Photo by Nick Fewings on Unsplash

Four Things Data Viz Practitioners Can Do to “Get Better at Design”

Desiree' Abbott
HCLTech-Starschema Blog
8 min readMay 10, 2022


So, you’ve gotten pretty good at making visualizations in your chosen tool. Your next challenge, should you choose to accept it, is to learn about how to make them look better. I’ve seen dozens of data viz community members just like you in the same position where you find yourself now, asking the same question: “How do I get better at design?”

Like a proper citizen of the 21st century, you take to Google. Ah, good old Google.

A Google search for “how to get better at design” yields a featured snippet from an article, called “9 Ways to Improve Your Design Skills.” The list items shown are 1 Subscribe to design blogs, 2 take an online course, 3 read books and magazines, 4 follow along with other designers/design agencies, 5 pay attention to great design to see what they’re doing right, 6 utilize templates, 7 recreate designs for practice, and 8 experiment.
Great ideas, but probably not the quick help you’re craving.

The problem with this approach is that it’s geared toward people who want to learn design as a general discipline, not necessarily as it applies to data visualization. For example, doing an online design course has you fighting off sleep at worst, or at best wondering how you’re ever going to have time to complete all of the videos and side projects — as well as do your actual job and live your life, and, ya know, sleep. And still, they never cover data viz!

Never fret though, I’m here to help. I’m not claiming to be some classically-trained design wizard, but I am here to share what I’ve learned on my journey thus far.

Please bear in mind that this is not meant to be a replacement for following the steps outlined in the aforementioned Google search, because that really is how to get better at design. These are just a few first steps you can take right now to level-up your design skills in the near term while you also pursue other avenues on your journey to data viz design nirvana.

And so, we present a tool-agnostic listicle about beginning to apply design principles to data visualization, in order of least to most difficult/subjective:

  1. Clean up your tooltips
  2. Declutter: Give your viz the Marie Kondo treatment
  3. Get it right in black and white
  4. Typography: Err on the side of simplicity

You’ll probably get the most out of this article if you have a visualization in mind which you’d like to “make over” or clean up.

1. Clean up your tooltips

Oftentimes, tool defaults don’t generate the most helpful tooltips. Instead, you might consider trying either the “complete sentence” approach, or the “key-value pair” approach. Regardless of which you choose, make sure to remove any extraneous fields and nonsense that the tool automatically puts there. You can even change up the font formatting just like you would for data labels or other text objects.

Shows 3 versions of a tooltip with the same information — the first (“not ideal”) is a result of the tool default and has messy field names, while the second (“better”) is a complete sentence and the third (“better”) is a cleaned-up version of key-value pairs, e.g. “First observed: August 2017”

If using the key-value pair approach, be sure that all fields have a reasonable, legible name that will help your user understand what they’re seeing, and again, remove any fields the tool automatically generated which don’t need to be present.

2. Declutter: Give your viz the Marie Kondo treatment

Just like the tidying-up queen recommends, we’re going to pull everything out of its deepest and darkest corners into the light, get rid of what we don’t need, and then put back only what we definitely want to keep.

First, let’s remove all borders, gridlines, axis lines, horizontal and vertical rules and row and column dividers. If there’s a line that’s not part of your data, get rid of it. These lines do not spark joy, so thank them for their service and chuck them in the bin.

Left image shows a small, simple dashboard with a scripty-font title, two charts (a colorful bar chart and a vibrant line chart), each of which have gridlines, as well as two horizontal dividers and one vertical divider. This will be the little viz we build upon through the article. Right image shows the same dashboard with gridlines and dividers removed.

Then, let’s add at least 30 pixels of space between all elements on the page. Increase the margins on the sides and top/bottom, add space between headers and the charts they head, and add space between charts. Give all your dashboard elements more than plenty of room to breathe — the more, the better. You might have to rearrange a bit, or increase the overall size of your dashboard or main container to accommodate all of the empty space, but you should still do it.

Left image is the same as the right image from previous, while the new right-hand image is the same elements with more space between the elements and the page borders, as well as between each other.

Finally, if you’ve enclosed any regions of your visualization using shaded boxes, let’s go ahead and get rid of those too.

Ok, now that we’ve got a much cleaner slate, we can really get to work:

  • Go ahead and add back any lines you think you actually need, taking care to keep colors and weights consistent — for example, you might decide that axes should be 1px and #4C4C4C color, so you should make sure all axes displayed adhere to that convention. When you do want to use lines, make sure they’re light and unobtrusive (as in, thin and closer to the background color), serving to make the important stuff stand out rather than competing with that important stuff.
  • Reduce space where it seems like too much. Like the Gestalt principle of proximity teaches us, things that are related should be closer together, while things that are not so related should be farther apart.
Left image is the same as the right image in the previous example, while the right image has the same elements with the addition of light grey axis rulers and slightly tighter spacing between dashboard elements.
Adding some axis lines works pretty well here, but we don’t need much more than that.

3. “Get it right in black and white”

Follow the adage (I don’t know that I would call it old yet) and put your whole viz in greyscale by changing all of your colors to grey tones — better yet, start your viz in greyscale in the first place.

Left image is the same as the right image from the previous example. The right-hand image is the same but all colors have been replaced by a medium grey.

Once you’re looking at a sea of grey, ask yourself these questions:

  • Can you still tell categories apart?
  • Do sequential and diverging palettes still work to help users understand the data?
  • If you are aiming for accessibility, is there enough contrast between objects or text and the background?*

Once you are confident that the viz still “works” in black and white, then you can start adding color back into the mix. Consider using a monochromatic palette because it will force you to not rely on color alone as an indicator, which is important when designing for accessibility.*

Left-hand image is the same as the right one from the previous example. In the right hand image here, there is a vibrant blue in use for the title, the bars, and the line on the line chart.
I love the monochromatic look, don’t you?

For further reading, check out Lisa Muth’s Datawrapper blog post, What to Consider When Choosing Colors for Data Visualization.

*On the topic of accessibility, this is my very favorite tool for checking whether contrast is high enough between elements and the background. Word to the wise: if you’re checking objects that will be at least 3 pixels wide, set the text size in the contrast tool to 18px bold — this will set the pass/fail criteria to the level of contrast needed for things like bars in a bar graph, slices of a pie or donut chart, etc.

4. Typography: Err on the side of simplicity

Ok, I’m going to be straight with you here: I’m actually still working on getting better at typography myself. Does anyone else remember the days of desktop publishing, when the tool would suggest different font “themes” — a set of 3 fonts that are totally different but go nicely together and give the document a certain look and feel? No? Fine, maybe I’m dating myself here, but I still miss those days. The main advice I have is that if you’re at this stage, you should err on the side of simplicity.

Consider approaching typography the same way we approached lines, spacing, and color: take it down to the bare bones and then start building up again. Make everything the same, basic font face (e.g. Arial or Times New Roman), style (no italics), weight (just use regular weight instead of bold or light), and size (I’d start with 12pt or 16px — yes, they’re different!) and then start changing the faces, styles, weights, and sizes where they need to be changed.

Change the scripty font to a more basic Arial.
Go from Arial to a bit more interesting but still thoroughly legible font.

If this is your first exposure to learning about typography, check out this little glossary of a few key terms.

The final reveal

Your journey to data viz design nirvana has only just begun, with small steps such as cleaning up your tooltips, decluttering by removing unnecessary lines, adding breathing room and simplifying colors and typography. This minimalist approach to design and to visualization isn’t for everyone, nor is it for every viz — but especially if you’re trying to help your users make data-driven decisions, it’s a really great place to start.

The left-hand image shows the initial dashboard we started with, while the right is the final product after simplifying the text, removing unnecessary lines, increasing the spacing, and applying color judiciously.
The final before & after. What do you think?

And now that you’ve gotten a great start, keep practicing these steps each time you build a new visualization. I even encourage you to start seeking out other design education resources as enumerated in our initial Google search result because once you truly understand the rules, only then will you know when and how to break them — and you’ll be able to break them beautifully.

About the author:

Desireé Abbott is a Senior Data Visualization Consultant at Starschema. She helps companies get that “aha” moment of truly seeing and understanding their data by turning their dashboard dreams into reality, right down to the last pixel. When not up to her eyeballs in charts and graphs, Desireé plays D&D, dabbles at piano and turns strings into things by crocheting. She’s currently living her best life with her husband and their fur babies: two cats and one very energetic dog. Connect with her on Twitter: @CallMeDeeray.