Tuesday, October 21, 2014

Mixed feelings about Tableau

I've been using Tableau (besides Adobe Illustrator) in my visualization/infographics classes this semester. After playing with it myself, I thought that it was worth a try. It's powerful and it certainly works well for analyzing large data sets but, after thinking a lot about it, I have some misgivings that I've shared on Twitter. I'm copying them below.

The main one is the comparison with Flash. I think that one of the main reasons why so many news organizations and professional designers aren't willing to adopt Tableau is the bad memories of the Flash debacle, a few years ago*. A proprietary tool that doesn't give you access to some sort of underlying code —Javascript or otherwise; something that isn't proprietary, in any case— after exporting your graphics won't be attractive to most people in my fields. If it becomes obsolete (and all technologies become obsolete, eventually,) your old work will become useless.

(*I am aware that the news and graphic design markets may not be interesting for Tableau, of course; Big money comes from other places!)

Anyway, here are the thoughts, plus links to some open source tools
(Correction: Gregor Aisch has pointed out that many of these tools aren't open source or even "free" in the strict sense of the word; Gregor has suggested his own Datawrapper as a tool that is. I should have included it in the mix.)

3 comments:

  1. Interesting comments Alberto. Let me dig a bit deeper on this one. First, the underlying format of any Tableau Visualization is simply XML. Opening up any TWB file in Notepad will show you the thousands of lines of code in an open format. When you reference exporting, I guess it depends on the purpose of exporting. If you mean to run stand-a-lone without Tableau then no, you can't export that code and load in HTML, but there is a Javascript API that will do this. Here is a tutorial on that functionality. http://tabsoft.co/1yVJCni

    The beauty of Tableau is the ease of use so that you don't have to code. I can create far more complex charts in R or D3 vs. what I can do in Tableau, but it requires a huge investment of time and coding. Tableau is working on better responsiveness and more flexibility with design. I imagine that future releases will get better and better in those areas. In the business world this rapid design is what makes it so powerful.

    You say it's too inflexible for news organizations. I guess that depends on who you are. The NY Times certainly has the talent and capabilities in-house to code on more advanced platforms (like D3), but Tableau has had tremendous growth all over the world with news organizations. Even our local Cincinnati Enquirer uses Tableau. Tableau also has a partnership with the IRE offering a free Tableau Desktop license to all members, over and above the free Tableau Public tool. That's a huge investment by them offering a very powerful tool free to the industry. Based on the viz of the day publications from news organizations all over the world it appears to be catching on quickly.

    As for teaching tools, we chose Tableau as our primary tool in our data visualization class because it's the quickest tool for people to get up to speed. We have masters students from analytics, information systems, business administration, economics, medical, marketing and even fine arts. It would be too much to try to teach the course material and a programming language together. By using Tableau it enables them to be able to create projects within a few weeks. And again, Tableau offers a one year license free to all full-time students as well as a license key for use to use during the class period for all students.

    So my two cents, I wouldn't simply abandoned this tool.

    ReplyDelete
  2. You raise some good points, Alberto. Thank you for your feedback.

    Tableau does let you export images and PDFs, and they’re nicely formatted for further editing in Illustrator, etc. That’s more the Dreamweaver-style use case you’re describing. Most visualization tools you mention require a hosted solution if you want them to be interactive and would face the same issue of obsolescence.

    Exporting to D3 is something that we’ve discussed, but we haven’t pursued this for several reasons. This isn’t just about the visuals, but also data access, aggregation, filters, etc. Making this work with the many data sources Tableau can talk to would be quite difficult. Additionally, the generated code might be quite inscrutable and of limited use. It isn’t a feature we support right now.

    In my experience, there is also a bigger question here about what we think people should to be able to do. I love D3 and I use it a lot. But I don’t think that it’s reasonable to expect everyone who wants to analyze data (particularly in in a newsroom) to know how to program and to understand the advanced concepts that make D3 so powerful. (That’s why you and I aren’t writing SVG or PDF in a text editor, even though we could. That’s what tools are for!)

    Data analysis solutions also do a lot more than just create the graphics: They help you look at many different aspects of the data quickly by making things like data access, aggregation, and filtering easy and fast. In D3, your data needs to be in the right shape and at the right level of detail, and you have to know what you want to build ahead of time. Changing the chart type means rewriting a bunch of code. That's the difference between building a visualization and doing data analysis. I think we really want to focus on the latter, with the former being a means to an end.

    Your points are well taken, and we’ll definitely be thinking about them for our future planning. We already do a lot to support data journalists, and we’re always open to new ideas to do that.

    ReplyDelete
  3. Thanks both for the comments. Jeffrey, I am using Tableau myself. I was just expressing a series of predictions, based on what I've observed so far in class. Tableau is really good for analysis, but somewhat lacking as a design tool. I am not going to drop it so soon, at least for exploring data.

    ReplyDelete