As soon as I got my hands on Tableau v6, the first thing that came to mind was Gapminder. Those new shiny features meant it would be possible to recreate the amazing Trendalyzer software in Tableau. Here’s how I did:
To do the viz, I took advantage of lots of the new features, many of which I have already blogged about.
There are 9 connections required to show the viz above:
Okay, I acknowledge that that is not a shade on Gapminder’s 250+ datasets. But, hey, I’m not Google, you know! Tableau coped well with 9 datasets. What was difficult was reshaping the Gapminder data in Excel in order to get the best out of it in Tableau. Don’t know about Reshaping? Then you need the Tableau Reshaper add-in…
There are lots of parameters in this viz. Mostly, I used them to enable the user to choose which Dimensions/Measures to display on the viz. You can read about how to do that in my earlier Joy of 6: user-built views post.
At last, the Page Filter grew up. A valuable but neglected feature in previous versions, it’s been beefed up enormously this time around. The viz takes full advantage of Page filters and the ability to Show History. Click on a mark in the viz to see how it’s changed.
The single, pretty major issue is that on a Server or Public view, you don’t get a Play button. Given the architecture of the Tableau Server, this makes perfect sense: it is simply not possible to get data from the server to the client quick enough to work. You need to download the workbook to see the playback controls. This is a shame, as it’s a cool feature.
My favourite “easy-win” from the new features is the ability to use tabbed views to create an About Box (read the blog post about this). In this viz, it allows me to create links to the source and other places, and add context to the viz. Without the About Box, I would either have to miss those out, or use up valuable screen real estate on the viz itself.
This project was not without challenges…
Gapminder data interpolates
On the Gapminder site, if a data point is missing, it interpolates the value. That makes for a nice smooth animation on their website. Unfortunately, Tableau cannot invent data in that way. See those lines through the circles in the image above? That’s where Gapminder interpolated in order to make a smooth interpolation.
Log and linear
Gapminder automatically switches between log and lin axis scales. This is really nice, as it removes the burden from the user, and makes everything fit properly. Alas, no such feature in Tableau as yet. I wouldn’t even expect Tableau to implement this – it’s a very niche feature.
Annotations: two problems.
See the light-grey label showing Year that appears on the chart itself? In Gapminder, that label fits the whole chart. in Tableau, area annotations are at the front, so if you try to recreate labels as large as they are in Gapminder, the obscure the data points. But that wasn’t the biggest issue. Oh no. The area annotation is specific to the Dimension being selected. Therefore, I needed to add a separate annotation for every single X-Y axis combination. That was pretty tedious…
Some Gapminder data is just plain wierd
I had to pick and choose the data I used for this viz. In some cases there were too many missing data points. And in others, the values just don’t seem credible. For example, if you choose the CO2 emissions Dimension, well, I just don’t believe those values.
This has been a pretty challenging viz to get right. Having a defined end-goal (Gapminder) to try and hit isn’t the normal way of developing in Tableau. It’s better to explore the data, find the story, and tidy things up when you get to wherever you’ve got to. Forcing Tableau to be like something else is trickier.
Tableau isn’t optimized as well as Gapminder. Tableau can’t match Gapminder for smoothness of animation or flexibility of showing the different data. But that’s not really what Tableau is trying to do. It does everything it needs to do in this project just fine. Most of your business applications will be simpler than this. I think the end result above is great. It was a great way to explore the new features.
December 4th was International Open Data Hackathon. Groups around the world got together to see what they could do with open data: scrape, viz, reimagine, play, tinker. Anything you could imagine, using any kind of open data.
Around 30 people gathered at the offices of White October for the Oxford event. I went along with my Data Viz hat on, armed with laptop and Tableau v6. A summary of all the projects we worked on can be found on Tim Davies’ blog. I’m going to focus on the work I did. Zarino Zappia (@zarino) and I first tried to do something with the Mozilla browser usage competition, but the datasets were just too big. Sure, Tableau can handle it, but our laptops didn’t!
Hacking open data is great fun. My view blended the arts council data with a full list of constituency names and latitude/longitudes that we found on Google Fusion.
2. Open data is frustrating
There’s so few standards, and often so many hoops to jump through before you can get going.
3. Seriously, Tableau 6 is bloody amazing
I know I’m a Tableau lover, but, really, it was a perfect tool for this. I had most of the viz up and running within a couple of hours. Blending the funding data with constituency data was instant. And just about everything I did was with the mouse. Some of the other groups created some great stuff, using code libraries and their excellent programming skills, but Tableau just lets you explore, shows your results instantly. There’s no script to code, no alt-tab-refresh to test every CSS tweak you make, and no wizard-type work where you change 5 parameters before seeing a result.
4. This viz could be just the start
Now we have the constituency locations, it would be trivial to start blending these results with any other metric (population, employment, etc) to start to gain some huge insights.
At the recent UK Tableau User Group, Mel Stephenson made a great set of points:
Screen real estate is scarce.
Filters take up lots of space
Filters don’t show any information.
How about making worksheets that perform two functions: work as a filter and convery information?
Get that right, and you have just increased the data-ink ratio significantly.
Here’s an example. Recently, Tableau posted a great viz showing the Social Media Race and Twitter/Facebook follower counts of leading Media players. Here’s a snapshot of that viz:
But check out that Filter – it takes up, what, 15% of the whole space? Why not replace the Filter with a worksheet that shows the number of followers each person has, and add an Action Filter to it. The new version of the viz is as follows:
What I like about the new version is that you don’t need to click anything to see who has the most combined followers (Lady Gaga). The only disadvantage is that to select multiple media players, you need to use a lassoo or a Ctrl+Click action with the Mouse.
First off, let me say that I do not take credit for the contents of this post. Credit is due to Joe Mako and Richard Leeke who responded to my questions on the Tableau Forum. This is a good time to say that if you aren’t using the forum yet, get over there to ask or answer all manner of Tableau questions from the trivial to the fiendish.
I am posting this solution because it is such a sweet use of calculated fields. Here’s the scenario: “I want the user to be able to see the top n customers’ average sales. They should be able to control the “n” value, and show or hide all the rest of the customers.” Here’s the answer:
I’m afraid I am not going to go through the solution. Why? Well, I barely understand it myself, and I don’t think I’d be able to explain it properly. I recommend you download the workbook and look at the forum thread to work it out. Otherwise, just be in awe at the work of Joe and Richard!
Flowing data posted nice and clear instructions on how to make bubble charts using R. It’s a great post, and I thought it would be a good thing to show how to create the same chart in Tableau. Readers can compare and contrast the ease of each. Let me know your thoughts in the comments below. Here’s the end result using Tableau:
You only need five Dimensions/Measures on the worksheet to make the view look the same as the one on Flowing Data. I added Colour to State in order to create the highlighting option in the dashboard above. Is this cheating – it’s 6 drag and drops – is that 6 separate steps? 🙂
Step 4: Clean up!
I prettified it a bit by changing the Mark borders and putting it into a well-proportioned dashboard.
The whole process took me 15 minutes. Once your up to speed with Tableau, you can get to work at lightning speeds.
Flowing Data’s post started off by mentioning Hans Rosling and the Gapminder project. Watch this space for more on Gapminder and Tableau….
This is the third post on simple tricks that are available in version 6. This one represents a common request we get. The client says “I want to compare the performance of x against everyone else.” What is x? Well, it could be Customers, Departments, Countries, anything.
The drop-down filter is a parameter control, in this example showing every State in the Coffee Chain connection. The Tableau devs have made it easy to create a parameter containing every member of a Dimension. Right-click on the State Dimension, and choose Create Parameter…:
That’s your parameter list created – show the parameter control on your worksheet by right-clicking on the parameter and choosing Show parameter control.
Next up we create a calculated field to return either the State selected in the parameter, or all the others. This is essentially a dynamic Group, with one lone member of the Dimension, and everyone lumped together in the “Other” category. In this case, you can call the “Other” category anything you want. Here’s the calculated field:
This is the field you drop onto your worksheet. Whichever shelf it is on, it will only show two members: the State selected in the Parameter, and everyone else lumped together.
This uses sheet tabs to create an “About this viz” tab. It is good practise to credit data sources and date your vizualisation. The About Box allows you to do this. I’ve used parameters to store the hyperlinks, and dropped them onto worksheets embedded into the “About this viz” dashboard.
Prior to v6, the choice was either to omit the credits, or squeeze them into your dashboard somehwere. No more!
One can always expect visually attractive stuff from David McCandless on his Information is Beautiful blog. Sometimes though, the visual displays potentially confuse viewers. Today’s great post on Wikipedia Banner Tests is an example. Representing the size of donations using a square is pretty, for sure. However, we tend to perceive the relationship of one area to another pretty badly. A linear representation is more effective:
Think back to David’s orginal squares view – would you really have guessed that the Jimmy Appeal raised fifteen times much cash? If so, you’re one of the few – most people would have said it was around 8-10 times larger. The linear view above makes that difference much more clear.
There’s also some interactivity on my view – you can change the measures being displayed in order to explore the data a little more.
Over on the forum, Hadbar asked a question about Error Bars. He asked the question back in the days of version 5, when it wasn’t straightforward to make a pleasant error bar. However, now we have the Joy of Six, we can use the new dual axis/multiple mark features to create some visually appealing error bars:
Tableau won’t calculate the error values for you, so you need those in your dataset. With this example, you need two measures:
Error – the magnitude of the error
Error lower – the lower limit of the error bar
Here’s how we build the viz. First make a regular bar chart with the relevant Dimension and Measure, as below:
To create the error bar, we build a Gantt chart on the second axis. It’s lower limit is the value of the Error lower measure, and it’s size (height) is the size of the error. Step one is to put the MIN(Error Lower) pill on the Rows shelf, and set it to use Dual Axis (click the picture to see details):
We’ve now got a dual axis graph, but there’s some way to go yet. The Error Bars need to use a different mark type, in this case a Gantt Bar. Right-click on the axis, and choose Mark Type…Gantt Bar:
Set the Size of the Gantt Bar to the size of the error. To do this, make sure the Marks shelf is showing the correct Mark. In this case, it’s MIN(Error Lower). This is a new feature in v6. In case you’ve not seen it, you can page through all the Marks:
Once you have found the correct mark, put the Error measure onto the Size shelf – that creates the Gantt bar with the height of the error. You’ll probably now notice that you have error bars, but they’re wide, and not in the right position relative to the bar. This fat bar problem is easily fixed – just slide the Size slider on the Mark shelf to be the smallest possible. That creates a nice thin line.
How to get the error bars to align correctly? Simple – right-click on the axis and choose Synchronise Axis. While you’re there, you don’t need to show the right-hand axis at all. You can’t properly hide axes in Tableau, but you can format them so they don’t appear. Right-click on the axis and choose Edit axis. Delete the title on the General tab, and in the Tick Marks tab, set the Minor and Major Tick Marks to None. Click OK and your axis has “disappeared”.
Lo and behold – you have some error bars. Play around with widths/colours and other settings to make it appear just how you want it. Coming next in the Joy Of Six series…. using parameters to create dynamic groups.