There is the cliché. There are lies, damned lies and statistics.
Statistics can be used to mislead. For example, "did you know that the number of employees leaving Engineering has gone up 50% in the last year?” A concerning statistic, indeed. Unless your Engineering department has doubled in size in the same period. Or maybe two people left last year, this year it was three. And two were people retiring.
I think the reason “statistics” gets a bad rap in the cliché is how the lie is delivered. It’s a form of misdirection and misleading the audience. You present some data that is otherwise factually correct, but invite your audience to draw their own opinion. Naturally the selection of the data leads to a false conclusion.
Now, “We have a turnover problem in Engineering”. That could be an outright lie. Despite the lie being ultimately the same, misleading to lie is somehow worse. It draws you in to participate in the lie. It’s sneaky.
This brings me to visualization. In my view, visualization has just as much power to mislead - perhaps much more.
You can use visualization to misrepresent. A simple example is shown above. The proportionality of the 71% is “incorrect” in that it’s not in line with the base data. You can use size and volume to indicate significance. Then to jazz it up, you put the 71% in bold and give it an appealing color.
Some examples are quite innocent. People may use Pie Charts without realizing there are issues with interpretation. Others are less so. A “favorite” of mine was a consultant who used to rank software products. They loved to use radar charts to show the different evaluation dimensions of the available products. The thing with a radar chart is you can manipulate the relative volume of the "stars" by changing the order of the spokes. Completely cynically, they’d play with the order till it “looked right”.
It’s a double-edged sword. Even these simple representations can be powerful, so they can be used and misused. 71% may really be significant in context. If you’re talking a 71% improvement in successful child vaccinations in a the 3rd World, well that may well be something you want to shout about (Keep up the good work!).
However, you can use a different representation. You could present the 71%. That might represent 413 children. You could then draw the parallel that this is the average size of a US Primary School. This lets people take the numbers (71 and 413) and contextualize it into a representation they understand.
At this stage you’re wandering from Data Visualization in the domain of the Infographic. Purist will argue there is a difference and there certainly is. However, both Data Visualization and Infographics take data and apply a visual transformation. They both rely on the fact that we process graphical information differently - in many respects more powerfully. Both use visual context to communicate.
Definitions abound, but I think of an Infographic as a story. It takes the audience through a structured graphical sequence to convey this story. Data Visualization has a different expressed goal. But, let’s face it, it has a similar capability.
If you are producing a Data Visualization, it’s important to think of the stories it will tell. You might want your visualization to be an impartial presentation of data. However, by definition and intent you want your visualization to convey something.
There are key questions to ask - What is in the mind of your audience when they see it? What are they seeking? How do you best draw attention to what is “important” or otherwise significant? Are there any inherent biases that you need to compensate for? What best informs their next decision or move?
This was front-of-mind for me when I was working on the designs for Culture Amp. The whole point of Culture Amp was that extracting employee data into something digestible is difficult. We wanted something that gave you the data visualizations you needed. Something that just worked and told you what you need. Out-of-the-box.
During this process I wanted the visualizations to borrow some ideas from Infographics. I wanted the consumption of the data to be a conversation. Start with the qualifying and headline information first, and then use cues and visualization to guide (but not dictate) how users then dug into and consumed the data.
Striking this balance was really important. We wanted to move away from matrices full of bright red and green numbers, where the detail was completely obscured. While it was a conversation, I wanted it to be a conversation the user got to steer. It’s taken some experimentation, but I think we’ve ended up with the right formula.