The key adds nothing but confusion and clutter
Removing the key means there is more white space and more focus on the chart
Modifying the shape of your bar can be the right fit for an infographic
Changing the shape of a bar can help it to punch out of a colourful infographic
Size your decorated bars by stretching them in one direction only
Stretching in two directions distorts the data
A 3D stepped bar chart- I guess - with arrows? Nothing is worse than ostentatious overdecoration
Tilted 3D bars and strange half-filled 3D pyramid bars. Both excruciating.
Bars with base-aligned icons
Bars with icons in circles next to the bars
An awful 3D bar chart
A slightly less awful, but still awful, 3D bar chart
A still from a Carbon Visuals animation about the amount of water we use
A still from a Carbon Visuals animation showing how 3D charts can work in motion graphics
Bar chart with unnecessary y-axis title
Y-axis title positioned vertically and unreadably
massive y-axis labels
Bar chart with truncated y-axis labels
An annotation can give context, in this case, the number of zeroes in a trillion
An annotation can place the unfathomably huge number in a real-world context
Bar chart with abbreviated x-axis labels
Bar chart with wrapped/ hyphenated x-axis labels
Bar chart with cryptic initials instead of x-axis labels
Bar chart with missing x-axis labels
Bar chart with tiny x-axis labels
Bar chart with overlapping x-axis labels
Bar chart with staggered x-axis labels
Bar chart with rotated x-axis labels
Abbreviation is OK for days of the week, months of the year etc
Hyphenation and wrapping is OK if that's where the word would naturally break
Initials and acronyms work if they are commonly understood
Gaps are fine for things like time series where it's easy to mentally insert the missing text
Vertical bar chart with only a few bars labelled
Bar chart with abbreviated labels on x-axis
Labels and y axis - too much information
rule-24-all-new-bg-22.png
Unnecessary gridlines
Gridlines and labels together are unnecessary and cluttered
Fewer bars, directly labelled, no need for y-axis
Rotated labels
Same content, but in a horizontal bar
Reading a vertical bar sends the eye roving across the canvas. This is both a positive and negative.
Horizontal bars are easier to scan
Unnecessary y-axis title
Y-axis title moved into chart title - horizontal and legible
A long y-axis label with too much wrapping
A long y-axis label with not enough wrapping
It's better to keep labels shorter and the title slightly longer
An abbreviated title, shorter labels and the full question text in a footnote. Perfect!
The original bar chart
Rotated is slightly clearer, but boring?
Sized squares are more intelligible and interesting?
Isotype charts bring out the human element of any story
Polar area charts are visually compelling and easy to read
Bubble tables are visually appealing and easy to scan
Bar chart with truncated x-axis, not starting at zero. It makes the difference between the bars look artificially wide.
Bar chart now starts at zero. Now we can see the full story. The Conservatives claimed to be increasing police numbers, but they're still lower than they were before austerity.
Bar chart showing tallest mountains and deepest holes on Earth. The zero line (ground level) is clearly shown.
Even when we are talking about Mauna Kea, a mountain which rises up from under the sea, we should show ground level (zero)
Bar chart which starts at zero. It shows number of male babies born for every 100 female babies. The key story is lost because the differences between the bars are small but important.
An important change over time story is lost because the bars start at zero. This shows the number of accident patients that are seen within four hours - and how this has plummeted.
A dot chart showing the gender ratio story - better because it doesn't have to start at zero so you can see the differences more clearly
A flagpole chart showing deviation from a specific value or point in time.
That Accident and Emergency story again. A dramatic change over time is hidden.
If your line chart doesn't start at zero, don't show the line touching the x-axis.
An area chart should always start at zero.
Often, starting a scatter chart at zero buries the story. This one shows correlation between age at first marriage (female) and number of children. But half the chart is empty.
In this version of the scatter chart, we don't start at zero and this allows us to zoom into the content and see the clear pattern. Marrying young correlates with a high fertility rate.
This shows the percentage of children at risk of poverty in Western Europe and the Nordics. Italy, Spain and the UK are at the top.
This shows the same data, but we have deliberately got a value axis stretching to 100%. This makes the differences hard to see.
This bar chart shows levels of vaccination in selected countries. Our value axis goes up to 100% so we can clearly see the total possible value (the whole population is vaccinated).
Some of the lowest-rated movies on IMDB. Battleship Earth is the lowest, then Gigli and Cats. The full range of scores of shown (0 to 10) so we appreciate how low their scores are.
A line chart doesn't have to start at zero. It means dramatic changes are more visible.
Vertical bar chart with a gap in the y-axis, showing the top 10 languages spoken in England and Wales. A gap in the 'English' bar.
A vertical bar showing Top 10 language spoken, this time with a lightning bolt in the y-axis to indicate a gap.
A bar chart showing the Top 10 languages spoken in England and Wales, with the 'English' bar looking huge, and dwarfing all the others
Another bar chart in which the largest bar is an outlier, dwarfing this others. This time, it is Covid case data, showing the UK's terrible performance compared to the rest of Western Europe
The 'languages spoken' story is broken into two. Part 1 shows how Polish is England's second language.
Part Two shows how English still dominates as a language, and Conservative tales of 'nobody speaking English anymore' are lies
A chart in which the 'other languages' just become one bar, which is then exploded out, to show its individual parts - Polish, Panjabi, Urdu, French and so on.
Other languages becomes a bar by itself. A table indicates the constituent parts of the smaller, composite bar
Bubble charts work excellently for outlier stories, as the bubble can disappear off the side of the chart.
A nested treemap, followed by a treemap in which the 'other languages' box is exploded out. There is space to label almost all the key secondary languages.
Another example of how outliers kill bar charts. This time, a polar bear's home range (106,000 square kilometres) makes all other animals' (often huge) home ranges look tiny.
We replace the bar chart with more of an explanatory graphic, in which the polar bear's home range is compared to countries (it is the size of South Korea, or five times the size of Wales).
Front cover of NYT in which unemployment data breaks out of its chart boundaries and cascades down the page.
Front page of NYT in which the number of Covid cases shoots up and the final bar in the bar chart breaks out of its container, and fills the entire front page
Vertical bar chart with a distracting border.
The same vertical bar chart without a border
Charts separated using grey background boxes, rather than borders
Charts separated using dark background boxes (rather than borders)
A series of bar charts with borders on every side of each one
A series of bar charts with subtle borders on two sides of the chart
Gridlines dominating the chart
The same chart without gridlines is clearer and cleaner
Gridlines can help when the data labels would be far too wide for the bars
Gridlines can also help when you have many bars
A confused, cluttered chart with too many gridlines
Removing gridlines and clustering the bars into groups makes for a more legible chart
A chart where the gridlines are more dominant than the axis. It is hard to read.
Gridlines should always be subtler than the x- and y- axis
A chart where the y-axis is perhaps unnecessary
If you have data labels, you don't need a value axis
A bar chart where just the top and bottom are bars are labelled
A bar chart with axis labels but no axis
A bar chart where the axis line is too solid and distracts from the chart
A more subtle axis line
Tick marks on a category axis are always a mistake
Tick marks on a value axis are rarely, if ever, needed
A chart where there is no differentiation in line colour or weight between the axes, gridlines or annotation lines. It is hard to read.
A bar chart where the gridlines recede and the annotation lines are clear
A summary of the key things to avoid when making a bar chart.
A summary of the key things to consider when making a bar chart
A within group comparison using a clustered column
A between group comparison using a clustered column
A table showing the percentage of people working in each sector across BRICS countries
A heat table where the similarities between South Africa, Russia and Brazil are clear
A bar chart table, showing how people earn a living in different BRICS countries
An abacus or dot chart - perhaps the best option. It is clear that the dots for South Africa, Russia and Brazil are arranged in a similar pattern.
A waffle chart showing the percentage working in each sector in BRICS countries
A stacked bar - with the first bar (agriculture) highlighted for easy comparison
A bubble table - more elegant than clustered columns
An isotype chart - easier to connect with than a clustered column
A simple bubble table, with just one variable (agriculture) highlighted
A waffle chart with just agriculture highlighted
Possible use for clustered column - where you have a huge outlier?
A clustered column where the final bar in each column (Netherlands) is a clear outlier
A bubble table showing Netherlands as an outlier when it comes to growing and exporting fruit and veg
A bar chart table -showing the Netherlands as the clear outlier when it comes to exporting fruit and vegetables
Clustered column showing changes in dating behaviour since the 1980s, and in particular the rise of online
A slope chart makes the change in dating behaviour more obvious - online dating has risen from 0% to 19% (the top option) between the 1980s and 2010s
Subdividing amoebas - on a linear scale. The first few days are lost as the numbers increase.
The exponential rise is seen as a straight line on a log scale
The kink on day 9 is invisible on a linear scale
On a log scale, the fact that the amoebas were quiet on day 9 is clearly visible
Actual v expected lines can get fuzzy on a linear scale
Actual and expected lines on a log scale are clearer when you are charting exponential rises
Targets vanish on a linear scale
With exponentially rising data, log scales allow targets to be clearly visible
Research results on a linear scale. It is clear that the majority cannot understand log scales.
Research results on a log scale. This closes the gap between those who cannot understand log scales and those who cannot understand linear scales.
A confusing log scale chart
A linear scale showing the same data - the changing numbers can be clearly observed
On a linear scale, the bars before November 1923 are almost invisible.
On a log scale, there is no obvious relationship between bar and number.
A badly-organised table, with too many borders and misaligned text
A clear table, with a clean layout. Easier to interpret than our log scale chart.
Sometimes showing the rate of inflation can make for a more readable chart
Here we've converted the numbers to an inflation rate to tell the story of Zimbabwean hyperinflation
If your outlier is huge, show it breaking the frame
Bubbles are great at outlier stories
Use analogies to make the story understandable. Here we say that a stack of 201 billion mark coins would (almost) reach the moon
Here we say that a pathway of 201 billion mark coins would stretch around the world 100 times.
A photo showing a German using banknotes as wallpaper
A photo showing German children using banknotes like Lego
Line chart showing the increase in people who don't mind what gender their boss is.
Line chart showing that most women in the UK now have driving licenses.
Here lines show the distance covered by various animals in an hour
Here lines show the distance between two islands
Lines here are used to show the difference between two positions (the UK v the US and their ability to beat up different animals)
Here lines are used to show the difference (or not) between left-wing and right-wing positions on various subjects
Here a line is incorrectly used to link distinct categories of 'brand value'
Here the line is replaced by distinct bars for distinct categories
Here a line is inappropriately used to show differences in behaviour between age groups
Here the line suggests change over time, when actually we are looking at changing tastes between demographic groups
A bar chart showing the percentage of female drivers in each age bracket
A series of bar charts showing how different age groups prefer different ingredients
Frequency polygon showing the distribution of ages for different genders in Hollywood films
A jittered strip plot showing the distribution of actors by gender and age in key romantic films
prev / next