Data Visualisation: A Handbook for Data Driven Design

by Andy Kirk

Traits of nothingness in your data or analysis – gaps, nulls, zeros and no insights – can prove to be the main insight, as described earlier.

You might recognise a point after which your ongoing efforts to refine may result in diminishing returns.

Keep notes about where you have sourced data, what you have done with it, any assumptions or counting rules you have applied, ideas you might have for transforming or consolidating, issues/problems, things you do not understand.

If you do not know something about your data, do not assume or stay ignorant.

If it is not showing what you expected or hoped for, you cannot force it to say something that is simply not there.

In the context of a visualisation project, editorial thinking is about determining what analysis you are going to portray visually to your audience.

this stage is the critical bridge between your curiosity definition, your data work and the design steps that follow.

Editorial thinking is concerned with making informed judgements about the content you intend to include in your visualisation. In my view this is one of the most defining activities that separates the best visualisers from the rest, possibly even more so than any technical talent or design flair. Before we move on to making design choices, you need to consider: given all the things you could show, what will you show?

You can rarely show everything – you rarely should show everything – so what are you going to show?

The term opinion can imply being impulsive or irrational but, in this context, it means you making discerning, informed judgements.

every visualisation you ever produced is the consequence of your subjective choices.

it is down to you to make the most reasonable call.

‘A photo is never an objective reflection, but always an interpretation of reality. I see data visualisation as sort of a new photojournalism – a highly editorial activity.’

City photos. Bar charts. Sketch templates. City statistics.

By considering some of the decisions involved in taking a photograph, you will find useful perspectives to shape your editorial thinking in visualisation. There are three particular perspectives to consider: angle, framing and focus.

There is only so much a single chart can show.

Before you choose to use a chart, you need to nail down what angle of analysis you want to provide.

I often find it far too easy to see everything as being potentially interesting to my audience, the curse of the analyst.

It is important, though, not to fall into the trap of lazy thinking that if you throw together multiple angles of analysis into your work, eventually one will serve the interests of your audience.

All the questions listed in Figure 5.1 reflect reasonable contributors towards collectively answering the curiosity.

Just like a photographer, a visualiser must demonstrate careful judgement about what data items to include, what data items to exclude, and why, for each statistical or chart display.

With visualisations, if you filter off too much of the content, it might disguise important context required for interpreting the significance and meaning of values.

One of the key motives of framing is to remove unnecessary clutter – there is only so much that can be accommodated in a single view and only so much your audience will be able to process.

This is the third editorial perspective and concerns which items of your data you might choose to emphasise, thus generating some level of focus for the attention of your viewer.

If everything is shouting, nothing is heard; if everything is in the foreground, nothing stands out; if everything is bold, nothing dominates.

The relevance of editorial focus is primarily associated with explanatory visualisations, whereby elevating key insights to the surface of a display is a key attribute of the experience they provide.

The angle is what fundamentally shapes the data representation approach. In simple terms, it determines which chart type is used.

The inclusion of such interactivity can be influenced by the editorial focus,

The inclusion of these captions would be a consequence of editorial focus to determine that these patterns in the data should be surfaced for the viewer.

As you will learn in Chapter 9, one of the key applications of colour is to create ordinal emphasis that brings important content to the surface and draws the eye’s attention.

This element of design concerns decisions about all of the positioning and sizing decisions.

The sequencing of the charts in the article is a function of editorial focus – what should go first, second and last, and why?

Turning data into a visual just because you happen to have it available is an aimless exercise. That is why we need to instigate from the origin of a curiosity.

Now that you have spent more time deliberating over your audience’s needs, maybe even asking them what they need to know, it is time to determine what is truly useful to them. This involves a blend of considerations:

Indeed, sometimes your audience will not really be best placed to know what is useful to them, in which case you may lead on what you want your audience to know.

Just because you have spatial data does not mean that the most useful angle of analysis will concern the ‘where’ of your data.

Although presented as consecutive stages, ‘working with data’ and ‘editorial thinking’ are quite iterative: working with data influences your editorial perspectives; and your editorial perspectives in turn may influence activities around working with data.

Representing your data visually involves the act of visual encoding. As visualisers, we encode our data using two main visual properties, marks and attributes. Marks are visual placeholders representing data items, such as distinct records or discrete groupings, depending on the form of your tabulation.

Attributes are variations in the visual appearance of marks to represent the values associated with each data item.

If marks and attributes are the ingredients, chart types are the recipes.

This would be an option to consider over the line chart when you have quantities for discrete periods (such as totals over a monthly period) rather than a purely continuous series of point-in-time measurements.

Sometimes one bar might be slightly hidden behind the other if the display concerns a before and after relationship.

An attribute of colour (usually the lightness property) is commonly used to distinguish contextual bandings behind each bar to aid interpretation.

However, a zero origin may be helpful to establish the scale of the differences depending on the subject matter being portrayed.

Any alternative method from this categorical family of charts would more usefully display the counts of text, such as a bar chart or a proportional shape chart where the word label sits inside a sized shape mark.

A histogram displays the frequency and distribution of quantitative measurements across grouped values for data items. Whereas bar charts compare quantities for discrete nominal categories, a histogram uses discrete quantitative ‘bins’ to form ordinal value groupings.

The plot is typically formed of a quantitative scale along which a line connects measurements of the frequency of each quantitative value. The line gets higher as the frequency gets higher. The connected line is then smoothed using various statistical techniques (that will depend on the subject context) and the area below is filled with colour to help visibility of the resulting shape.

The ‘candlestick chart’ (or ‘OHLC chart’ used in stock market analysis to track the opening, highest, lowest and closing prices of stocks) uses a similar method and is often used to show the distribution and milestone quantitative values for events that encounter constant change, such as stock market analysis over a given time frame based on showing the opening, highest, lowest and closing prices.

The total of all sector values must be 100% and the constituent parts must be exclusive and representative of a meaningful ‘whole’, otherwise the chart will be corrupted.

When you have multiple sectors, you might choose to emphasise only two or three parts through editorial selection.

The role of a pie chart is primarily about being able to compare a ‘part to a whole’ than being able to compare one part to another part.