Where is the joy?

Where is the joy?

I haven’t gone all contemplative on you for this post, nor am I channelling my inner Black-Eyed Pea, but instead this is a post about the recent phenomenon of the Joy Plot. And it’s fair to say it’s resulted in a bit of a division of opinion. Those who know the derivation of the Joy Plot will see what I did there … those who have no idea what a joy plot is but have good recollection post-punk gothic type music from the late seventies might have some kind of confused association going on …

As far as I know, the term “joy plot” has only recently been derived. It owes its name to the iconic cover of an album by Joy Division from 1979 (below)


Robert Kosara has an excellent EagerEyes blog – and in his recent post here, he mentions that the term was coined earlier this month by Henrik Lindberg, with examples here


It’s hard to give an accurate, exact definition, so instead I’ll give a vague one here:

  • A succession of horizontally close-packed area charts over time, usually showing defined peaks
  • The lines are sufficiently close together that the peaks overlap/obscure the lines behind (giving a sort of 3D effect)
  • Looks a bit like the cover of “Unknown Pleasures”

It turns out that the original artwork was based on radio / magnetic waves from a pulsar star. But I think the recent popularity of joy plots (edit: I deleted the words “explosion in” – let’s not overdo this) is down to a brand new addition of a module making similar chart types available in R.

To explain how I stumbled on to the idea of creating a joy plot, first I probably need to explain that my blog posts are slightly out of order. My last blog post was based around the choice of colour for a tennis visualisation I was planning, and after a long process of obtaining, poring over and cleaning the data, I was left with my finished visualisation. Cue my next post (currently about 25% written) about working with datasets over a longer time in larger projects, and showcasing my resulting visualisation. But as is my wont, I continued to be distracted by other ideas for new projects.

Since my aforementioned tennis dataset consisted of the individual game/set results for every Men’s Singles match at Wimbledon from 1968-2017), it seemed a shame not to use such a rich data source for new visualisations. As will become apparent, my visualisation considers the full dataset of male tennis players and compares the journeys through their individual careers by looking at the +/- scores of cumulative games won. But it doesn’t split the trails into individual lines for each player. What would be a good “small multiple” type way to show this? I began to try one or two things.


My first attempt (above, not annotated), might be described as a series of sparkline area charts. This shows the years from 1968 to 2017 on the x-axis, and the top 40 tennis players on the y-axis, in terms of total number of wins. Players are ranked in order of debut – Roger Federer is about the ninth player down, showing the thickest area marks over a particularly long career. At this point, I’ve already had the light bulb moment that this looks a bit like the album cover I’ve seen in discussion with the current trend of joy plots, so have deliberately made the chart very minimalist using white marks on black backgrounds.

But this isn’t a joy plot because there’s no overlap. Now it’s important for me to say at this point, that although it seems like stating the obvious, some people might assume I haven’t thought of what come next: I know that overlaps aren’t good. I know that 3D effects aren’t good. If I wanted to get analytical value from a chart, this might not be a fantastic visualisation but it serves a purpose, especially if annotated. But I am interested in discussion, in technically challenging myself, in new styles, in art within visualisation, and post-punk 1980s music, All good reasons to pursue a joy plot.

As might be expected, this needed a bit of thinking “outside the box” – with apologies for the cliche, it required less orthodox processes within Tableau. Once I realised how to overlap the individual lines, I reached a few possible joy plot alternatives: either fully opaque or partially transparent, and with or without grid/zero lines.

version 2bversion 2Dashboard 1a

The second and third options here are starting to look quite nice. Option 3 here shows some transparency, allowing all values to be seen (I’m well aware that I’m not showing player names/axis values at this point). Option 2 in particular, from a distance, looks like clouds drifting across the screen.

At this point – a quick (very high-level) technical note on how I did this. From a Tableau point of view, putting player detail on the row shelf will always lead to a non-overlapping visualisation, so the answer lies in plotting co-ordinates. I looked around for some examples and inspiration and found that Bora Beran has recreated the original already. See his blog here and you can download his Tableau dashboard which does that. Looking at Bora’s work, I could see that it was done creating a polygon for each row, with each of the points pre-determined rather than calculated within Tableau. So, I exported the data from my first chart (top forty men only), added some bottom “corner points” and some path indicators to distinguish the polygons, and re-imported to a separate project. Then I just need to offset each polygon by a multiple of the rank, and we’re left with overlapping horizontal polygons. Feel free to download the workbook from my visualisation below for any further insight.

Rightly or wrongly, I decided if I was going to do a joy plot, knowing its aesthetic advantages and its analytic disadvantages, I wanted to go the whole hog and go close to the album cover. I changed white shading to black, and here is the final result. Note that my interactive version adds a splash of colour to highlight individual players, in an admission of the downsides of overlapping trails (click on the image to access the interactive Tableau version).


Its worth noting that, as mentioned at the top of the post, this is absolutely not to everyone’s taste. Bora Beran in his blog faithfully imitating the original Joy Division album cover calls it dataviz art, and that’s exactly what it is. You certainly won’t find this in any instruction textbooks.

Here was reaction from Matt Francis – former Tableau Zen master and one of the people I respect most in the community. Now clearly he’s not overly impressed with joy plots which is both fair enough and understandable. But the point he makes is 100% right!

And here is Andy Kriebel’s response – Head Coach of the Information Lab’s Data School and organiser of Makeover Monday. If there’s someone in Tableau data visualisation you should be paying more attention to, then I don’t know who it is (hint: there’s nobody).

Certainly, I had other similar comments. There were those who appreciated the appeal of the chart but didn’t agree with it. Those who, while grudgingly accepting that I was going to attempt a joy chart, preferred the transparent white cloudlike earlier iterations to the final black version.

So should I just show a bar chart instead to visualise the top 40 men of the open area in Wimbledon? Maybe – the below took me two minutes flat and makes it abundantly clear the top forty winners of games and the dominance of Federer, Connors and Becker.

Sheet 11

But this won’t be remembered, discussed or explored. Investing time in exploring and understanding a visualisation can lead to a greater understanding and recollection of key information (that’s not an asserted fact, it’s just my own hypothesis, based on how I interact with my favourite visualisation types). Being eye-catching, unusual and fun can be plus points to any visualisation, if used in appropriate situations.

Back to Matt Francis’ question above. To paraphrase – instead of investing so much time devising and iterating on a joy plot, shouldn’t I just consider whether it should be used in the first place (the implication being that because it hides data, that it shouldn’t be)? Well, I’ve done that, and come up with the answer: “yes, it should”.

The key advice that will always be offered when considering a data visualisation is “consider your audience”. Therein lies the answer. If my audience were the general public; if I’d been commissioned to write an article for tennis fans or newspaper readers; if I were writing a report showing findings of my research of 40 years of results, there’s no way I’d create a joy plot. I’d probably spend time making the above bar chart look slightly prettier, include that, and move on.

But my audience isn’t the general non-visualisation consuming public. My audience consists of the following:

  • Data visualisation enthusiasts, whether sticklers for analytical best practice or fans of artistic less functional projects
  • Blog followers, many of whom enjoy a debate on pros and cons of various methods
  • People professionally interested in a variety of visualisation techniques and chart types
  • People who know what I’m like, and who enjoy my, still novice, foray into data visualisation

So, I make no apologies for my joy plot. It hasn’t brought joy to everyone, but despite some of the constructive criticism above, some of you loved it, and I thank you for it!

Of course, the sensible thing would be to leave it there. After all, I’ve got personal projects to consider, not least an out-of-sync tennis blog post to complete. But, being me, I stumbled upon Henrik Lindberg’s next idea and couldn’t resist  Depeche Plot, anyone?



Of course this also has its faults: it’s possible for a line to cross the circle twice, becoming a chord, for example. And although it’s explained with annotations, it’s not going to be a particularly intuitive chart. But, like my joy plot, it’s possible to explore more and find out everything necessary by hovering and examining tooltips. It’s imitating album art, contains genuine explorable data, and is fun. And yes, I’m wondering if there are any other album cover data art possibilities. Why not?!

I’ll continue to find joy in as many ways I can visualising data. Even if it means looking beyond best practice bar charts!



What did I learn from Tableau Conference 2016? (part 2)

What did I learn from Tableau Conference 2016? (part 2)

Part 1 was posted a few days ago here, featuring numbers 1-8. We’ll jump in at number 9.

9. Tableau is moving into data preparation.

With the announcement of a three-year roadmap followed by the developers on stage, Tableau announced a raft of new features large and small. These are all well documented now in a number of blogs (and Tableau’s own website here). As an aside – on the way is the ability to import from PDF. Wow! But the one I want to focus on is Project Maestro. This is Tableau’s name for a suite of data preparation tools within Tableau itself. As a one-man team at work, I spend a lot of time cleaning, manipulating, reshaping and pivoting, but high-end data shaping tools such as Alteryx are prohibitively expensive. This seems to be a direct rival, and on speaking to people who use Alteryx and the like, they have good reason to be worried! I’m excited to be able to have much more data-wrangling available within the visualisation software. It feels like a game-changer, and it’s likely to raise the game of Alteryx (other data preparation software is available!) too. I’m sure they’ll do things to stay ahead of Tableau’s data preparation capabilities.

10. An artistic solution is often better than a complex solution

Michele Tessari of Tableau gave a talk about Artful Data. I was expecting a session about nice artistic visualisations but I really liked the way he balanced the artful approach with the analytical. Best example was in calculating and displaying the median of a dataset. The string of complex calculations left most of us open-mouthed in complexity. And yet, set up a curve showing cumulative percent, and hey presto the 50% mark is the median. Done in seconds, more easily, just as accurate, and a clear example of the advantage of an artful visualisation over a complex set of calculations.

11. Make sure your visualisation stands up in greyscale

There’s so much talk in data visualisation about correct use of colour. For that reason, this one piece of advice from Andy Cotgreave (which I haven’t heard before) seems counter-intuitive at first, but makes perfect sense. If you have a well-chosen colour palette then it should be perfectly readable in greyscale (just black, shades of grey and white).

Rather than critique some visualisations out there, I’ve decided to put a couple of my own more colourful works under the proverbial microscope. Time to admit that the first one (a France tile map using a large range of pastel colours) has not worked well in greyscale. The second (a heptathlon-themed visualisation) has worked nicely in my opinion, with the full range from green through yellow to red distinguishable in the Overall Position column.

One out of two looks OK – perhaps I should go back to the drawing board with the France visualisation!

France tile map – some colours are hard to make out in greyscale
Heptathlon data visualisation – originally used palette of red, yellow and green


12. #VizForSocialGood

There were many separate sessions on a common theme – the idea of using visualisation for a good cause. Schedule clashes were such that I missed sessions about the work Tableau Foundation are doing to support non-profit organisations, as well as a great talk about the work of Tableau geniuses Datablick and how they are using tableau mapping to help eradicate malaria in Zambia. I didn’t attend that talk because I don’t really have experience mapping in Tableau, but have heard nothing but good from those who attended and those who have taken part. But in addition to this, the final gathering of the conference introduced the idea (and hashtag) of #VizForSocialGood (from Chloe Tseng/@datachloe again). I like this. As we continue to learn and improve, it’s important to work on personal projects you want to work on and enjoy (as I alluded to in the Steal Like an Artist section here). But I think that should include not just cricket, or music, or books, but important causes, especially as presence in social media networks grows. I refuse to use such a cringeworthy term as a New Year’s “Vizolution”, but it’s definitely on my list to do more of over the next year. And soon.

13. The Double Pivot

Another technical takeout here, which I won’t write too much about. I have to include innovations in survey data reporting though. Zen Master and leader in the field of Tableau survey data reporting gave a talk on new ways to visualise survey data given recent improvements in version 10. Most prominent was the “double pivot” –  a way to pivot demographic data and merge back with rest of the pivoted survey data. The upshot is a great way of showing traditional banners of demographic/dimension data. Survey data is my bread and butter professionally, so it’s raised some interesting possibilities for professional work. So it’s included in my list.

14. Let the Data Decide the Visualisation

Another tip from an Andy Cotgreave talk – is “Let the data decide the Viz.” I like to think it’s obvious but it’s such an important thing to bear in mind that it bears repeating. Particularly within the confines of an environment where you might be keen to try something new. You can’t just learn how to do a sunburst chart and decide you’re going to use it for your next visualisation. Data first, then visualisation type, not the other way round. Below is a very famous visualisation which Andy referenced – creatively and brilliantly set up to represent descending blood representing casualty numbers. But it only works so well because the data fits. Unnamed_CCI_EPS

15. A word on Tableau’s community

Leaving the best to last, but I have to be careful here. A few months ago, Ben Jones of Tableau Public wrote this excellent blog pos about building a thriving community: http://dataremixed.com/2016/07/building-a-thriving-community/ – in particular including this graphic


I said I need to be careful here – with the welcome I got from so many of the community at the conference I’m in danger of rolling down the back of the hill straight into the love fest! I can’t talk highly enough about the welcome I received. It seems that if you (a) do a lot of visualisations, (b) tweet a lot, (c) have a blog that people read (this one, believe it or not) and (d) are English, for added novelty value, then there are so many people to meet, and so many people who recognise you. People, some of whom I didn’t know but many of whom I admire, introduced themselves, shook my hand and just wanted to share conversations and experiences (and sometimes beers!). I thought of mentioning everyone here who I met who I spent time with and made my experience better in their own way. But it’s no exaggeration to say there are too many. In fact, much too many. And to single out just a few would be unfair. But you all know who you are, and thank you all.

Added to this was a genuine big surprise. Tableau Wannabe Podcast Hosts Emily Kund and Matt Francis host the “Vizzies” awards, for various categories  voted for within the online Tableau community. The (joint) award of “Notable Newbie” for me, presented on the final morning, was the icing on the cake for me. And it was a hell of a good cake even before then.

Here’s a link to Matt and Emily’s podcasts including one with more details on the Vizzies (I’m not sure the awards are published yet).

And here it is.


Back to Ben Jones’ blog post and diagram, it’s true the “love fest” isn’t the most constructive way for the community to grow. Perhaps we can make an exception for 13000 or so like minded people meeting face to face (this sounds bad doesn’t it?!). But I love the idea of “iron sharpens iron”. The result of the face to face contacts and friendships I made is that the number of people I can call on for constructive criticism as well as help and advice has greatly increased. I’ve already been pleased, since the conference, to be constructively talking to those I previously only knew tenuously. And I genuinely think that one week in the year spent talking about visualisations, meeting people and socialising (talking about anything but visualisations!) leads to the other fifty-one weeks with a stronger support network. Of course there’s no need to wait for national congresses – we are all encouraged to join local user groups and meet-ups, and I’m delighted to have taken part in the inaugural Birmingham Tableau User group already since returning from the conference.

16. And finally …

That’s it really – I could have a miscellaneous section on non dataviz-related weird and random stuff, such as (a) people take pigs onto planes from Austin airport, (b) a previous governor of Texas could have been my ancestor, (c) the giveaway squishy Tableau brains make great cat toys (and I admit that’s a touristy “Keep Austin weird” mug accidentally in the background, or (d) the local Voodoo do(ugh)nuts were amazing, even if shaping them into a “T” for Tableau was perhaps more suggestive than intended, but it’s not that kind of blog. Oh, go on then. Hoping to continue this blog series in a little under twelve months time from Vegas in 2017!


Is it OK to steal?

Is it OK to steal?

So, is it OK to steal? Of course not. In the data visualisation context, that would be passing someone else’s work off as your own. Anyone would agree that would be the wrong thing to do.

But that would make a very short and uninteresting post. At the very least, I’m shooting for quite long and uninteresting (anything better is a bonus). So, is it OK to base your visualisation heavily on someone else’s work, to the point where it’s quite obvious that’s what you’ve done?

Here’s a case in point – a recent visualisation by Rody Zakovich. Normally I’d make a link to the interactive version but in this case I’ll give a link to Rody’s public Tableau page, mainly because he is prolifically good and there’s a lot more great visualisations than just this one.

boston-red-sox-visualizedAnd here’s a recent visualisation from me (click through for interactive version):


I make no bones about it – it’s pretty much the same viz. Names, colours, logos and timescales have changed, and the rest is either the same, or an equivalent to take account for the fact that UK football does have “winning seasons” or World Series, but does have a 96-team four-tier structure, promotion/relegation and the FA Cup (not that Birmingham City have ever won it). Oh, and I changed a z to an s!

If I was critiquing Rody’s original, I’d say that a sunburst chart isn’t the most effective way of being able to ascertain exact information, or to easily compare season versus season numerically. As any visualisation nerd will know (and I count myself as one), the radial arrangement of years doesn’t really mean anything.

But I’m not critiquing it, because that isn’t the point of his visualisation. There’s no getting away from the fact that at a glance it looks appealing. You can tell good years from bad years, see periods/etas where the team was strong and where the team was not so strong. It’s a bold, striking visualisation and if I was a Boston Red Sox fan I’d love that in poster form on my wall. Now, I’m not a Boston Red Sox fan (surprisingly for a Brit, not because of baseball ignorance, but because I’ve been lucky enough to go and see their divisional rivals the Tampa Bay Rays play a couple of times, but that’s another story). But, I am a Birmingham City fan. Hence, my remake.

It has the same pros and cons of the original. If you want easily readable, accessible and comparable stats, you can just go to Wikipedia or the club website, where the figures were sourced from in the first place. But as an “artistic” visualisation in one place, I like it (though I say so myself). My hope is that Birmingham City fans and visualisation fans would like it too, though I admit these are both pretty small pools of people.

But because this was so similar to an originally published piece of work, the first thing I did was to contact Rody. He was delighted to talk through the project, the code, and the tricky bits, and really keen to see the final product. Though I wanted to do as much as I could from scratch, I certainly knew after discussing it what it made sense to change and what it made sense to leave exactly as it is. After I published and publicised my visualisation, he was the first to like, retweet and promote it. My version (as you can see) acknowledges him for the original Boston visualisation – he didn’t even think it needed acknowledgment but I wanted to do so.

And then, also that day, I witnessed a similar example – Andy Cotgreave from Tableau published a small post entitled “It’s the small dataviz things” on his blog. A small comment on a really nice small touch on some recent data visualisation. Hang on a minute, isn’t this exactly the same as The Little of Visualisation Design on the Visualising Data website? This has been done 27 times already – this looks blatant to me!

But before I could worry too much or formulate an overwhelming opinion on the subject, I noticed that Andy posted online along with a playful acknowledgement that he was blatantly stealing the other Andy (Kirk)’s idea (I can’t remember the exact wording). Was Andy K annoyed? Worried? Angry? Of course not – his reply was that it was great to see his idea leading to more spin-offs. A great positive way to look at things and further evidence to me that if you like someone else’s idea, if you use it with very little change, but you acknowledge and interact with the original creator, then there’s plenty  of room for all versions.

With the discussion and content of this page now firmly in my mind, I sat down a few days ago to write this entry. I wrote the title, fired up a podcast, where Matt Francis and Emily are interviewing Alberto Cairo, with the idea of writing this up, multitasking a little round the house, and listening to the podcast. An hour and half later, I had still only written the title. Testament to Matt, Emily and particularly Alberto, the podcast was enthralling, and you should listen to it here:


As a result of this – I didn’t get this post written (or anything else done), but wrote down several ideas for future blog posts, some of which you may see soon. However one thing stood out – the phrase: “Steal Like an Artist” which was discussed towards the end of the podcast. This was a phrase that Alberto recommended and encouraged to data visualisers looking for great examples and influences/influencers in the field.

So what does this mean? The main thing is that everyone should almost be encouraged to “steal”. If there’s someone whose work you admire, copy it! Copy it well, understand it, acknowledge it, and in due course adapt it, so that you find your own style. Some people think Dali influenced Picasso, others think the reverse, but it seems there was a certain amount of “stealing like artists” going on, helping both become arguably the two biggest names in early 20th Century art. But going back further in time, most of the true master artists did this. A lazy google of “Steal like an artist” almost exclusively reveals the title of just one  well-acclaimed book. I won’t publicise or recommend it, because I haven’t read it. But I’ll steal(!) one line from its amazon page which says:


Nothing is original, so embrace influence, collect ideas, and remix and re-imagine to discover your own path

So that’s why I highlighted the word “artistic” several paragraphs previously. I’m not an artist and never have been (as is pretty obvious here), so need to be reminded to think in those terms. Now, I will remember to steal like an artist. In particular, in data visualisation (and even more in particular, in the use of Tableau, where public projects are downloadable to all), the key thing to do is understand what you are stealing or adapting for your own benefit so that you can work towards including your own influences. And, if anyone ever wants to steal from me, then I know, unlikely as it seems, that I will have made it!