The little of visualisation design: Part 26

This is part of a series of posts about the ‘little of visualisation design’, respecting the small decisions that make a big difference towards the good and bad of this discipline. In each post I’m going to focus on just one small matter – a singular good or bad design choice – as demonstrated by a sample project. Each project may have many effective and ineffective aspects, but I’m just commenting on one.


The ‘little’ of this next design concerns small-multiple grid maps and a neat way of providing a state legend. The project in focus here was produced by Nathan Yau and maps the spread of obesity across the US over the past 30 years.

lovd-part26a

As you can see, in the top left there is a reference guide that explains which grid cell relates to which state. With any grid map there will always be some degree of state position/neighbour compromise so this helps the reader to immediately (being the first thing you logically see) orientate themselves before they then move through the sequence of yearly patterns.

lovd-part26b

This feels a far less repetitive solution and not as visually intrusive as it would be to include the 2-digit state labels in each map and in cell. I can imagine Nathan toyed with the idea of having cell borders around each state label but I like that he didn’t do this – it is sufficiently legible through just relying on white space.

The little of visualisation design: Part 25

This is part of a series of posts about the ‘little of visualisation design’, respecting the small decisions that make a big difference towards the good and bad of this discipline. In each post I’m going to focus on just one small matter – a singular good or bad design choice – as demonstrated by a sample project. Each project may have many effective and ineffective aspects, but I’m just commenting on one.


The ‘little’ of this next design concerns the decisions involved when including arrows in your work to act as pointing or directional devices. A recent project that uses arrows comes from the Guardian, titled ‘Where is the riskiest place to live?‘ by Josh Holder, based on the world risk index. As you can see in the project there are several features of the chart that have associated captions, connected by arrows, to help draw to the surface some key or interesting insights from the data. (Note, there is a second use of arrows to indicate the general ‘Low’ or ‘High’ risk direction along the x-axis but I’m going to focusing on the ‘pointer’ arrows in the body of the chart.)

lovd-part25a

There isn’t an instruction manual for the most effective way to design and integrate arrows into your display. Instead, as with many things in data visualisation, it is about judging what will offer the most legible and elegant solution to suit the overall visual balance. There is a really nicely judged lightness to the visual weight of the arrows in this piece. They are short in length and consistent in appearance, which ensures they supporting the display not dominate it. Perhaps the most unusual feature is how the arrow head points towards the caption, rather than the other way round.

lovd-part25c

Arriving at the ‘best’ choice involves a surprisingly high number of decisions: What arrowhead* to use? What colour? How long/how far away should the caption be from the item of interest? What weight to apply to the line? Will it be straight or curved? Which is the start and which is the end? Where specifically to point the arrow to and where from? The next time you see an arrow, just remember how much thought has gone into what you might otherwise think is a straightforward choice.

(* Reminds me that Jane Pong ran a quick survey on peoples’ preferences for the arrows available in Illustrator)

The little of visualisation design: Part 24

This is part of a series of posts about the ‘little of visualisation design’, respecting the small decisions that make a big difference towards the good and bad of this discipline. In each post I’m going to focus on just one small matter – a singular good or bad design choice – as demonstrated by a sample project. Each project may have many effective and ineffective aspects, but I’m just commenting on one.


The ‘little’ of this next design concerns the challenges of handling long labels. In this post I’m possible breaking the theme of this series as I’m not so much offering a solution, rather more a ‘heads up’ to flag the possibility of this issue in your work. I’m referring to one of my own recent projects here, ‘Filmographics‘, looking at the ebb and flow of the fortunes of actors’ movie careers. As you can see in the screen shot below, when you choose an actor from the menus at the top an illustration of their face is displayed in the page body and their name is presented within this image.

lovd-part24a

The challenge faced in this case was judging the best font size for the actor’s name label within the (self-imposed) space constraint of the image width. We found a nice size for working with 59/60 values but then we had Arnie: at 21 characters in length, ARNOLD SCHWARTZENEGGER would be the single instance whereby the preferred label font size would cause the surname to be split over two lines. To make it small enough to accommodate Arnie would have the consequence of the name’s being (in our view) an insufficiently prominent title/identifier for all the other actors in the dataset.

lovd-part24b

In the end, with time running out, we made a decision to accept the rather inelegant compromise of sticking with our preferred font size that would be suitable for all except Arnie: ‘Good enough’, ‘It’ll do’ is often a call you have to embrace when time resources diminish.

lovd-part24c

With the benefit of hindsight we maybe could have looked to programmatically handle a custom font size for Arnie alone. We could also have handled the placement of the actor’s name entirely differently so as not to present us with a challenge of dealing with this spatial constraint. The main point to make here is to reinforce the importance of developing a deep and early acquaintance with all the physical properties of your data values, not just the range of quantitative values you’re facing but also the length of potential labelling assets.

It’s the little things like this that cause the big headaches.

The little of visualisation design: Part 23

This is part of a series of posts about the ‘little of visualisation design’, respecting the small decisions that make a big difference towards the good and bad of this discipline. In each post I’m going to focus on just one small matter – a singular good or bad design choice – as demonstrated by a sample project. Each project may have many effective and ineffective aspects, but I’m just commenting on one.


The ‘little’ of this next design concerns an interesting approach to labelling y-axis scale values. The project in focus here comes from the Washington Post, titled ‘How terrorism in the West compares to terrorism everywhere else‘ by Lazaro Gamio and Tim Meko, putting into context the relative levels of terrorist-related deaths in the West.

LoVD-Part23a

As you will see, the labels for the y-axis intervals are located within the chart. Specifically, they occupy the vertical space aligned with January 2016. Normally, we would locate these annotations to the left of the chart, maybe to the right, sometimes on both sides to assist the reader perceiving the chart’s values. It has never occurred to me before to think about positioning these labelling devices somewhere in the middle of the chart but I think it is a really innovative way of optimising their value.

LoVD-Part23b

The specific choice of using the January 2016 horizontal position to present these labels is interesting because January isn’t right in the middle of the chart. It may be a logical month to choose in order to create a subtle breakpoint or divide between judgments of 2015 and 2016 values. Most likely, it could be influenced by the presence of ‘nothing’ – the zero value of terrorist-related deaths in Europe/North America during this month creates the ideal (*) emptiness to occupy the labels. This approach continues in the bottom chart, where there ARE values to plot in this space, but the interval labels only begin from where the bar heights end, thus there is no obfuscation.

(* yes, I’m aware how the notion of design decisions being helped by zero death values and, conversely, being impeded by higher death values sounds – just remember I am just looking at this solely through the lens of a design challenge, not what the subject represents)

New project: The Pursuit of Faster 2016

To mark the completion of the Rio 2016 Olympic Games, I have been working, with my trusted lieutenant Andrew Witherley, on designing a new version of ‘The Pursuit of Faster‘ project. This visualisation explores the evolution of medal winning performances across all Olympic Games since 1896 as athletes strive for that ultimate pursuit of being faster than the rest.

DesktopFullView(640)

It portrays the patterns of improvements in the results of time-based events whether it be on foot, in water or on water. There are several sporting events where relative speed determines medal success but this analysis is purely concerned with results from races where absolute speed is the measure of success.

Choose a sport and select an event to see how Gold, Silver and Bronze winning times have changed over the years, for both men and women. Hover over the medal markers to reveal the actual results (you can switch on/off this semi-transparent pop-up, though, using the provided toggle). Below the main chart you can learn about the most successful countries in each event, the gaps that exist in result times between genders, a normalised measure of improvement over time, and analysis about the margins of victory between Gold and Silver medallists. To learn more about the background to the project, how to read/use it and notes on how data has been handled, you can click on the project’s ‘About’ button.

This project was originally launched in 2012 as part of a pre-London Olympics visualisation contest (securing a runners-up prize) but I wanted to employ a fresh design and incorporate all the subsequent results data from the 2012 and 2016 Games. The project now has a more adaptive design for desktop, tablet or phone. The image above shows the full-screen view on desktop, this next image shows the full-screen view of the mobile version. The only difference is that the main chart is transposed and the mouseover results are switched off in the mobile version.

MobileFullView

The little of visualisation design: Part 22

This is part of a series of posts about the ‘little of visualisation design’, respecting the small decisions that make a big difference towards the good and bad of this discipline. In each post I’m going to focus on just one small matter – a singular good or bad design choice – as demonstrated by a sample project. Each project may have many effective and ineffective aspects, but I’m just commenting on one.


The ‘little’ of this next design concerns interactively handling outlier values or series. The project in focus here is the latest ‘2016 Election Polls‘ by Wilson Andrews and Josh Katz, plotting the ebb and flow of national polling averages for the presidential nominees over time.

LoVD-Part22a

With T***p and Clinton dominating the election narrative, it is easy to forget (especially for non-US people like myself) that there are other independent candidates in the race, albeit with a significantly lower share of the polling numbers. That’s where the ‘Show Gary Johnson’ button comes in.

LoVD-Part22c

It makes sense to editorialise this project and just present analysis (in its default state at least) relating to the two main nominees. The chart’s y-axis scales are therefore designed to best fit these higher value series. However, by offering the ‘Show Gary Johnson’ button, readers can reveal the polling data for Johnson at which point the chart is reconfigured with the y-axis origin set to zero in order to reveal these lower value trends.

LoVD-Part22b

I love this feature so much I will do my utmost to ensure ‘Show Gary Johnson’ becomes permanently woven into the lexicon of this field.

The little of visualisation design: Part 21

This is part of a series of posts about the ‘little of visualisation design’, respecting the small decisions that make a big difference towards the good and bad of this discipline. In each post I’m going to focus on just one small matter – a singular good or bad design choice – as demonstrated by a sample project. Each project may have many effective and ineffective aspects, but I’m just commenting on one.


The ‘little’ of this next design relates to the possibility of offering a different emphasis on presented data through a small change in a chart’s composition. The chart in question is some analysis (only source I can find is here) by Robert Mann titled ‘Who Lies More: A Comparison’ about the degree of truthfulness or otherwise (as per PolitiFact’s independent ratings) of 50+ statements made since 2007 from each of the array of presidential candidate.

LoVD-Part21a

Given the focus of the analysis as a reader is largely about seeing the overall polarity of truthfulness, I would probably have pivoted the stacks either side of a centrally positioned baseline, with the ‘falser’ three categories (Pants on Fire, False, Mostly False) to the left, maintaining the connotation of being negative) and the ‘truer’ three categories (Half True, Mostly True, True) to the right. You could argue for ‘half true’ to be on either side, I would say in this era of politics it is probably true enough to be seen as a positive. This layout would still facilitate readability of the component sizes, just as as before, but it would also provide a more immediate overall view of the general balance of the integrity in these peoples’ statements.

(This is something John Nelson has written about previously, check out his article from 2011 and *update* Jon Peltier discusses how to make the proposed redesign of this chart).

The little of visualisation design: Part 20

This is part of a series of posts about the ‘little of visualisation design’, respecting the small decisions that make a big difference towards the good and bad of this discipline. In each post I’m going to focus on just one small matter – a singular good or bad design choice – as demonstrated by a sample project. Each project may have many effective and ineffective aspects, but I’m just commenting on one.


The ‘little’ of this next design involves criticism of my own work (“About time”, you say) and concerns the matter of design thoroughness. In my book, this is something I position as being part of the pursuit of elegance in your design execution. The graphic in question is one I quickly compiled and tweeted out yesterday morning, looking at analysis of Matt Damon’s roles in the the Jason Bourne movies, to mark the release of the latest movie in this series.

LoVD-Part20a

The issue I want to highlight here is the simple failure to be consistent in my use of a colon following the ‘Release date’ label in the captions below each movie poster. I missed two colons. A small matter you might think and you would be right but ever since publishing the graphic I’ve been agonising over this mistake. Remember, this is the ‘little of visualisation design’ and these type of errors (left in through a failure to thoroughly check work) demonstrate a failure to either find time or care enough to pay attention to the smallest of details. These are building blocks of quality visualisation design so care over every last detail.

LoVD-Part20c

The little of visualisation design: Part 19

This is part of a series of posts about the ‘little of visualisation design’, respecting the small decisions that make a big difference towards the good and bad of this discipline. In each post I’m going to focus on just one small matter – a singular good or bad design choice – as demonstrated by a sample project. Each project may have many effective and ineffective aspects, but I’m just commenting on one.


The ‘little’ of this next design concerns the use of colour to emphasise the primary editorial focus of a display rather than to represent different categorical associations. The analysis titled ‘How Far Is Europe Swinging to the Right?‘, by the New York Times, offers a political-persuasion breakdown of the voting patterns across the six most recent national elections for a selection of 20 European countries.

LoVD-Part19a

Typically, in most countries, each political party would have some established association with a given colour. Had this analysis being displayed using the party categorical colours then you would see a rather overwhelming technicolor extravaganza but a project that was editorially flat. However, the focus of this analysis is on the emergence of ‘right-wing populist and far-right parties’ and so colour (red) is only used to emphasise the constituent party stacks that are affiliated with the hard right. The categorical sorting of the stacks within further aids the sense of editorial orientation for the reader.

LoVD-Part19b

The little of visualisation design: Part 18

This is part of a series of posts about the ‘little of visualisation design’, respecting the small decisions that make a big difference towards the good and bad of this discipline. In each post I’m going to focus on just one small matter – a singular good or bad design choice – as demonstrated by a sample project. Each project may have many effective and ineffective aspects, but I’m just commenting on one.


The ‘little’ of this next design concerns the sensible positioning of categorical labels. This Daily Chart, by the Economist’s data team, offers a view of the identified political persuasion of people living in selected swing states of America. The chart is a variation of the connected dot plot, with a separate row for each state.

LoVD-Part18a

Typically, charts like this would have categorical value labels right-aligned to the left of the vertical axis. However, in this case, the labels are positioned with immediate proximity just to the right of the highest value – which is the value used to order the categories vertically. This approach aids readability, making it just that little bit more efficient to perceive the values and their associated categories.

LoVD-Part18b