Tableau Twb Dimension Names Being Read in Twice

I am beyond excited to introduce the kickoff (and astonishing) "how to" blog post from DataBlick Partner Noah Salvaterra on the DataBlick site! In addition to working full fourth dimension with clients, Noah has been instrumental in creating some of our unique service offerings, such as the Tableau Personal Training Programme, and a Virtual Middle of Excellence Program that we hope to announce by TCC15. If that wasn't enough he likewise recently lured both Joe Mako and Jonathan Drummey to the DataBlick squad.

Noah has been called "ridiculous", for his ability to make merely about annihilation happen in Tableau, every bit readers of Jonathan Drummey's Drawing with Numbers have already seen. This post is no exception. I knew this was cooking, and my jaw still dropped.

Thank yous Noah!

-Anya

Tweet to @noahsalvaterra

Can I make this in Tableau?… You tin now.

Chord Diagrams are a chart that is often used for visualizing network flows that represent migration within a organisation, the one on the left for instance, shows changes in cell telephone selection for a collection of users. I'll introduce the nautical chart blazon a scrap more than carefully later, and provide a reference, but by the terminate of this mail my hope is that you'll have a sense of how you might build this blazon of chart with your own data in Tableau.

Joe Mako suggested the possibility that he will visualize the data from this post in a more effective way. I call back that would be a great follow upwardly and I certainly welcome that sort of feedback. I'm fine with the possibility that something improve could plow up. The information underlying the chord diagrams in this postal service is included at the end. So feel gratuitous to have a crevice at it. My opinion may non exist that disarming on this topic, but the alternatives I've seen so far, such as stacked bars, rut maps or in the case of geographic data just throwing it all on a map and hoping it makes sense accept their own problems equally well. What I think most folks tin can concur on is that Chord Diagrams aren't the right choice for every visualization. Searching the internet, I constitute more bad than good, and then I chose the examples here with some care. My hope is that past enabling chord diagrams in Tableau, we can allow some experimentation and mayhap establish some guidelines to know when information technology might exist a good selection, even if that answer turns out to exist never.

Can I make this in Tableau? When it comes to chord diagrams I've seen a lot of responses that make me cringe. Chord Diagrams aren't built into Tableau, and I wouldn't expect them to be added whatsoever time soon, but lines and polygons are sufficient to make just about annihilation you'd like. Seriously… I'd like a Tesla, and here is the workbook. If it isn't loaded yet, consider that I'k building a car in a data visualization tool, and quit reading so fast!

This workbook raises an interesting question in terms of best practices. The data underlying this workbook was parsed from a cad file. What data visualization best tells the story of this data? Should I go with a stacked bar chart or a scatterplot? BTW, if you've got your blush-depth glasses handy, you're going to want to grab them and download this workbook. I'g not going to go deeper into the techniques backside this workbook, it there is demand I could certainly exercise a followup to my original 3d mail service that that appeared on drawing with numbers. I'm hoping this workbook will make an appearance at Allan Walker and Anya A'Hearn's Minority written report talk at TCC15, somebody please salve me a seat.

At that place is very lilliputian math in the Tesla workbook (OK, bated from the rotation matrices), only for the chord diagram I burned through a couple #2 pencils before I even began to build the workbooks in this mail. Figuring out the underlying math isn't for everyone, but I never doubted the possibility of creating this chart blazon in Tableau. If y'all merely want a picture of a chord diagram, you're probably better off getting information technology from Circos. Your workbook will certainly be a lot faster. A chord diagram may be too complex to take in all at once. Images of chords have graced the encompass of more than a couple scientific journals and magazines, merely such charts don't unremarkably practise much for me, they may be beautiful, if that sort of thing does it for you, but they aren't terribly illuminating. Interactivity is essential to this presentation, so if you lot work at one of those companies who pastes tableau screen captures into powerpoint, this probably isn't the correct choice for you. In do, I'd be probable to pair this chart with one or ii simpler charts that back up the presentation and make full in some of the areas where the chord itself might be weak, for example a bar nautical chart could evidence the aggregate picture, which should clarity any defoliation from the donut and to aid with interactivity. As I sketched out the requirements for this project in my caput, interactivity was at the top of the listing. Chords and arcs should answer to hover and pick equally they exercise in the D3 versions of these charts.

Another requirement, or peradventure merely a wish, came out of a chat with John Edward Abdo, Founder and CEO of Palette software at last years Tableau conference. His idea was an app store for Tableau that could contain business specific visualizations, and then frequently used chart types could be at your fingertips in seconds even if they weren't included in the show-me carte du jour out of the box. I liked the thought of existence able to plug new visualizations types into the show me menu if the process could work seamlessly with Tableau. I got hung upwardly on for a while on the thought of seamlessness, essentially dismissing the idea. I've done enough of hacking in Tableau, merely I wouldn't even know where to start with hacking the show me menu. If any disgruntled developers desire to aid me out, call me. Brusk of that, Tableau would need to be involved in creating the infrastructure for such an exchange, and that assumes that opening prove me to developers is even feasible. Withal, the seed was planted and when I started thinking about chord diagrams, information technology started to grow. Creating a few chord diagrams in Tableau might exist interesting, but after some initial novelty it would be about as useful as a Tableau Enigma car. So I decided to call back a scrap on how I could create the workbook in a general enough way that the structure would exist reusable by others. I'd like to find a mode to generalize this farther. I tin't hope to exist available to support every issue that arises, that it will piece of work in every state of affairs, or with whatever dataset. Simply let me know if you discover an outcome or if you lot apply information technology to make something crawly. I look forward to hearing from folks that use this post or have ways to push it further. Currently the structure of this visualization requires the granularity of the information to exist at the level of cells in the transition frequency matrix. I think it could be pushed to piece of work with finer granularity datasets directly by using LOD calculations. I decided to go out that on the table for now, if anyone wants to take a crack at that, please feel free to practise information technology and allow me know how it goes.

Clear enough? Image from Bradley Voytek'southward original mail service on Uber.

I started with a couple well known chord diagrams, that provided a convenient reference during development. One of my favorites is an example that comes from Nadieh Bremer's blog Visual Cinnamon. It shows the device choices of a collection of a population that has who owned multiple devices over the course of the survey. As well, Nadieh's mail service includes an elegant introduction to the chart type and how it tin be used to tell a story. Below is a brief summary of her presentation, in many cases using or paraphrasing her words, this may be sufficient for those who are already familiar very familiar with chord diagrams, and too serves to points out some of the features in the Tableau version that may work a chip differently, for the definitive introduction to chord diagrams I volition defer to Nadieh's mail service. The second case I'll provide is based on Mike Bostock's chord diagram postal service showing Uber rides past neighborhood in San Francisco which seems to be the most frequently referenced chord diagram I could notice. Mike'south chart is a response to a post by Bradley Voytek that appeared on Uber'south site, including the map pictured above.

The boundary, is an exploding donut chart… some folks are going to accept trouble getting past that, but I'm non sure the verbal per centum comparison of these arcs is necessarily primal. I'thou non looking to distort data, merely when the complexity of the story makes avoiding distortion difficult or impossible, so it becomes a question of how to avoid distorting the story that is being highlighted. I included the volume in the arc as the tooltip (and included a bar chart in the interactive version below). For example, 38.04% of respondents currently own a Samsung (their previous telephone may take been annihilation on the nautical chart, including Samsung).

Individual chords prove migration between phone brands. This chord shown here represents respondents who currently own a Samsung and previously owned a Nokia equally well every bit those who currently own a Nokia currently and had a Samsung as their last phone. The arc on the Samsung side spans eight.75% of the circle, which means that eight.75% of respondents ain a Samsung currently, but had a Nokia earlier. one.two% ain a Nokia now, but previously had a Samsung, that means Samsung took gained a lot more than customers so Nokia. Since Samsung had a net gain over Nokia, the color of the chord is set to match the Samsung arc.

Some customers remain loyal to their current make. These are shown as hills on the chord chart, which tin can be though of as chords looping dorsum on themselves. nine.8% of the respondents use an Apple tree phone at present and also had an Apple previously.

The collection of chords higher up bear witness those who have an Apple tree phone at present, or who previously endemic an Apple phone. All of the chords are the color of the Apple arc, so Apple had a net gain from every other brand. Additionally, with the exception of Samsung, the chords are very narrow at the other finish from Apple. That ways Apple lost very few customers to these other brands.

Here is then interactive Tableau version of this chart:

The Swap:

Tableau isn't built for matrix calculations, matrices and vectors would exist powerful data types, but building this construction for an arbitrary nxn matrix as it stands is likely beyond my abilities, at the very least, as Joe would say, "the Juice isn't worth the squeeze." I chose to permit some minor pre-processing of the data to get it into a alpine format, that is, where each row contains data relevant to a single prison cell in the matrix. There is an app for that… well technically it is an excel plugin. I've included an excel workbook at the finish of this post which has some stride by step instructions on preparing your information to plug into the chord workbook. It is of import that the construction and relevant field names match with the samples, or there will exist some cleanup required after you plug them in. I've washed this cleanup ten-15 times myself by now and trust me; it is easier to get this right before you swap out the information. Annotation, swapping out the data isn't a technical term in Tableau, I am doing my all-time to avoid terms like replacing the data source here, since that is already a process within Tableau (and it isn't a feature that volition exist used). The process of replacing a information-source involves having two open up information sources inside tableau at the same time, essentially substituting all references to one with the other. In this case there is but e'er one data source. The modify happens when the information source properties, in particular the folder containing the underlying text files is contradistinct, so the visualization updates to a new underlying data rows, but the structure it uses for the visualization is only updated, not replaced.

Then say you lot decide you'd like to recreate Mike Bostok's Uber Chord Diagram in Tableau. That example is included beneath, so in that location is no demand to do and so, but going through the steps may aid to setup using a data set of your own. The excel workbook provides step past step instructions on the pre-processing, depending on the structure of your source data the procedure might exist a scrap different, merely the goal is to get a dataset with the right structure and names, that can exist swapped in for the underlying data in the Tableau Workbook. You can attach boosted information to rows or columns as desired before of subsequently reshaping the data if you want boosted data to be accessible in the view. The reshaped, renamed data should be saved twice in the same folder (I recommend a folder for each Chord Diagram) equally ChordData1.txt and ChordData2.txt, additionally, a couple pocket-sized text files will need to be in the folder, called 2_Copies.txt and 5_Copies.txt these can be exported from the attached excel spreadsheet or you can apply the samples attached at the stop of the post.

The prep work is pretty minimal. I don't run into much way around the reshaping, fifty-fifty if I was to first from scratch. I mentioned that the data may need to be aggregated to the level of a matrix, and six or then fields need to accept the verbal same names as in the original excel file, though you can always modify them back to any yous like subsequently you swap the data. If that is the bad news, here is the skilful, one time you've done that you're basically washed. Right click the data source and select edit information source, then click on the directory for the source data (the red link on the left hand side of the data edit screen) and change to the folder where you placed your files. Double click one of the files to complete the swap and return to the dashboard. That is it… the chords should update to the new information source and y'all're washed.

In case you missed that, here is a video of how to change the Uber chord workbook into the phone chord workbook in near a infinitesimal:

I promised the underlying data used to create the workbooks included in this post, then here is that, in addition to fueling alternative visualizations of this data this workbook includes detailed pace by step instructions on the pre-processing that took place:

ChordData

As well, here are the underlying twb, tde, and csv files used in this post:

ChordFiles

Despite the frequency of the "Can I make this in Tableau?" question on the Tableau forum, Chord diagrams haven't made an advent nonetheless in Tableau. I don't similar "Tableau tin't practise it", to exist the reason behind ones visualization choices. I've spoken with folks on both sides of the contend on this nautical chart blazon, some feeling that chords should never exist used and some who similar them for i use case or some other. I'm in that second camp, but I'm non trying to sell annihilation in terms of best practices. Chord Diagrams may nowadays some perceptual challenges, in fact I'1000 sure they practice. In spite of that, however, I take trouble throwing them out as a whole. Some of them are certainly bad, peradventure even most, but while these charts can exist complicated to wait at, they also stand for a very complex set up of relationships. In the case of migration information at least, I'm not sure there is a perfect choice. But I guess we will see what Joe comes upward with.

Don't forget to ship me your Tableau chord diagrams! I'thou going to stick them all on the fridge.

carltonhavock.blogspot.com

Source: https://www.datablick.com/blog/2015/08/27/diy-chord-diagrams-in-tableau-by-noah-salvaterra

Share this post