packaged as stand-alone presentations for ad-hoc kinds of analysis or be connected to data stores for continuous updates and performance tracking.

Here are some of my Tableau favorites

(not all of them new and in no particular order)

  • explore outliers simply by highlighting their data points
  • the summary panel, which aggregates selected data points on the fly
  • highlight the same data points across multiple, linked graphs
  • highlight important data with color while graying out the rest
  • the many self-documenting features
  • the automatic re-sizing of graphs depending on what else is on the screen
  • the worksheet metaphor
  • cloning worksheets
  • customize user prompts for filters and other data controls
  • hide confidential details by distributing aggregate data only
  • on the fly grouping and data aggregation

By no means is this an exhaustive list, but these are just some of the ways in which Tableau makes data exploration more effective and easier to accomplish.

Some Tableau growing pains

Before this post sounds like a PR piece, let me mention a few minor points that worked out less well.

The Personal Edition supports database connections to text files, Excel and MS Access.  Just to see what would happen, I used a very large table in MS Access and found that Tableau was slower than using Excel Pivot tables against the same table.  Not many desktop users would attempt using a million row table in Access and MS Access is no speed demon either when it comes to serving up data for Tableau.  Using a much smaller table with less than 100,000 rows worked like a charm.

Working with large tables can be made more efficient with a feature that allows the user to suspend table queries until all the query criteria have been set up.  Good if you already know how to structure the graph, less useful if you’re trying to explore the underlying data.

For some types of analysis I like to color code states based on my own data but I could not find a way to do so.  The Tableau folks are very concerned about proper data visualization techniques and thus worry about coloring states based on data, lest the geographic size of a state distorts the viewer’s interpretation of the data.  On the other hand, they do provide the ability to color states based on census data that is preloaded and ready to use.  Maybe there is a Jedi technique for shading states based on my own data, but so far I have not found it.

All in all, though, I have been very impressed with the rich feature set that Tableau provides and have started using it for data exploration.

Additional Reading

Thank you to Michael Cristiani (twitter.com/mcristia) and Daniel Murray   (twitter.com/DGM885) for recommending the following:

“Freakalytics.com- Home” http://bit.ly/W8KwE

Blog: “#Tableau 5.0 Active Dashboards and Enterprise Scalability” http://is.gd/MEyr and http://is.gd/MEBc

Server videos on the Tableau web site: http://www.tableausoftware.com/learning/training/ondemand

Tableau 5.0 Whitepaper -> Stolte, Chris. “Enhanced Visual Analysis by Linking Multiple Views of Data”: http://is.gd/MEt0

“Seems to be PRish, but some real good info about the Server 5.0 there: http://is.gd/PB8x”

6 comments to Tableau: Picturing Data on a Virtual Canvas

  • Michael:

    as promised, just a quick follow-up: extracting the data as you suggested did speed things up considerably. Once extracted, filtering sped things up even more. Thanks for the tip.

    Christine.

  • Christine, glad you like the new version. We’re always working on performance and it would be great to discuss in person. I second Dan’s request for you to come to the Customer Conference if you can– I’d love to meet up there.

  • You are welcome, Dan. I saw a few DMs from you, so I’ll switch to Twitter shortly and get back to you there 🙂

  • Christine;

    Thanks for the mention. You have a very nice blog which I just subscribed to. I also agree with Mike’s comments regarding the MS Jet Engine…it’s a bit clunky. Will you be attending the Tableau User Conference in July? If so, I’d love to meet you and share some data visualization stories.

  • Michael:

    thank you for the tip. I will try it and report back.

    Christine.

  • Christine,

    Thanks for the review here.

    The slowness with local files (Excel, Access, text) is most likely the responsibility of the MS Jet Engine, not Tableau. One thing you can do with local files to speed things up is Extract the data to using, either as part of the connection process or after the initial load of the data (Data | Extract).

    MANY BLESSINGS!
    Michael W Cristiani.