Here are my notes:
DESIGNING PERSUASIVE CHARTS with Scott Berinato
"People read charts like they read books" Scott said. There are a lot of things you can't control and information is read in the order it was presented, thus making building charts difficult at times. People also naturally gravitate toward things that stand out, like colours and outliers, and almost immediately start to form narratives.
He used 5 examples to talk about misleading charts:
1. Ideas that Don't Exist
This chart, presented in congress, shows as if "abortions have risen above cancer screening", Scott said, calling it "a deliberate attempt to mislead".
Personally, I always go back to the statistics rule of "correlation does not imply causation".
2. Look at Axes Labels
Scott used this graph to illustrate how deceiving a cumulative bar chart can be, showing growth when there was none.
In fact, when you separate out the revenue individually, there is a decline.
Basically, this is a very ill-suited chart for the message.
3. Pay Attention to the Spacing
Scott argues that perhaps there are no truly objective charts, but rather, each serves it's own purpose.
 |
| Wide spacing between "Years" |
When a chart has a much wider spacing, the fluctuation of the line does not appear as drastic as if the same chart had a much narrower spacing.
 |
| Narrow spacing between "Years" |
4. Truncated Y Axis
Truncated Y axes create a more dramatic story, which sometimes could be misleading. This chart looks as if the average job satisfaction really plummets throughout an employee's career.
However, if the entire Y axis is shown, the decrease looks unremarkable.
Scott said that some scientists may look at very limited ranges of data where truncating the Y axis becomes appropriate. There are no hard rules, just think about whether you are exaggerating the story unnecessarily.
5. Dual (Y) Axes
Dual axes charts measure 2 data points in the same visual space.
First of all, although we're looking at care sales between Tesla and other brands, the 2 charts have completely different units (one in percentage increase, the other in dollar increase). Then, when looking at the green line, proportionally, it looks as if Tesla shares are projected to increase 25% (a quarter of the chart) when in reality, it will only increased about 2% (Y axis on the left does not contain the entire 100%).
Since we're looking at Tesla vehicle sales compared to other vehicle sales, Scott thinks this chart is a more appropriate representation.
Q: Common Decision Points?
This depends on the data you choose to show. For example, the following graph shows the sales of vinyl records between 1993 and 2014. It appears, quite justly so, that the sales of vinyl records have "sky-rocketed".
However, if you start the graph in 1973, then you'll see that the "peak" is not a peak at all.
Scott then compared the sales of vinyl with the sales of other physical/digital/streaming album sales, and the proportions becomes apparent.
Q: How do you know when you've crossed the line?
Use the golden rule, and ask yourself whether you feel deceived or mislead by the chart. When choosing the right representation, ask yourself if you are "zooming in on the message or are you distorting the truth".
Q: How do you know charts are accurate?
Evaluate all the ways charts can be misleading. For example, pay attention to whether the Y axes are truncated and the story is in fact more dramatic than it really is. Or when encountering a dual axes chart, analyze the data individually / separately first before comparing the 2 together.
See Scott's book Good Charts: The HBR Guide to Making Smarter, More Persuasive Data Visualizations for more info.
