Make your data speak for itself! Less is more (and people don’t read)

I’ve gotten some pretty horrible data visualization examples lately from an analyst company that shall remain nameless. So, I started educating my people on what good data visualization looks like, how size and subtle colors and position are excellent ways to convey information without screaming it at them. If you’re still using bright backgrounds on your graph — here’s looking at you, kid. You want a high data-ink ratio. That means that more ink (in most cases, digital ink) is used to convey data than for any other elements.

In case you’ve missed the age of information overload, just know that people won’t read through the text —(are you still there, reader? ;-D) — your graphs have to be rock-solid. They have to stand on their own — they can’t be open to interpretation. They can’t require that people read the accompanying text to understand. They shouldn’t detract from the information with lots of bright colors and bold fonts — use these elements sparingly! You can’t put the cognitive burden on the user — if you make understanding your research too hard, it certainly doesn’t make them want to invest in your company. How you convey information to people is at least as important as the research itself!, screams my inner tech writer and outer UX self (disclaimer: I was a tech writer for years :-D). After all, you’re trying to win people over with your data, not befuddle and confuzzle them!

Here are some of the highlights I’ll discuss in this article:

  • Using different colors wisely only when it adds information
  • Not adding chartjunk
  • Having a high data-ink ratio. Colored backgrounds detract from the data being conveyed
  • Using the right type of graph to convey the information
  • Understanding the meaning of particular colors, like red and green, and using them sparingly to indicate problems and successes, respectively.
  • Putting chart values as close as possible to the graph section it details. Avoiding a legend if at all possible.
  • Use element size, color saturation, and positioning to convey information

Here are some examples of bad data viz and better yet, how to fix them:

Overuse of color detracts from the actual data

Color should be used judiciously and only when using color conveys MORE information.

This chart has many things wrong with it:

  • Random use of color and having some categories the same color creates the perception that these categories are somehow related. This introduces cognitive burden as the user struggles to find a connection between for example, speed of product/service delivery and mobile strategy integration. Why is new business models all by itself in blue, we wonder?
  • Red is generally used to show an error, and orange a warning in charts. But here, that best practice, to put it in the words of the famous Mark Twain (who I admit I idolize :) ), “this rule is flung down and danced upon”. Speed of product delivery is in red, but represents the biggest gain in this chart.
  • A bright blue background further confuses the user. Using bright, saturated, attention grabbing colors (especially red and orange and yellow) causes the eyes not to know what to focus on first. Users end up jumping all over the place, instead of actually taking in the information you want them to.
  • As if all the red wasn’t enough, a red box is added, further distracting the user, even more so as there is nothing to show why that red box is there.

The Redesign:

Here, I’ve reduced to only the blue color, reduced the weight of the graph lines, removed the background (white background is most readable for most users), and added a small bracket indicator to replace the red box, complete with text. Nothing is left to the reader to figure out. Remember, readers won’t spend time reading the accompanying text and will only give your chart a quick glance to see if merits further study. You don’t have another chance to make a first impression!

Heavy treatment calls attention everywhere — what’s important?!

Here’s another culprit where the whole chart jumps out at you. But what’s the critical information you want to convey? Here’s what’s confusing you:

  • Everything is bold — the lines, the circles, the values
  • Text on legend seems to be important but is wrapped on 4 lines, impeding scannability
  • Bright blue background makes the whole thing vibrate
  • Grid lines have too high a contrast with the background and the data. They should be subdued, conveying information subtly, but not screaming it.

The Redesign:

Here, a more saturated bold treatment and text was used to convey the important parts, and size was used to immediately convey differences in value. Text was shortened and length of text column widened for faster scanning.

Color, text, arrows — oh, my!

There are many things that make the following chart difficult to understand.

  • The one that pops out is the arrows. What on earth do they mean? I couldn’t find anyone who knew! Internal dept systems at 49% goes to Customer supply chain at 36%? What?
  • And why all the red and orange?! These are attention-grabbing in themselves, but against the blue background, they seem to vibrate. Not what you want, unless it’s a psychedelic black-light poster like we got back at Spencer’s (in the US) back in college ;-D!
  • Unnecessarily long labels contribute to lack of immediate comprehension.
  • Crazy long title! Does this agency get paid by the character, like in Dickens’ day? ;-D

You probably wouldn’t guess what those arrows mean in a million years — I’m hesitant to even tell you. Apparently, they’re supposed to be trend lines. Hmm…well, not to me, and not to anyone else I asked.

Here, I simplified, after talking with the major stakeholder — is showing a decreasing trend of internal API usage that important, when what we really wanted was to show an increasing trends of external API usage? Is it worth cluttering up the chart to convey both values? In this case, we decided it wasn’t.

The Redesign:

Here, reducing the colors to just two, both shades of blue, where the external expansion is shown as more important with the more saturated blue to draw attention to it. A simple curly bracket indicator is used to replace those mysterious arrows on the original chart. Title and legend text are shortened to ease in comprehension and scannability.

Begin at the beginning!

Charts that start randomly in the middle of nowhere ought to be expected to explain themselves and how and why they came to be there (yes, Mark Twain reader, you may recognize one of my favorite author’s way of speaking there. I’m a polymath meaning I draw from many schools of knowledge — and I love Twain — Samuel Clemens is the author I’d most like to have met, if I invent a time machine someday, or you do! :-D).

In addition to the background and over-saturated, attention-grabbing colors we’ve discussed at length already, here another data viz rule is flung down and danced upon — to convey information effectively, charts should start at 0. Anything else and you’re trying to twist the data to show something else! Be truthful with your data!

Here, the cake layers make no sense. Why do they start in the middle? It makes it hard to compare. Bars in the same color convey a relationship that isn’t there. And the bars don’t seem proportional, either.

The Redesign:

This one is super-easy. Just begin at the beginning, as all good storytellers do! I also reduced to the blue theme. As you’ll see in some of the later graphs, I start using saturation to convey meaning as well, but here, they are just different shades of blue :).

Convey information consistently

If at first you use one method to convey a scale, don’t use a different method the next time just so that you can add more color! You’ll probably recognize this low to high scale used on a line graph in an earlier example, but it’s been made even more complicated!

Here, besides the background and colors, which I’ve harped on enough ;-D, and the red border, the legend and colors are incorporated into the x-axis. The grid marks really aren’t even necessary since the percentages are on the bars themselves.

The Redesign:

I told you I was going to start using saturation to convey meaning, and here it is! YAY! By using increasingly more saturated blues, we convey meaning — low competency to core competency. Time, effort, and money being put into something is indicated by a higher saturation. Oh yeah, and the curly bracket to replace the mysterious unlabeled red box above. The legend is simplified to only the important information.

Color has meaning — use it to convey information!

Here, the important information is not the No plans to section, but the Yes section and the No, but plan to section. The No plans to is particularly unimportant. That indicates that they’ve thought of this, and have discarded the idea. There’s little potential here. So why red and orange? Why use orange to call attention to the least important segment?

The Redesign:

Here, I looked at what the colors meant — orange for No plans to made no sense at all, nor did red for Plans to — red and orange are used for error and warning states, respectively, remember? No plans to means the light is off, it’s gray. I’ve also incorporated the labels right into the segments — no eye saccades for my viewers — they don’t have to jump between the chart and the legend, which further impedes comprehension and scannability.

And then, let’s look at modernity in pie chart. Donuts are much more widely used. Space to separate segments makes it harder to compare. The other thing I love about donuts (no, not to eat- I’m weird, I know) is the potential for the inner circle. I haven’t used anything here, but you can show something important such as number of customers surveyed, or even a simple picture to convey more information. Oh, the possibilities! :-D Yes, I freakin’ love UX — does it show? ;-D

Don’t make your readers think!

Oh, the colors! What do they mean? They’re all bright, but convey not meaning whatsoever! Get rid of them, data visualization grasshopper! :-D

The Redesign:

I had to stretch a bit here to move the Not important label into the segment, but felt that it worked better than having huge segments with the labels outside. Remember, if you can’t put the labels in the segments, right next to them is the next best thing. This is good if you have to deal with sending designs for translation, for example. By right next to, I mean adjoining, not in a legend. Here’s another article I wrote where I discuss pie chart design and color usage at length, for color-blind users, in case you’d like more information:

Color saturation and size

Oh goodness! Could we get any more boring than this? What’s important? Everything is bold, which means that nothing will be paid attention to. Nothing stands out, because everything stands out!

The Redesign:

This is probably one of my favorite redesigns :). Here, I used a less saturated blue to show internal focus increasing to a more saturated blue to somewhat or completely external customer-focused. I also used size to convey immediately where most companies are. And by grouping internal and external legends into two different blues and totaling the percentages of each, it cuts down on the cognitive burden 5 blues would cause. Trust me, I tried it :).

Hope you’ve enjoyed learning about data visualization and its importance in conveying information and especially, in getting them to read and more importantly, comprehend it. This can’t help but get more people talking about it! Good data visualization makes your company look good! What potential customers are going to go check out a company that shows them confusing charts? Would you expect that company to be able to help you? If your answer is yes, “I’ve got some ocean-front property in Arizona…” :-D

Happy data analysis and visualization!