Update: It appears that this chart and other data visualizations have been removed from the website and report. I’m hoping that means that the authors will be refactoring them with improved graphics. Meanwhile, I’m going to leave my post below as is. There are good lessons and tips to be shared.
I know. I hear you. It’s still January and we might just have a winner, one that will be impossible to beat during the next 12 months. Incredible. As you may recall, in late 2011 I awarded Stupidest Bar Chart to a doozy from Klout. That bar chart was confusing, but not in the way this one is. First, put down your beverage of choice. Then take a look at this:
Yeah. That…chart. It’s kind of like a horizontal stacked bar chart. I don’t understand anything about it, though. This chart comes from an infographic at Deloitte.com on Analysis Trends for 2014.
Maybe zooming in might help?
Nope, doesn’t make it any clearer. In fact, it’s just as crazy, but bigger. Call it Big Crazy DataTM.
Here are the issues and questions I have about it:
- What do the colours mean? If this were a stacked bar chart, the grey and blue areas would indicate different data. It appears that only some sections have data. But I’m not sure.
- What is the scale? Normally a bar chart would have an axis that indicates some measure and all the bars would be graphed against that axis. This has no axis.
- Why do some bars have signed numbers and one have a range? Why are some numbers unsigned? Even some delta numbers are unsigned.
- What do the relative sizes of the sections mean? In one bar we see a blue section labeled 285, but it’s larger than a section labeled 425-475.
- Where numbers appear, do they describe the section they are on or the section next to the number? I’m not sure
- What does the relative position of the blue section mean? I’m not sure.
- Why are some of the labels in light grey and some in dark grey? I’m not sure
- What are the units of measurement for these numbers? Are some percentages? Units of 1000s? 100,000s? Are they of people? Positions? Something else? I’m not sure.
- Do the endnotes there explain the numbers? No, they are just citations for reference materials used to create the report.
Maybe the chart has an explanation inside the full document, Analytics Trends 2014: (And why some may not materialize)… No, same chart, no text that directly explains any of the numbers. To add some irony to this, the report itself is about Analytics and even covers trends in visualizations.
A Picture is Worth A Thousand Words, Unfortunately.
The report has something to say about data visualizations used in data analytics:
There’s no question that visualization has become a critical capability for organizations of virtually every shape and size. Easy-to-use software makes complex data accessible and understandable for almost any business user. From discovery and visual exploration to pattern and relationship identification, today’s visualization tools easily affirm the adage that a picture is worth a thousand words. Or, in this case, numbers.
This is especially true with big data, where visualization may even be a necessary capability for driving insights. That’s why visually oriented tools are rising in prominence for many big data applications. Users get to understand, explore, share, and apply data efficiently and collaboratively—often without the need for analytics professionals. And that’s where the risk comes in. In their eagerness to dive into data, users may choose polished graphics over thorough data preparation and normalization and rigorous analysis—glossing over important insights and analysis opportunities and potentially producing erroneous results. [emphasis mine]
Keep reading the report from that section. The irony burns.
What’s Going on with this Bar Chart?
I’d bet that the Analytics professionals at Deloitte know much better than this. The webpage and report for Analytics trends is beautiful to look at. I’m guessing that a graphics designer has taken these numbers and created a beautiful, yet meaningless graphic with numbers. And just as the report predicts, people who don’t understand how to best use visualizations can gloss over important insights and analysis opportunities and potentially produce erroneous results. This report has some great points. And it’s pretty. Very, very pretty. But the distraction of bad visualizations makes difficult for me to actually see the points the authors are trying to make.
My guess is also that this data, as a set, had no business being put together in one chart. I’m not sure, but they don’t seem to have the same measures or even be the same type of data. So putting them in one chart won’t help. This was a page in a report needing a graphic, so someone made one.
Jamie Calder ( @jamiecalder) helped me “see” the story this chart is trying to tell: think of it as a math equation. That might get you there. But it’s still not an appropriate use of a bar chart. And Josh Fennessy ( @joshuafennessy) has pointed out that this isn’t supposed to be a bar chart at all. It’s supposed to be a waterfall chart. But it’s dressed up as a bar chart, so I’m going to still leave as a contender for Worst Bar Chart of 2014. Let’s just call it a self-nominated chart. Martin Ribunal has found what is most likely the original chart from which this chart was most likely
copied inspired by and has listed that in comments below.
What Have We Learned About Data Visualizations?
- The best data analysis can be invalidated with bad data visualizations.
- If you develop content, insist that you say in the final published work. I know this is difficult in large corporate entities, but it’s important to ensuring that your goals are met.
- The more accessible we make self-serve BI and data visualization tools available, the more responsibility we have to educate, train, and mentor those using these tools.
- Show your visualizations to other people. Ask them what they see. Ask them if they are confused, what conclusions they might have and what questions they still have.
- Choose the right chart type to fit your data. Then use that chart correctly.
- If you needs a graphic image, don’t mimic a recognized chart type.
- If you add a chart to a document, you should actual reference it in the text in the way that helps the reader understand it.
- If your chart has numbers, you have to say what those are number of, including some sort of unit of measure. And your graphics should correctly portray their relative size.
- If a chart leaves viewers saying “I’m not sure” more than once, it’s not working.
- Loving your data means loving how it is presented, too.
What Would You Ask?
What other questions do you have about this…graphic.? How would you improve it?
I can’t bring myself to call it a bar chart any more. But it’s still dressed as a bar chart, so it fits the nomination category. If you find a bar chart or any other data visualization to nominate, let me know. I wouldn’t want something worse than this one to go unrecognized.
It is a capital mistake to theorize before one has data.
-Sir Arthur Conan Doyle
Experts often possess more data than judgment.
Sir Doyle and General Powell seem to have conflicting points of view about data, but I’m not sure they do. I love my data and yet data alone won’t solve many problems. I have to figure out which data to use, how current the data needs to be, and how to use that data with other data and my own experiences and biases to get to a decision that’s right for right now.
We in the data profession pretty much spend our days trying to get quality data to the right people as quickly as we can. We provide analytic services in hopes that management can turn that data into good decisions. We add biases, we filter out biases, we support a lot of guessing. None that makes Sir Doyle or Gen. Powell wrong.
What if we provide data and analytics to organizations, but mostly management just makes guesses? I recently sat in a meeting where we were asked to keep adjusting the data rules until the analytics would show exactly what values the end users wanted to see. Of course this is a fine balance: end users need to set the requirements around how data should be processed to produce the analytical solutions, but at some point we data pros can’t get sucked into using decision systems to justify bad decisions.
I’ll leave you with a quote from Mr. Heinlein:
To get anywhere, or even live a long time, a man has to guess, and guess right, over and over again, without enough data for a logical answer.
–Robert A. Heinlein
That might be how a man should live, but organizations need to ensure they are working with good data and great analytics. At some point your competitors will stop flying by the seat of their pants via SWAGs. And they will most likely be making better decisions than your organization. If your analytics are there only to make end users feel better about their guesses, you’re doing it wrong.
Subscribe via E-mail
- September 2016
- August 2016
- June 2016
- May 2016
- April 2016
- March 2016
- February 2016
- January 2016
- December 2015
- November 2015
- September 2015
- July 2015
- June 2015
- May 2015
- April 2015
- March 2015
- February 2015
- January 2015
- December 2014
- November 2014
- October 2014
- August 2014
- July 2014
- June 2014
- May 2014
- April 2014
- March 2014
- February 2014
- January 2014
- December 2013
- November 2013
- October 2013
- September 2013
- August 2013
- July 2013
- June 2013
- May 2013
- April 2013
- March 2013
- February 2013
- January 2013
- December 2012
- November 2012
- October 2012
- September 2012
- August 2012
- July 2012
- June 2012
- May 2012
- April 2012
- March 2012
- February 2012
- January 2012
- December 2011
- November 2011
- October 2011
- September 2011
- August 2011
- July 2011
- June 2011
- May 2011
- April 2011
- March 2011
- February 2011
- January 2011
- December 2010
- November 2010
- September 2010
- August 2010
- July 2010
- February 2009