Red & Green Color Scales? Maybe for Christmas

  

Tis the season!

I’m as guilty as the next guy. When I first started out building data visualizations and dashboards I was quick to select a Red-Green color scale to differentiate between good and bad. In my defense, in the United States green is often associated with good results, red with bad results. Additionally, it didn’t help that many of the available tools defaulted to the red-green scale as the vendors weren’t focused on sound data visualization practices.

So what’s the drawback of using the red-green color scale?

First, color meaning varies greatly between cultures. i.e. Red typically  symbolizes “bad” in western cultures…in China red symbolizes good luck, happiness and long life…all positives. As for green, western cultures often associate green with “good”, whereas North African and South Pacific cultures associate green with corruption and danger respectively.

For more information on color meaning by culture, try a site like ColorMatters. There’s a lot of resources that touch on this topic. We encourage you to leave comments on the post if you have other examples to share.

So, beyond the varying meanings from culture to culture, why else would the red-green color scale be less than optimal?

If you’ve studied visual perception and design, you’ve probably come across the subject of color blindness. Two common forms of color blindness, Deuteranope (aka  Green blindness) and Protanope (aka Red blindness), will essentially render the red-green color scale useless for those affected by either.  Below are examples of what these users would see:

Normal Vision:

Deuteranope (Green blindness):

Protanope (Red blindness):

Our suggestion: use an orange-blue scale. While the colors will still appear different to the color blind, the hues will still provide enough differentiation for analysis. It may take some explanation up front, but consistency on this will lead to quicker processing.
 

Cheers,

Kevin A. Taylor 

Data Sushi: Shimmy Shimmy Ya

In the infamous words of the late O.D.B., (I will resort only to the abbreviation since this is an “all-ages” blog) in his collaboration with Mariah Carey…

“Ooh Baby I Like It Raw”

Hold on…We’re talking about data.

Raw Data vs Cooked Data

All too often we get requests to visualize data that has exactly 1 layer of granularity…meaning we can’t drill into or interact with the data. We’re left with only a static visualization. That might have been acceptable before, when we were limited to print media and power point slide decks (even PowerPoint can include interactivity these days).

But today, in the digital age, we have the ability to explore and interact with data on the fly…changing the conversation both in media as well as the corporate meeting room. But to do this, we MUST provide the raw data! It’s one thing to generate new questions; it’s another to be able to answer them.

Last February I was blessed with the opportunity to attend Tapestry 2014 in Annapolis, MD. One of the presentations I enjoyed the most was “Jock Dreams of Data Sushi,” delivered by visualization expert Jock Mackinlay.

 Jock Dreams of Data Sushi

Data Sushi, as Jock describes it, is “a visualization that is beautiful on the outside, and contains RAW data on the inside”.

Below are the reasons Jock gives for why we should demand raw data when possible followed by the excuses we hear for not doing so:

Reasons:

1. Increases Dwell Time
2. Validate the Author / Designer / Developer
3. Encourages Conversation

Excuses:

1. People Have Cooked Data
2. Too Valuable to Share the Data
3. Hard to Share Raw Data

I encourage to you to watch Jock’s 15 minute presentation as well as to promote raw data behind your visualizations.

Cheers,

Kevin A. Taylor

Can Data Visualization Save Lives?

If you’re reading this blog, you’ve probably already heard plenty about the ever expanding presence of data visualization and infographics. Not just on dashboards and in board rooms, but out in the real world, on billboards, on fast food cups, virtually everywhere.

Beyond helping to bring some clarity around KPIs and Corporate Metrics, data, and specifically data visualization, is being leveraged in every facet of life. Things like “data for good” where visualization is being used to help resolve world hunger.

But what about in our everyday lives…at home? Can visualizing data make a difference?

I’m sure we could come up with an extensive list of examples, but one I want to talk about today is weather forecasting, especially as it relates to storm tracking.

It seems like just yesterday that when it came to reporting tornadoes, the news was always reactionary. You only knew a tornado was in your area if one had been reported from the ground. The obvious downside to this was there was no warning system. People were often blindsided. No longer is this the case.

Take a look at the following radar map:

This one shows a fairly strong storm system with a pattern in the bottom left (“The Hook”) that is often an indicator that a funnel cloud has formed,  or could be forming. Although these pictures are older, the same pattern was shown recently as storms passed through North Carolina. When this pattern is spotted, our local weatherman, Greg Fischel, will drill down into the following radar visualization that shows wind direction.

Here, the Red shows wind moving away from the radar satellite and the green shows wind moving towards the satellite. As the two wrap around each other, it’s a tell-tale sign that a funnel cloud has formed. Greg will typically zoom in on the edges to get the wind speeds and then provide some annotation (arrows) as he has done here to help the casual viewer understand exactly what is going on.

Having this data allows the local news to broadcast warnings to individual subdivisions, providing predictions of when a tornado is likely to hit your precise location with accuracy to the minute.

So if anyone asks you if data visualization has ever saved lives, the answer should be YES.

This is just one example of the power of data visualization. I’d love to hear other examples that you have come across.

Cheers,

Kevin Taylor

"Age Stickers" - Data Visualization Should be Unbiased

What I’m about to write about should be considered by confession of geekness. I’m perfectly fine with that. The sad part is my wife now knows just how big of a geek I am too.

In the “real world” it’s often easy to find examples of poor data visualization design. It’s a bit more rare to find a really good example.

Sitting on the couch one weekend, watching TV with my wife, I saw a Prudential Retirement commercial. (yes a commercial, they still exist in the DVR / NetFlix era). Essentially, people are asked to place a sticker on a chart to mark the age of the oldest person they’ve ever known. By the end of the commercial you are made to believe that people are living longer so you should invest more in your retirement.

Here’s a link to the 1-min commercial:  http://www.ispot.tv/ad/7IhP/prudential-age-stickers

I literally got up and shouted “I LOVE IT” and my wife looked at me like I had 6 heads. Oddly, I was going to write about how good of an example it was until I sat down to write. When I did, I started to break down the visual and the story it was telling and realized how biased it actually is.

What if the study had asked the same people if they knew someone that has passed away between the ages of 50 and 65 and plotted those stickers? Would you still want to increase your retirement investment if it were likely you wouldn’t reach retirement age?

The fact is, you can make a chart tell just about any story you want by using partial data. In order to represent the data accurately, you should include all the data.

In this case, a more responsible way to visualize this data might have been to show average (or median) age of death(based on all death data) over time. If the line is trending up, you have your unbiased visual to market. If it’s trending down, don’t twist the data to tell otherwise. Although this is marketing and honesty doesn't necessarily sell investment packages.

Cheers,

Kevin Taylor

Do Not Taste the Rainbow

Like every other pixel of ink on a visualization or dashboard, colors added or colors omitted should be done so consciously.

Unless you’re a delicious fruit flavored candy, there’s no need to exploit the rainbow. Random, over use of color can lead to confusion and mental exhaustion for your users. If you’ve ever been to Las Vegas you have an idea of what I’m talking about…sensory overload is a real thing!

Yet dashboard after dashboard, we see color used as if the designer is paid by the # of colors used or as if they were on Crayola's payroll. Examples can be found all over the web.

Here are 10 guidelines for using (or not using ) color when designing your visualizations. These are considered best practices and not necessarily laws.

1. Use color to make things POP. If everything has a color, nothing pops. Everything seems of similar value and the user isn’t sure where to focus attention
2. Use muted colors for categories instead of bright bold colors.
3. Use low saturation colors for large areas and high saturation colors for small areas.
4. Be consistent with color selections. i.e. if Sales is blue on one visualization, make sales blue on all visualizations
5. Do not use color by entity for a bar chart that already has labels as the colors do not add value:

Bad:

Better

6.  When comparing one entity (i.e. mocha) against other entities (i.e. dark roast, decaf, etc). make the entity of interest a predominant color and mute the others (i.e. use gray)

7.  If using color for dimensions, think about semantics (i.e. grape = purple, cherry = red, lemon = yellow). Not doing this can severely slow or potentially alter an analysis.

Bad:

Better:

8. Similarly, if you are using company colors match them perfectly. The details make the difference. Don’t attempt to eye it. Use a tool like “Get Color” to obtain RGB values.

9.  Limit your dashboard to a single color palette when possible.

10. Be considerate of your color blind audience. Red-Green is the most common, followed by Blue-Yellow. Consider using a color blind palette like the one created by Maureen Stone for Tableau Software:

These are just 10 tips that came to mind. Please leave other tips in the comment section below!

Cheers! 

Kevin Taylor

Do You See What I See?

One of the most important steps in designing visualizations…and one that is all too often ignored entirely…is to make sure that our visualizations deliver the message they were intended to deliver.

We tend to take for granted that our beautiful creations are telling the story we aimed to tell. Even with written words we seem more prone to seek out confirmation. For example, have you ever wanted to fire off an email while on an “emotional peak” but instead you run the email by a confidant to make sure the message is clear and rational?

We should do this with all of our data visualization and dashboard designs.

It’s not enough to rely on our own interpretations because in reality, as the designer our interpretations are often jaded.

We tend to believe the message is clear because we know all the nuts and bolts of the chart or graph…because we designed it. Since we understand the data and since we know the intended message, we can no longer view it from the eye of a “first-time-viewer”.

Additionally, we may invest a significant amount of time and resources to get a particular view created. At this point we tend to build an emotional tie to the work and effort and are less likely to discredit the visualization, even if it doesn’t tell the story in the best manner.

At the end of the day, the main purpose of any visualization is to inform. When we inform, we want to do so in a clear manner that can be understood by our audience.

Ultimately we should aim for understanding that can be achieved without a need for human instruction. When this can’t be accomplished we need to provide the appropriate context to allow our visualizations to stand on their own. This can be done through instructions, annotations, intuitive titles and other features in your dataviz tools.

Just be careful not to take away from the visualizations. Instructions can be suppressed so they don’t take up valuable real estate and so you don’t have to view them once you know how a view works(i.e instructions on demand). Consider Tufte’s Data:Ink Ratio when adding annotations.

But most importantly, ask a co-worker or a friend or whoever, just ask someone else to validate your intended message. Ask them what message they receive rather than stating your intended message and asking for confirmation.

Cheers,

  

Kevin Taylor

Blog About the Blog

Ah yes, my first public blog post. Better not screw it up!

So this 1^st post is simply a blog about my blog. For those of you in the corporate world, think of it as a meeting about a meeting…just try not to dread it as such.

Most of my topics tend to be basic and have most likely been written about before. I will do my best to give credit to the pioneers in this field who have equipped me with this knowledge and have paved the path for my own personal success. There are so many great authors and bloggers in this dynamic industry and I encourage you to seek them out and absorb their thoughts with great passion. I’d try to name them all but I know I’d fail and many of these folks have become a bit like family so I don’t want to risk that.

So what can you expect to find when you follow my blog?

Here’s a rough list of guidelines. Like the guidelines in visualization, they are not laws and I will apologize now, for they will be broken on occasion.

I Will:

•         Post at least 2 blogs every calendar month (will shoot for 4)
•         Keep blogs to 400 words or less (in appreciation of your time)
•         Write with candor and request you do the same in return when leaving comments
•         Write in compliance with what the pioneers before me have laid out as visual best practices
•         Keep my content applicable for beginners and aspiring experts alike
•         Convey my passion for this subject
•         Write about Color, Chart Types, Visualization as a Process, Visual Analytics, Chart Types,     Dashboard Design, Tools, etc.

Visualization, or Visualisation for my friends across the pond, is not exactly rocket science but the devil is in the details. The commonly agreed upon tips, techniques & best practices we’ll discuss can be a differentiator between amateur results and professional results…and ultimately the difference between conveying your message clearly or not. In the end “Seeing is Believing”.

As I am new to the blogosphere, I welcome any suggestions you may have! Anything from subject matter you might want me to write about to functions you see on other blogs that are missing from mine.

And finally, my blogs are merely my opinion. They do not necessarily reflect the opinions of my past or present employers.

I hope that these posts will serve primarily as a spring board to deeper 2-way conversations.

P.S. I already broke the 400 word “rule”

Please feel free to reach out to me directly or simply follow me on Twitter @KevinTaylorNC

Will post 1st true content on Thursday morning (10/14).

Cheers!

Kevin A. Taylor