Don’t you hate dodgy line graphs? Line graphs draw the reader’s eye along, joining the dots. That’s good for trends; going up, …. and going down. Brilliant for indicating change over time. Sadly it’s all too common to see dots joined inappropriately. And here’s an example…..
With truth, lies and weirdness, these charts all seek the truth. Is there one graph to rule them all?
Here’s what I see as the context:
a). We should expect politicians to tell the truth.
b). Telling untruths should be a fail.
c). Pants-on-fire lies are a huge fail.
d). I don’t assume because someone tells lies about, say, healthcare that they lie about, say, border control.
e). The goal is to consider each candidate and to compare all 5.
So how could this data be presented clearly?
For any one candidate,a pie chart does a nice job of indicating the share of ‘true’ versus ‘pants on fire’ statements. That’s what pie charts are good for; bit of pie adding up to a total.
Did you look at each graph in turn? And then; attempt the visual leap of imagination to compare candidates? Maybe you attempted to ‘place’ one pie on top of another. That’s hard work.
Another way the pie charts lose is the amount of horizontal screen space on the desktop. They pretty much illegible on smartphones. There’s a better way.
Using a table allows you to get a lot of data on screen in a small area. But numbers alone can be hard to read. Conditional formatting allow each cell to be coloured (colored) automatically depending on the cell value. With the candidates / conditional formatting in rows, you can quickly see where each candidate leans.
The weirdness here is that ‘pants on fire’, and all big liars gets green. Both true and lies are equally green/ok. If you want to compare across candidates, the shades of green are misleading; you are back to reading numbers.
We want to go beyond just individual performance.
To focus just on category winners, you might opt for a conditional formatted data bar (below left).
Presenting all candidates / statements results in the dizzy clustered column on the right. Busy, busy. Okay for a caffine high perhaps. I also find it plain weird that ‘false’ gets a nice tall column reaching for the sky.
I like what Avinash did with the category winners again using Excel (automatic) conditional formatting on a table. This time the colour scale goes from green through yellow to red.
This is an example of how we include our values in data. For me in this format it would have been more intuitives to split the categories 50:50. Half good/green and half bad/red. This would mean;
more ‘half true’ statements > less half true statements
The weirdness is the top percentages are green in two cases, and red in others. (It was really tempting to say; “that’s pants”). Two statements, ‘true’ and ‘half true’ use a higher percentage/greener as better. e.g. true. The other 4 statements use the inverse: a higher percentage is bad/red e.g. pants on fire.
Do you get a good sense of the individual candidate? How did the mental gymnastics of reading the numbers without ‘seeing’ the colours. e.g. Rubio’s has the most half true statements; 22% / deep red, second only to 24%/orange mostly false.
My experience in sharing conditional formatted tables is; you will need to explain them. When you’re used to them, they’re super and fast. Get used to them if you do analysis.
Great Chart Expectations
To summarize our expectation for this chart:
i). Green is good. Greener is better. Red is bad.
Colour (or if you’re American ‘color’) choice should align to cultural expectations. If you’re familar with the psychology of colour so much the better.
For the colour blind, the difference between the red and green I choose was invisible. I have a ‘test candidate’ for this at home who was unimpressed with my version. Estimates indicate this impacts around 8% of men (in whom it’s more common) and about 1/200 women. (In my extended family ‘test population’ all are men.) Here’s are the colours from my graph before I fixed it:
For the millions with red/green colourblindness the colour constrast looks something like this:
ii). Avoid weirdness; don’t mix your coding within one data visualization.
iii). More truth is better, and better than lies i.e. sort first by ‘true’.
iv). Less pants-on-fire > more pants-on-fire. After truth, sort by least ‘pants-on-fire’.
v). Lies aren’t neutral, they’re a negative i.e. below the line of acceptable.
This can be represented by multipling the lies by minus 1 so they’ll appear below the midway line.
vi). To give equal visual weight to truth versus BS, the axis should be formatted equally above and below zero. Without this manual step a larger portion of the graph gets allocated to lies.
N.B. I’m inflicting my expectations/values here.
vii). Align your data presented in charts with cultural norms. In this case left-right aligns with reading norms. And the top right hand corner of a graph is usually the ‘best quadrant’.
Truth & Pants Compared – The Data Visualization
So here is it. How does the ‘data’ leave you feeling?
Here’s what it did for me:
Clinton best on truth
Not much difference between Clinton and Sanders.
Compared to truth tellers, the liars are off the chart. Sad.
With 20 years high-tech marketing & product development experience from Boston to Billund, Berlin to Bangalore, Jane has managed teams and tech products with millions of installs, and millions of revenue (annually).
She's researched and developed marketing strategy for global markets, and established the blueprint for product & marketing strategy in many new teams.
As an intrapreneur turned entrepreneur, she changed vowels in 2014 and founded JEM 9 Marketing Consultancy. Today she works with CEOs, founders & marketing leaders to assist them in understanding and reaching customers.
Speaker on market research and product management.