The little of visualisation design: Part 41

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 confusion that arises through careless colour choices. The specific subjects here relate to a pair of example charts used on TV during recent Wimbledon (BBC) and Cricket (Sky) coverage in the UK. It is rather uncanny how both examples above demonstrate the same issues, even though they are published by different broadcasters.

In each of these displays above and below, you will see there is contextual colouring on the ground to provide some notion of whether the tennis serves or cricket deliveries were hitting the right areas. In both cases red, yellow, green are committed to what is essentially the background colouring (btw, the issues of colour-blindness are another matter). The problem demonstrated is the way the red and yellow colours are then re-used to be associated with a second, different meaning in the foreground data layer. In the tennis graphic above, red means an ‘ace’ and yellow just means other serves. In the cricket graphic below, red means ‘edged’ and yellow means ‘missed’.

By associating individual colours with two distinct meanings this causes unnecessary confusion for the eye and the brain to identity and remember what these meanings are as the reader pans across the graphic. A reasonable rule to follow is once you commit a colour to mean something you should preserve this exclusive association within the same view. In both these examples, it would be straightforward to find different visual cues through alternative colours or maybe filled/unfilled shapes to create more distinct representations.

The little of visualisation design: Part 40

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 continues the theme raised in the previous #LittleVis posting where I discussed reducing unnecessary thinking. The specific example here comes from a chart published by the BBC in this article about the slowing of the UK housing market.

The line chart shows the upwards and downwards trends of two house price trackers changing over time, simple enough stuff until you glance at the x-axis labelling at which point it is quite easy to becoming confused about judging the ‘when’.

You might expect that with the most recent value being a June figure for 2017 that the axis label intervals would follow maybe an annual cycle for each June going back a few years. Instead, there’s a real mismatch going on. Firstly, the majority of the labels follow an 18-month interval, which feels incongruent with the year-on-year values plotted. Secondly, these 18-month intervals are pivoted around the months of July and January. Thirdly, there is missing interval label which leaves a strange 35 month gap between the final and penultimate label.

Chart apparatus like this should assist not hinder. Annotation is arguably the least technical design matter to handle, just being guided by common sense can often be a good enough indicator of what the best choices should be.

The little of visualisation design: Part 39

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 making small changes to eliminate unnecessary brainwork imposed on your reader. The chart in question was published by Mike Galsworthy on Twitter and analyses some data from Survation’s tracker on the Customs Union.

As you can see there is a colour legend positioned on the right that explains the series representation of each line in the chart. The inefficiency here is created by the distance that exists between the legend and the series lines themselves. There is an unnecessary amount of extra brainwork required to store the colour meanings from the legend and then retrieve these associations when focus is moved over to the task of reading the chart. With the colour hues being quite similar it is quite easy to forget which line represents which opinion, especially as the legend ordering does not match any value ranking.

This is a very crude cut-and-paste reworking to illustrate how the burden on the brain can be reduced by moving the series labels alongside the relevant lines. In this case I felt positioning them to the right of the more recent values would be the most logical placement.

This is not a big issue (hence the post series title), the chart published is clearly intended to be a quickly turned-around piece of analysis and you can still see and process what the chart is showing. This is about how small decisions influence the optimisation of a visualisation’s accessibility: always seek to eliminate unnecessary thinking so that the brain can instead focus on the task of understanding what it is saying and what this means.

The little of visualisation design: Part 38

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 a nice styling approach to annotating charts. The work in question was published in The Washington Post by Reuben Fischer-Baum and covers analysis that answers this question: “Does your team botch the NFL draft?“.

There are many things to enjoy in this project but I was struck by the subtle impact of the hand-sketched style of the chart annotations. As you can see these notations point out tips for how to read the chart but also illuminate certain features of data that are of particular interest and/or offer some domain context that might explain the effect seen.

By presenting these annotations in a completely different style to the chart’s standard look-and-feel it creates a neat distinction with, for example, the chart’s value/axis labels. In contrast, these items of chart apparatus are (rightly) relegated to a somewhat utilitarian role with the sketchy notes jumping to the surface, offering small snippets of commentary in written form that you might otherwise verbalise and point to if you were presenting it live. The example shown above is just one of several charts that employ this technique

The little of visualisation design: Part 37

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 small way in which a chart can immediately lose any sense of elegance. The chart in question was published in The Times overnight showing some latest modelling data from YouGov about the upcoming UK election.

If you look down the y-axis you’ll notice the irregular space used to accommodate the ‘Plaid Cymru’ label. Unlike most of the other parties, no abbreviation or acronym has been used here which essentially means a double-row space is used, breaking up the visual rhythm of the arrangement of the bars.

I know I’m probably guilty of being that guy but even though it might only be ‘only’ a bar chart we should still care about its appearance. This irregularity in space usage creates a jarring interruption to the flow of reading. Consistency is one of the most important concerns in visualisation design – consistency in line spacing, consistency in alignment, consistency in font, consistency in category label format. Why not reduce the label to PC? If you are interested in politics you’ll know what this stands for just as much as you’ll know what NI, SNP or LD stands for (incidentally, why not use ‘Con’ rather than C for Conservatives?). Here’s a photoshop reworking…

The little of visualisation design: Part 36

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 a clever way of cleaning up space and maximising the role of gridlines. The visualisation in focus here is from a few years ago and is one of my favourites for teaching, it comes from the New York Times and looks at the achievement of Peyton Manning breaking the record for the number of career touchdown passes.

What I like here is the cropping of the horizontal gridlines indicating the position of the 400 and 500 values along the y-axis scale. Only three players have ever reached or exceeded these figures in their careers so it makes sense just to provide this reading assistance to the right of the chart where these lines are plotted. This in turn cleans up the overall display on the left, enabling the title and lede to occupy the empty space left behind by not having fully extended gridlines.

The little of visualisation design: Part 35

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 value of consistent composition and style when using photo-imagery as an encoding or annotation device. The visualisation in focus here comes from the Washington Post and reflects on ‘Trump at 100 days‘.

It might appear to be a relatively simple matter, but gathering and preparing photo-imagery, especially people-based subjects like in the sequence of Presidential head-and-shoulders in this piece, can involve a lot of effort sourcing, editing and compiling each one. It is easy to take the elegance of the final solution for granted, especially in the pressured environment of media/news publishing: it is far more obvious to the eye when this effort has been absent – inconsistent face sizes, a jarring range of different facial expressions, subjects facing different directions, cropped foreheads, lack of care over the original image outlining/cropping.

Look closely at these and if we are being extraordinarily picky, you’d say the JFK (looking slightly upwards) and Eisenhower (facing to the right) images feel slightly inconsistent with the rest and I might have re-sized Clinton’s face a little larger to reduce the impact of his prominent shoulders, but there are always contextual factors to weigh up. Otherwise these look super nice, occupying the circular bubbles really neatly and with a suitable, cohesive overall style.

Other matters also come to the surface: what image would be most representative of Barack Obama, the young fresh-faced ‘yes we can’ guy or the grey-haired sage ‘I need a holiday’ guy? Do I have permission to use this image? The overall point is that this stuff isn’t always straightforward.

(Incidentally, look through the full project, see the ‘Washington DC’ diamond being used, perhaps similar to #LittleVis 33?

The little of visualisation design: Part 34

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 is a somewhat agonising choice, such is the variety of brilliant ‘little’ things on offer to select (a ‘we are not worthy’ Wayne’s World bow to the mini, multiple Marimekko charts), but in the end I decided to focus on the very final feature included in the project. The visualisation work here comes from Periscopic and is a study of the ‘The Emotional Highs and Lows of Donald Trump‘.

The simple thing that just struck a chord with me here – perhaps amplified given the subject matter – is the respect, care and attention given by the creators to equip the audience with comments on the degree of accuracy this data portrays. The data plotted is gathered from emotions as analysed and recorded by the Microsoft Emotion API. It is an inexact science, a technique still developing in its sophistication but nevertheless it is a worthy approach to automate and summarise the detection of emotion through video, as described in the comprehensive descriptive text.

By specifically including the ‘Accuracy’ statement, the folks at Periscopic demonstrated the first principle (in my view) of good visualisation practice: ‘trustworthiness’. This transcends the goal of accuracy in a way because not everything can or will be 100% accurate. This doesn’t mean we can’t still portray this data, rather it means we have a responsibility to be transparent, equipping an audience with confidence and/or an understanding about what data they are seeing, what shortcomings may exist and advising on the amount of ‘pinch of salt’ they need to shape any interpretations or conclusions.

The little of visualisation design: Part 33

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 is a rather subtle matter concerning an example of how to squeeze out an extra final percent of creative design thinking. The project in focus comes from Dutch studio Clever Franke and is a new tool called ‘District Mobility‘, which visualizes the transportation demands in Washington DC and enables people to understand the bigger picture.

The simple matter to comment on here is the repeated utilisation and symbolic consistency in the use of the diamond-like shape of Washington DC. This is such a fantastic shape to work with, in the first instance, as a region for plotting thematic analysis and, secondly, as a visual archetype for different interactive and data mark features across the entire project.

In amongst the urgency and pressures of developing a visualisation solution, the best designers in this field have that extra sixth-sense and presence of mind to spot opportunities like this: to see patterns of form emerging that can be used in a way to bind a project together as a whole. Sometimes these things can prove to offer only gimmicks or represent metaphors that are too stretched and/or cliched, but when they work well they can really enhance a piece.

The little of visualisation design: Part 32

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 judging your maximum axis scale when handling outlier values. The project in focus comes from the New York Times in an article ‘Is Terrorism Getting Worse? In the West, Yes. In the World, No.‘. One of the charts looks at the annual trends of terrorism-related deaths in the USA going back to the 1970s.

Clearly, in this dataset, the 2977 victims of the tragedy of September 11th 2001 stands as a huge outlier compared to the general trend and one may have normally plotted this chart with a maximum range up to ~3000 to accommodate this largest value. However, in this case, the analysis is focused on the framing the underlying trend which supports the view that terrorist atrocities are of relatively lower prevalence in the US compared to other regions of the ‘west’ (the full article provides prior context for this with other analysis).

In support of the main point, the maximum y-axis value is astutely capped to accommodate the second highest value (the 170 victims of the Oklahoma City bombings), which slices the top of the rising peak for 2011. The designer is not remotely seeking to downplay the significance of 9/11 nor diminish the significance of the loss of life of every one of the victims, but the point here is to position this count as outrageously larger compared to the norm. Sometimes cropping large values is justified to position them as (legitimate) non-typical outlier values and to allow a clearer perception of the pattern across the smaller values.