Bad Visualisations

A collection of weird graphics and visualisations from here, there and everywhere. Don't follow if you love pie charts and are unwilling to change your mind about them

[badvisualisations]

The graphic adds no value. To get any sort of information from this you need to read the (rather dense) text associated with each graphic. 

It’s another case of a graphic designer trying to be too clever and destroying the information in the process. If the use of the same shapes in different sizes for different bars is bad enough, here the designer has tried to use different shapes for each city. So there is no way at all that the size of any building says anything about the quantity it is trying to depict.

Also there are two quantities being represented here (runs and average), and it’s not clear what the bar is trying to show.

Now I’m starting to wonder if this is intended to be a graphic at all. 

Source: https://www.espncricinfo.com/story/_/id/27358389/which-top-cricket-city-win-world-cup

[badvisualisations]

This is from Britannia’s investor presentation, sent to me by reader Varun Shenoy.

Graph is all good, but where are the numbers? It shows that market share has increased, but without telling you what the number has increased from and what it has moved to, it’s incredibly hard to trust the data. 

Also - saying “largest competitor” obscures things a bit. It seems like there’s more than one competitor, and different competitors are the largest competitors at different points in time. Presenting each of the major competitors’ market shares as a line (with numbers / axes ) would have enhanced the graph! 

[badvisualisations]

Pie charts are bad, but they are at least okay if you’re showing parts of something that add up to 1. 

I really don’t know what the graphics designer here was thinking - maybe following a “rule” that if percentages need to be shown, a pie needs to be used. 

Needless to say, the data doesn’t add up to 100%. The sectors of the pie don’t represent mutually exclusive and collectively exhaustive parts of anything. This is absolute nonsense. 

Source: https://www.espncricinfo.com/story/_/id/27143430/kane-williamson-hand-steadies-new-zealand-ship

[badvisualisations]

I spent a few minutes trying to figure out what the Y axis is here, and so far have no answers. Why would you distribute the points across two axes if the information is contained only in one? 

Source: https://wellcome.ac.uk/sites/default/files/wellcome-global-monitor-2018.pdf

[badvisualisations]

This is a screengrab from Star Sports. 

This “graphic” just overwhelms us with numbers without giving us any information. I mean - what is the relevant data point here and what message is the broadcaster trying to convey? Couldn’t they have simply picked the relevant measure (either number of sixes or number of sixes per game) and done a simple line plot to show the trend? 

Also, issues with the data quality exist - there was no World Cup in 1995, bu there was one in 1996. 

HT: Siddhartha Vaidyanathan

[badvisualisations]

This is a great example of “data mining”, where you look for some pattern that fits the pattern that you’re seeing now, so that you can extrapolate to make some random prediction. 

And if you look at the two lines that have been stretched and overlaid to make it seem like they overlap, they haven’t even done a good job of that - it’s hard to see how the two lines are similar. 

Now leave the bad data aside and let’s stick to the visualisation - What’s with the background? The shadows of the lines? The overall UX?The selective axes labels (0-10-30-50 on the left) and so on. 

Source: https://www.cnbc.com/2019/06/12/tesla-looks-like-netflix-did-in-2011and-it-may-see-a-similar-recovery.html

[badvisualisations]

The representation using a choropleth isn’t bad, but the reason I’m putting this here is the choice of intervals. I understand it is the default algo of datawrapper, but why would you draw boundaries at 15.4% and 33.5%? There is no regularity to this.

And when you use intervals like this, the choice of intervals can make a huge difference to the message in the graph. 

A better way would be to use a continuous scale, and do away with the intervals altogether. That would take out the arbitrariness, and help readers compare states better

Source: https://www.livemint.com/news/india/india-s-workforce-is-masculinising-rapidly-1560150389726.html

[badvisualisations]

Cricinfo at it again. When will people learn that using non-standard figures in place of bars actually hides information? 

Minor points:

1. The scale on the left is useless

2. Not clear what the number in the red circle is. 

3. Each batsman is playing a different shot. Again distracts from the graph

Source: http://www.espncricinfo.com/story/_/id/26927142/how-india-beat-australia

[badvisualisations]

1. Why would you stack bars with quantities that have different units? 

2. In the light of 1, this is minor but, why would you choose near-identical colours two quantities you’re stacking? Whose father what goes in picking contrasting colours? 

[badvisualisations]

If we could see past the colour scheme, this might be a good graph, but the choice of colour scheme makes it impossible to understand. Also the graph itself is not clear on what the Y axis is, and we need to perhaps turn to the accompanying article to figure that. 

The graph itself is simple - showing some measure of voting among men and women in India across time. But that atrocious colour scheme means it’s virtually impossible to see what’s happening. 

What prevents people from using simple graphs? 

[badvisualisations]

Check out the X axis. It’s really hard to find a pattern in terms of the scale that has been used. 1 and 1600 are as far apart as 1600 and 1913, and 1935 and 1953. It’s insanity

I know what has happened - some bright analyst at BankAm Merrill Lynch made this in Excel. And instead of making an area chart (which is what it should be) they made some sort of a bar graph. So all the X axis data points they had got spaced evenly. And so resulted in absurdities like you see on this graph. 

Source: Bank of America Merrill Lynch

[badvisualisations]

1. As a rule, the Y axis of a bar graph needs to start at zero. This violates that. Makes it seem like Indian women are one-fifth the size of Latvian women or something. 

2. Why are the figures scaled on the X-axis as well? Adds no information

3. Generally not a fan of using random figures being used in place of bars. what’s wrong with bars?

Source: Several people on twitter sent this to me

[badvisualisations]

This is not all. You need to click through to this article on BBC, where you will find a few other such pictures. They use five pictures to convey the sort of information that one simple bar graph could have shown. And in dispersing this information across five graphs, they make it absolutely impossible to compare across these graphs.

Source: http://www.bbc.com/capital/story/20170530-the-avocado-toast-index-how-many-breakfasts-to-buy-a-house?ocid=ww.social.link.twitter

[badvisualisations]

I cannot even. 

1. Quantities are neither mutually exclusive nor collectively exhaustive

2. The colour scheme! 

This is simply an atrocity. 

Source: http://www.thecricketmonthly.com/story/1181193

[badvisualisations]

Too many series on one plot here. And I don’t know why we have a mix of lines and bars - simple two-axis lines would’ve been good enough.

Also, the extra set of green bars to show growth is a bit confusing, especially since the high growth period happens at a time when the base was really really low. 

And then the X axis labelling is atrocious (I guess this is Excel default). You don’t need to label each tick. 

I would possibly draw this graph with a logarithmic y axis. and dispense with all the bars. 

Source: https://twitter.com/MohapatraHemant/status/1120212609045655552

[badvisualisations]

The problem is with the left set of graphs. When you see a set of bars, you assume they all represent the same “property”, with one bar for each category. Here, while all bars have the same dimensions, they all represent different quantities, and so this representation is misleading.

A sort of waterfall chart might have been better. 

The ordering of the bars doesn’t make sense as well - loss is defined as revenue minus expenses. So putting the loss bar in the middle doesn’t make sense. Also, expenses needs to be showed below the axis since it’s a negative number. 

I can go on but will stop

Source: https://the-ken.com/story/grofers-bet-on-value-customers/ (paywalled)

[badvisualisations]

I fail to understand why graphics designers still continue to use non-standard bars for their bar graphs. Without those legend numbers, it would be very hard to determine, for example, that de Villiers’s number is a bit more than twice of Russell’s. 

source: http://www.espncricinfo.com/story/_/id/26574308/no-complaints-andre-russell-batting-position-dinesh-karthik