May you chop axes? No…you must!
To chop or not to chop – that is the question for many information designers. We already know that we should never cut the axes of bar and column charts. But what about line charts?
In a recent entry on perceptive priority, I claimed that both versions of the chart below are acceptable. Even when you chop the axis of a line chart, the proportions are not distorted. Ultimately, the angles in both charts remain relative to the changes in value.
 |
 |
|
Redesign: zero – maximum scale |
Redesign: minimum-maximum scale |
Nevertheless, some prominent visualization experts insist that you should never omit the base line – even in a line chart – because it exaggerates the differences. ‘Exaggerating’, mind you, is not ‘distorting’.
Others, including myself, feel that you should utilize the available space to realize the largest possible differentiation, which in this case means the largest possible resolution. This next important rule in designing graphics forces us to concentrate the axis around the relevant areas (i.e. the values). We can easily communicate the range in a simple sentence without images.
Of course, it always depends on the context:
When it comes to management reports, we should be able to assume that managing directors are capable of interpreting a labeled scale for their own data.
In a glossy brochure, however, people often misuse a minimum-maximum scale to dramatize small differences.
If you read this blog regularly, however, I am sure that you will be able to understand the difference.
Where this (chopping axes) really goes wrong is when multiple graphs are presented — for example, if there’s one graph per category, or separate graphs showing different metrics (say, sales and costs). Not including the zero point is quite dangerous in this case because readers can’t tell how the changes in categories or metrics compare to each other.
Monday, February 23rd, 2009, 10:23 pmYou are absolutely right, but the “always include zero” commandment often enough makes it even worse. Why not use logarithms?
Tuesday, February 24th, 2009, 12:00 pmVery valid points.
Personally I first determine who my audience is…as you mention managing directors should be capable of reading the axis and not (pardon the pun) between the lines.
Another point, is that should you want to actually illustrate significant movements/difference etc. then I may chop the axes, and vice versa.
I would argue it wholly as it depends on the situation at the time. Unfortunately in this case, there’s no golden rule.
Tuesday, February 24th, 2009, 4:54 pmI like Tufte’s advice to aim for an average angle of 45 degrees (bumpy, not spiky, not flat). That tends to give a picture that lets us read the true story most accurately.
That said, I think Chad’s point is very important – we must be consistent when the audience may be comparing between charts.
Thursday, February 26th, 2009, 1:57 pmHi Nicolas,
I recently was doing a bar chart for delivery accuracy percentage for a number of stores where the percentages vary very marginally, between 95 % to 99.99 %. When the bar chart is part of the dashboard, the differences almost disappear. In this situation would it not be advisable to start from say 70. yes the rule of proportionality will be broken but the differences would clearly stand out.
Friday, February 27th, 2009, 4:24 pmOur dog-matic Bella would say: Always consider your design first, before daring to bend my Bella Reporting Standards. Paresh, I don’t know your data, but it seems high delivery accuracy percentages is what you would expect. So, maybe you better show deviations from average or expected values and everything will be fine. Maybe this example is clarifying how this could be done.
Friday, February 27th, 2009, 7:23 pm