The Worst Bar Chart of 2014? I think not.
First, it’s not a bar chart. It’s a waterfall chart. Now, beyond that, there are lots of things ‘wrong’ with the implementation of it that I would like to address here.
Before I begin dissecting, a little explanation of a waterfall chart. At first glance, a waterfall chart can easily be mistaken for a bar chart. It has bars, and is a chart, but that’s where the similarity ends. Typically a waterfall chart is used to tell a story. A story of how data is moved from Data Point A to Data Point B.
A classic example is inventory. I give you, Exhibit A. Ok, this is an extremely simple example I made with MSPaint – the graphics tool of all power users – still, it should help to prove my point.
This chart, remember it’s not a bar chart, tells the story of HOW July’s ending Inventory Total made it to August’s Inventory Total. Green bars are increases, Red bars are decreases – data labels for clarity. At a glance, you can see if the store is receiving equivalent amounts of inventory as they are selling, etc. Pretty neat visualization I think.
That’s a waterfall chart in a nutshell, so let’s move on.
Looking at the example that Karen provided, I see a few things (or more) wrong with it. Here are the big points (and those easily fixable with MSPaint).
First, the colors. The blue portion of the chart is supposed to show movement. It doesn’t. Which direction does blue go in? One section it’s moving left, another, right. It’s also a starting point and it’s an ending point. Additionally it’s used to show the deficit created by the 2018 supply vs. demand. So, I propose modification #1 – new colors.
In this new example, green signifies an increase, red a decrease. The ending point of the red ‘bar’ is how we arrive at the Projected 2018 Supply starting at the 2008 Employment. (156 + 161 – 32)
Second, the data labels. In this example, the data labels are not related to the movement of the data. All of the labels are oriented towards the right terminator of the bar, but the right terminator is not always the ending position of the bars movement. Some quick rearranging of the data labels will help to make it a bit clearer that where each bar starts, and where it ends.
Finally, the last ‘bar’ on the chart doesn’t follow the same format as the others. It’s not showing movement. It’s showing the portion of the total that is a deficit. Additionally, the data labels here DON’T EXPLAIN THE DEFICIT BUT RATHER THE TOTAL. This is against so many good rules of visualization it hurts my head just thinking about it. To fix this final bar we would want to modify colors, change the data label to a scale (showing 475 being the total) and then highlight the deficit with the correct number in question as a data label.
There are additional confusions created by alternating the row label color (why?), but again, not easily solved with MSPaint, so just imagine how great it looks when labels are all one color.
Here’s how I would have visualized a waterfall chart using this data. What do you think? More or less clear than the original version?
In specific rebuttal to Karen’s post, do I think this is the ‘Worst Bar Chart of 2014?’ No. Mostly because it’s not a bar chart. Still, it’s a pretty awful chart and while I’ve shown a few ways it can be better it took WAY to much explanation for it to be an effective visualization. If you have to explain your chart in words, then why visualize it at all?
To reiterate one of Karen’s most impactful closing points “If your chart leaves viewers thinking ‘I’m not sure’ more than once, it’s not effective.” If you are in the business of publishing visualizations to the public, please show you charts to someone who doesn’t know the data as well as yourself. Their response will be good litmus test of whether your chart will be effective out in the wild.