Tableau Access for CS765

in Posts

Tableau is a commercial data analysis/visualization system. It is very popular and common (it is what most of the Unviersity data platforms use).

We will provide you with access to Tableau for class, and recommend it as an option for learning about grammar-based visualization description.

There are three reasons why we recommended Tableau for class (pre-2026):

  1. It is an excellent way to see how the concepts of visualization grammars are applied in practice.
  2. It is an excellent tool for exploratory visualization: once you have a little familiarity with it, you can very quickly make different visualizations even from large data sets. This ability to rapidly explore both design and data provides a great way to learn about both: experimenting with different views is a great way to learn about a data set. And how to choose visualizations.
  3. Tableau makes very good visualizations. It has an amazingly fast data engine. And it makes charts that look good, have sensible defaults, …

In the past, a tool like Tableau was the only way to do #2: you simply can’t program fast enough to explore both design and data. And doing #3 was hard: it required substantial implementation effort.

But, that was the past… Now with an agentic programming tool, you can have your favorite AI put together programs that generate visualizations quickly, and it knows all those data handling tricks and visualization features to make nice, performant visualizations.

So, Tableau is less of an obvious win. Especially because Tableau is a big, complex system. While it is designed to be “user friendly”, it is still a powerful tool. It takes some time to learn. Our experience is that students in the class can usually learn enough about how to use it quickly to do the assignments that we ask. But it can be frustrating at first.

So this year, my recommendation is more tepid. I still think Tableau is a great way to learn about visualization, and a great tool for doing the assignments. However, I am less convinced it is worth dealing with its steep learning curve, because you can get its advantages other ways. The benefits (Tableau is a great way to learn visualization) balance the costs (its hard to learn) - in the past, the ability to do exploration tipped the scales.

Rather than providing training in class, we point you towards the many great resources on learning Tableau on the web. I do provide two tutorials to get started with the data sets for class: Tableau Tutorial for CS765: Getting Started with Census Data and Tutorial Tableau Walkthrough 2: Life Expectancy .

For 2026, you must experiment with some grammar-based system (but you can use Vega-Lite or Altair instead of Tableau).

The course staff are not Tableau experts.

How to access Tableau

In order of preference…

  1. Tableau makes the software free for students, so you can also your own student license. This has the advantage that it doesn’t expire when class ends. Info on the Tableau for Students webpage.
  2. Tableau provides “course licenses.” You can download the software and use the key for the class. Instructions are on Canvas: Tableau License Information (Canvas).
  3. You can use Tableau Public Online - the free, online version that works in the web browser. This is a limited version, but it should be sufficient for class. Go to Tableau Public, create an account (or sign in), and try “creating”. More about Tableau public at Getting Started With Tableau below.

Why Tableau?

You might be (justifyably) wondering… why do we recommend people use Tableau? Normally, I don’t like to promote any particular tool, particularly not an expensive commercial one.

Some reasons (in no particular order):

  1. Tableau is actually an example of how the theory we discuss in class (thinking in terms of data types and encodings) manifests itself in practical ways. Note how Tableau doesn’t give zillions of chart types: it lets you directly control encodings.

  2. It emphasizes the choice in building vs. using tools. As Computer Scientists, we often prefer to write programs - which has upsides and downsides. I really want students to experience good tools so they can better understand this tradeoff.

  3. Tableau actually is quite good at what it does. While it definitely is a “power user tool” it represents a really well-thought out approach for helping people do visual analysis of data. Even novices can explore different design choices to explore data.

  4. Under the hood, it is pretty amazing (from a systems perspective). It handles very large datasets without effort. It can often produce charts faster than my python scripts can merely read the data.

  5. It actually produces good charts (in terms of the technical “art”). It includes a lot of design experience, best practices, and research in how it makes choices. It has numerous ways to encourage you to “do the right thing.” This frees you from the details (what font do I pick? what spacing for a bar chart?) and lets you focus on finding good encodings for the data.

  6. It embodies a ton of research. Many pieces of it were the subjects of research papers. Indeed, the whole system itself came from a research project (which spawned a number of very influential papers).

  7. The Professor who ran the original research project and was one of the founders of the company (Pat Hanrahan) is a Wisconsin alum and Turing Award winner. He gave a distinguished lecture here in 2022 (about other work).

The biggest downside is that it is truly a power-user tool. It is designed so that experienced users can do amazing things, not necessarily to make it easy for a first time novice. That said, I find it to be an impressive balance of not being too inaccessible, but also able to have lots of things hidden. Once you get used to a few basic concepts, you can do a lot very quickly.

Getting Started with Tableau

After either installing the software or going to the online site…

You can create a Workbook and connect to an Excel or Text (CSV) file that you upload from your computer. The first row in the file should contain the column headers. Then you can simply drag the fields from the Data pane on the left (representing columns in the Excel or Text file) out to “Columns”, “Rows”, or “Color”, “Size”, “Text”, etc, under “Marks”. These fields are broken up into dimensions and measures. Dimensions are categorical fields (usually discrete), and measures are metrics or numbers you want to analyze (often continuous). You can change the method of aggregation of the measures to “Sum”, “Average”, “Median”, etc, by using the dropdown menu of the fields. In case you know the data you want to look at, but don’t know the best way to view the data, you can select different dimensions and measures while holding down the control key and click on the “Show Me” button on the right to display a list of common chart types that can help you start your analysis. For more detailed instructions and sample datasets, you could watch the “Getting Started” video from the official Tableau Free Training Videos website. You may also find YouTube videos such Edureka, freeCodeCamp.org, or Penguin Analytics useful as additional examples.

The current recommendations to give students (from Tableau):

The way I like to learn these things is by diving in and fiddling around with it. Some things to do (in rough order)…

  1. Make sure you can create a new workbook (use the “create” button )
  2. Connect to Data (initially, you might want to upload a file, in the future, you may want to connect to one of the data sets we provide).
  3. The Key to Tableau is making sure it understands the type of your data. If Tableau understands your data, it will make it easy to create visualizations that show it.
  4. Tableau has the concept of “dimensions” and “measures” - which are a little more complicated than just “categorical vs. numeric” variables. Understanding this concept is important. If a variable is the wrong type, Tableau won’t let you do things. Tableau has “green pills” (measures) and “blue pills” (dimensions) that you can drag into visualizations.
  5. Creating appropriate dimensions and measures from your data is important. For some things, this requires right clicking on a variable and either “convert to measure” (to change its type), “change data type” (if it is misinterpreting numbers as strings), or even computing fields.
  6. Often you need to explicitly tell Tableau how to interpret your data correctly, for example, taking a set of discrete choices and aggregating them by counting (or a set of continuous values and aggregating them by counting buckets).
  7. You can use different variables by dropping them different places on the workspace. Rows and columns are an obvious place, but you can drop variables onto the icons for color, size, etc. This lets you put mutliple variables into a visualization. For any of the “pills” in the visualization (for example, in the columns, rows, color, …) you can change its settings by right clicking.

I find that playing with things usually reveals many of the things I want to do.

There are two “walkthroughs” showing my initial attempts to work with the 2026 data sets (see A Tale of Two Data Sets...) at Tableau Tutorial for CS765: Getting Started with Census Data and Tutorial Tableau Walkthrough 2: Life Expectancy .

Some quirks with Tableau Public

Tableau public is free - you just need to create an account. As I understand it, the main “limitation” is that everything you make is public (which is OK for class). It isn’t ideal, but we are having a problem with the “class site” for Tableau online.

The actual editor and visualizations seem to be the same as the regular online tool (from my quick, non-expert experiments). However, there are a few quirks…

  • When you first publish (save) the workbook, check “show sheets as tabs”. Otherwise, when you view (or share a view) of your workbook, the viewer will only see one of the worksheets (and I am not sure how it picks which one).
  • When you log in to Tableau public, it wasn’t obvious to me how to get back to something I was working on. Click on your “profile” (the person in the upper right corner) and pick my profile. As best I can see, you can click on it to open it in “view mode” then click on the edit icon (top toolbar).
  • If you want to share something, use the share icon (in either the top or bottom toolbar of the viewer - I don’t see how to share from the editor).

It should be good enough to do the assignments in class. I was able to very quickly recreate Tableau Tutorial for CS765: Getting Started with Census Data. You can see it here.

Resources for Learning About Tableau

Everyone has different tastes in how they like to learn these kinds of things. If you find something useful - a website, a video, etc, even a tip you on what you figured out, please share it with your classmates!

I’ve made a Canvas Discussion Board for discussions about Tableau.

Official Tableau Documents

Tableau recommends the Data Analytics for University Students guide. Although, that seems to just be a way to get to Getting Started as a Creator. This page discusses the desktop, but online is very similar.

Some extra notes for class…

In no particular order…

  1. Our goals include getting you to show us your visualizations. We are not testing your mastery of the tools.

  2. Many (including me) have had a hard time exporting from Tableau. The export menu is hidden (it’s under the worksheet menu… Worksheet > Export > Image). SVG is usually the right choice. I think there is a separate option to export an image of a Dashboard (if you make one of those).

    • If you are stumped on how to make Tableau export, it is OK to take a screenshot. You might need to crop out the menus and other UX elements (unless you intentionally want to show them - as I often do in the tutorials).

  3. If you can’t get the tool you are trying to use to make the exact picture you want, it is OK to edit it (unless we say otherwise). A really common paradigm is to take a picture from (for example) Tableau, and then load it into PowerPoint or Illustrator to add titles, captions, legends and callouts. Avoid changes where you might make unwanted distortions in the data (where your edit actually changes the encoded data).