Charts

Charts

Charts

Charts should be used to show an explicit message by allowing the reader to see the numbers, trends, patterns, relationships and exceptions in the data.
Compliance audit Financial audit Performance audit Review
Reporting
Ref: 18.341

Principles

We use charts in our reports to show an explicit message. Charts allow the reader to see the numbers, trends, patterns, relationships and exceptions in the data. If we are unsure of what the numbers mean (no trend or exception emerges), then we should re-consider the use of the chart and perhaps use a simple table instead. The main advantage of charts is that they are perceived by the visual system of our mind, which is much faster than the verbal system, which processes text and tables Charts should be self-explanatory, that is you should not need to read the report to understand them.

Instructions

Chart choice

To be successful in communicating with charts, you have to choose a chart type suited to the message. You should organise the display or illustration so that the point is obvious. Do not over-complicate matters by attempting to show two or more disparate messages. Avoid overloading the chart with data and variables. Use the simplest style of charts available. Simple bar charts communicate more easily than component bar charts. Simple line charts are more accessible than layered line charts. Visual perception is quite good in comparing lengths, following lines and detecting patters in two dimensions. Visual perception is not so good in comparing areas and colours. The possible messages can be associated to data relationships: how something changes with time (time series), how to items change together (correlation), which items are bigger (ranking)… Some chart types are better in showing some data and relationships, so there is almost a chart type for each kind of message.
Message Best choice Alternatives
Changes over time Time series chart conveys the connection between data points. A minimum of 3 points is needed to show a pattern. Limit the chart to five lines. Time series chart As more lines become indecipherable, use multiple small time series charts. Multiple time series chart For a single series, vertical bars are possible; however, bar charts are better for non-continuous data. Bar chart for a time series
Ranking
Comparisons
Note: always sort values.
Bar chart allows quick comparison between large amounts of data. Horizontal bar chart Pie chart is possible and can be visually appealing. However, avoid with many components or similar values, because it is difficult to compare the items. It is very difficult to compare data across two or more pie charts. Avoid this! Bad example of a pie chart
Part to whole Bar chart Vertical bar chart Use stacked bar chart, if several series. Stacked bar chart Pie chart is possible for a small number of items, which are not of similar size. Good example of a pie chart
Deviation from a single value(e.g. average) Bar char showing the difference from the given value Bar chart chart showing difference
Deviation from a trend over time Time series chart Plot the differences and draw the reference line. Time series chart showing difference
Distribution Histogram summarises vast amounts of data that fall into numeric ranges or categories. Histogram with bars Box plot is good for comparison several distributions Boxplots Line histogram can be used to show the shape of distribution. Histogram with a line
Relationship between variables (Correlation) Scatterplot reveals relationship by displaying a number of points. Scatterplot
There are many further chart types available. More complex charts should be used cautiously. If you would like to use a complex chart, consider: what is the message? Does the chart convey that message? Is there a simpler method of getting that message across? It may be preferable to use several simple charts to convey the message than to use a single complicated chart.
Strengths and weaknesses of selected complex charts

General design tips

After having chosen the chart type that helps perceive the message, each element of the chart (title, axes, labels…) should help to tell the story clearly and no useless element should remain, i.e. without nothing missing and nothing extra obscuring it. The following tips should help you to keep the design simple, yet effective. Do not follow them without thinking. Try variations and see which one makes your chart better.

Integrate text and the chart

Keep the chart on the same page as the text. State your message in the text and mention that this is also shown in a chart. For example, "We found that cows are 25% happier today that 10 years ago (see Figure 1) … Figure 1 - Evolution of cows’ happiness" Write your text while looking at the chart. Avoid sending the chart to an appendix.

Colours

Do not depend on the use of colour for interpreting facts. Design your charts using shades of one colour (grey or other colour). If you need more colours because 5 shades are not enough, there may be too many things in one chart. Add more colours only if they add information (example: green=ok, red=alert) and do not clutter the chart. Several hues of one colour are usually a good choice. Prefer soft colours in general and strong ones only to highlight exceptions. Respect the colour palette provided in the ECA visual identity guideline. Finally, print the chart in black and white to check that it will be ok for people with colour blindness or if not printed in colour.

Title

Charts must have a title. The title can explain and reinforce the message ("Unemployment descending in Spain") or only say what data is charted ("Unemployment in Spain 2005-2013") or both things.

Axes

  • Axes should be labelled at least with the units used (€, m2, %), optionally with the full description and units like: 'Oil price ($/barrel)'.
  • Remove decimals (e.g. 10.00 to 10) and reduce the number of digits by using the large numbers names (e.g. 10 000 000 to 10 million)
  • Place the axes near to the data (near the longest bar) or the most important data (the most recent value in a time series). The same axis can repeat on both sides.
  • For bars the axis must start at zero. Otherwise we show only the tips of the bars and not the data (= length of bar).
  • Do not use 2 vertical axes with different scales. It is very confusing. Either use only one scale, or separate in two charts or convert into percentages of change.

Text within the chart

Text within the chart should be large enough to read, and presented horizontally. Try to write the names the series on the chart. If it does not look good then use a colour-coded identification with a legend. Avoid data labels: if readers need to know the values, give them a good table rather than a cluttered chart. Avoid unfamiliar abbreviations.

Non-data ink

’Data ink’ are the components representing the data (lines, bars or dots) and ’non data ink’ are the rest. Non-data ink should be kept only if it contains useful information (titles, axes, legends, reference lines, important grids) and if it does not obscure the chart. Similarly, avoid any superfluous decoration (e.g. cluttered backgrounds, gridlines, multiple fonts, borders, bold and underlines), unusual formats or special effects such as 3-dimensional graphics.

Alignment of charts in the report

It is important that charts within the same report be presented in a consistent and comparable manner.

Showing accurate data

If there is a need to show accurately more data than is possible in a chart, consider the following possibilities:
  • Use a chart with your message inside the main text and include a table with the complete data in an appendix.
  • Add a data table to a chart
    Example
  • Use a table enhanced by in-cell/in-line charts.
    Example

Design tips for specific chart types

Bar chart: Arrange bars in size or logical order. Broaden bars for greater impact. Vertical and horizontal bar charts are equally easy to understand, but horizontal charts are easier to label, and readers expect money and time to be vertical. In a paired bar chart, ensure grouping highlights your message. Pie chart: Limit the number of slices, ideally five. Arrange slices in order, largest to smallest, starting at 12 o'clock. Label slices individually, do not use a key legend. Include the overall volume or quantity in the title. Histogram: Succinct explanation and simple design are essential. Make sure that readers know what the histogram represents by including the chart's message in the title. Ensure axis labels are clear and coherent. Choose between 5 and 20 categories - fewer than 5 gives no discernible pattern, more than 20 is unwieldy. Histograms are drawn vertically rather than horizontally. Scatterplot: Readers may not know how to interpret, so include an explanation of its conclusion. Label scales carefully. Trend lines are optional. Small multiple charts: Use same scale on all charts and consider adding the average (reference level) to all charts.

Resources

includes two examples describing the process of chosing the chart suitable to the message and designing the chart to convey the message most effectively.
This tool than can help you choosing colours.
can be used as a starting point for finding the type of chart most suited to the message you want to convey with the data
same as above but in the PDF format

References

Blogs

Books and articles

Last Modified: 21/10/2020 23:45   Tags: