Diving Into Tableau
First Impressions After One Week
I’ve long been dazzled by the crisp colors and interactive displays of advanced data visualization. It wasn’t the first thing that drew me to data science, but it wasn’t long into my programming journey that I was envious of those with the skill to create truly great visualizations. Publications like Informationisbeautiful and the NY Times keep pushing the boundaries of visual communication with their stunning graphics and displays. I’ve always been struck by how a well-crafted visualization can communicate big ideas with clarity and grace. So when I learned that we were using Tableau in my Data Analytics class, I was more than excited — especially when I discovered that I could had access to a year-long student license to use the software. I’ve been interested in Tableau for a while but until now, the price tag kept me away. When I received access, I couldn’t wait to learn as much as possible.
After spending a week on Tableau, I can say that I’m impressed. I feel that Tableau lives up to the hype around it. It excels on both ends of the usability spectrum — it’s easy to prototype and build many versions of a graphic, and that product is generally easy to interpret by the end user. While I didn’t initially find the Measure Names and Measure Values to be intuitive, the drag and drop interface drastically simplifies the entire process. Implementing calculated fields and filters made it easy to visualize exactly what I wanted to. Overall, I found it to be a very accessible tool.
Below is one of the I created. I used the FBI’s Uniform Crime Report data for my hometown of Anchorage, publicly available on the municipality’s website. The dataset contains values for cleared and reported violent and property crime, from 1986 to 2019. If you want to read more about the analysis I did, it can be found here.
I had a lot of fun playing around in Tableau, but I did find it to have one some drawbacks. It felt incredibly limiting that creating visualizations was the only thing I could do in Tableau. While you can create calculated fields and Tableau Prep can be used for data manipulation, I didn’t like that I could do it all in the same place. Compared to the freedom of a notebook style interface, I felt that the number of things I could do was confined to a small set of choices. I felt limited.
I think this is the tradeoff with using a tool like Tableau. It’s very intuitive, has a low bar to entry, and makes great visualizations. It’s useful for visually-guided analysis and for telling stories with data. It allows for easy communication from the data to its audience. However, its strengths end there. By comparison, a general purpose language like Python enables better, well-rounded analysis. In Python I can graph all quantitative data with a single line of code, quickly get descriptive statistics, and handle missing values. Whereas Tableau is a great tool to create visualizations once I know what I need to communicate, using a programming language tells me where to look to discover patterns worth communicating.
At the same time, though, I’ve generally found visualizations made with matplotlib and seaborn a bit lacking. They are certainly useful — I can use them to create graphics for data analysis and they help demonstrate patterns in the data. However, they don’t have the same expressive power harnessed by interactive visualization packages. Due to my preference for using a programming language and wanting to get more out of my visualizations, I’m more interested in learning Dash, Bokeh, or Shiny. I believe that these packages will allow me to create the interactive, beautiful visualizations I want without feeling restricted when performing data analysis.
I think Tableau is an excellent product. For those without a coding background, it provides a way to create visualizations for data analysis that is extremely intuitive. Compared to either programming or other software like Excel, the bar to entry is set extremely low. And even coming from a coding background, I find that I get better visualizations in Tableau than with Seaborn for the amount of time spent. However, I feel I can do much better analysis with code and don’t mind the extra effort to learn another visualization package. For that reason, I see packages like Dash and Shiny as offering the best of both worlds.
What are your thoughts? Do you prefer Tableau, Shiny, or a separate package? Let me know in the comments below!
