Using Tableau for Data Science

What is Tableau?

Perhaps you’ve heard a growing buzz in data science circles about Tableau. The Seattle-based software firm—which, according to Fortune, “pioneered the concept of ‘visual analytics'”—got its start at Stanford University in 2003 and launched an initial public offering in 2013.

Tableau’s mission is to make spreadsheets, databases, and other information sources simpler for the average person to use. Co-founders Christian Chabot (CEO), Pat Hanrahan (Chief Scientist), and Chris Stolte (Chief Development Officer) began by combining a structured query language for databases with a descriptive language for rendering graphics. The result? A database visualization language called Visual Query Language or VizQL. VizQL underlies Tableau’s eponymous software application, which queries relational databases, cubes, cloud databases, and spreadsheets, and then generates a variety of graph types. These graphs can be combined into dashboards and shared over a computer network or the internet.

Today Tableau boasts more than 50,000 customer accounts, including the World Bank, Coca-Cola, Exxon Mobil, Homeland Security, Pfizer, Fannie Mae, Gallup, Nike, and Adobe. Allrecipes executives are using Tableau to develop a 360° view of consumers. More than 600 employees at Cornell University perform analysis with Tableau: They manage contributor relations, visualize faculty salary statistics, and track which students are in which classes. Tableau helps Zillow provide self-serve business intelligence that empowers business owners to answer their own questions.

In 2017, for the fifth year in a row, the Gartner annual report on business intelligence and analytics placed Tableau in the “Leader” square of its Magic Quadrant, which shows the relative position of each competitor in the business analytics space. Tableau led the field in ability to execute.

Tableau currently offers five main products:

• Tableau Desktop — “fast analytics for everyone” — allows you to explore and visualize data in minutes, connecting to data and performing queries without ever writing a line of code.
• Tableau Server — “collaboration for any organization” — allows users throughout an organization to access live interactive dashboards in a web browser or on a mobile device.
• Tableau Online — “business intelligence in the cloud” — is a hosted version of Tableau Server that requires no set-up.
• Tableau Reader — “view Tableau Desktop files” — allows anyone to publish interactive data online.
• Tableau Public — “for public data” — lets you read files saved in Tableau Desktop.

Tableau Public and Tableau Reader are free, and free trials are available for the other products.

Tableau for the Data Scientist?

Given that Tableau aims to “make analytics easy for analysts, executives, IT, everyone,” data scientists may view the software as, at best, beneath them and, at worst, a threat to their livelihood.

Second point first. Popular and powerful as it is, Tableau is but one of many, many pieces of software used to glean insights from data. Note that in Lukas Biewald’s depiction of the data science ecosystem, Tableau is one of six tools in one of four subcategories in one of two partitions of one of three columns. So even if everyone from the corner office to the reception desk is proficient in Tableau soon, plenty of programs will remain under the near exclusive purview of the data scientist.

So perhaps the aspiring data scientist should just leave Tableau to the amateurs? Think again. Tableau Senior Software Engineer Robert Morton argues that his company’s software complements such data science standbys as R and Python. While Tableau is not the best tool for cleaning or reshaping data, and its relational model doesn’t allow for procedural computations or offline algorithms, it does excel at data exploration and interactive analysis.

“While both R and Python can be used to generate visualizations,” Morton writes on Quora, you must have a specific idea of what question you wish to ask of your data and how you wish to format the answer.”

With Tableau, Morton explains, you can quickly re-formulate your question with each new answer that emerges from a visualization.

Learn Tableau

Sold on learning Tableau, or at least checking it out? You could be up to speed in as little as a few weeks thanks to Tableau’s three training options:

• On-Demand Training: Learn online and at your own pace with free recorded sessions that cover everything from connecting to data and creating views to administering Tableau Server and building funnel charts.
• Live Online Training: Scheduled to cover a specific topic at a set time, real-time online learning sessions promise to make you a “Tableau pro.” You have to register for these instructor-led sessions, but they’re free.
• Classroom Training: Tableau offers instructor-led classroom training across the globe, on-site or in a virtual classroom.

If you’re the bookish type, you might consider Tableau Your Data!, which goes beyond the basics of Tableau Desktop’s interface and explains best practices for creating effective visualizations for specific business intelligence objectives.

Finally, for a running list of all of Tableau’s irksome quirks, surf over to Chris Gerrard’s blog Tableau Friction.