Reviewing Google’s Six Data Visualization Principles

Recently, I was lucky enough to stumble upon a great article written in 2019 by Manuel Lima, senior UX engineer at Google, that was full of interesting facts and data visualization tips.

In this article Lima starts off by explaining how this list even came to exist. He shares that back in 2017, a group of brilliant Google employees from different backgrounds (graphic design, research, engineering) came together to create a complete guide to data visualization.. and let me tell you, they thought of everything. Shapes. Colors. Fonts. Icons. Moving Elements. You name it, they addressed it.

One of the byproducts of this helpful and informative guide was a list of six principles for designing any chart. These principles aren’t rocket science, but they are details that many people (including me!!) bypass and push under the rug. As a result, data visualizations are not as impactful as they could potentially be.

As I was reading this article, I couldn’t help but think of how all six of these guidelines were heavily applicable to my every day work as a consumer finance analyst. So here are six principles for designing any chart, as instructed by some of Google’s most passionate and brilliant minds.

1) Be honest.

We’ve all been there. You’ve been researching and cleaning and pivoting and melting and imputing data for weeks. It is finally time to visualize your data in a beautifully formatted chart. You call plt.show() and the unthinkable happens.. 

Your hypothesis doesn’t seem to be checking out. Maybe your graph looks like the one on the right side, but if you just tweak the y-axis juuuust a liiiitle biiiit, the growth of the interest rates can actually look quite quick over time.

Please, do your best to resist falling victim to this. Data is beautiful. But, data is also honest. If the data is telling you that the growth of interest rates by year has not been notable, then that’s okay. It just means there might be other secrets hiding in the data that you haven’t found yet.

2) Lend a helping hand.

This principle was actually quite enlightening for me personally. I had never thought of it before, and I was surprised. You should always know your stakeholder before building a data visualization that they will consume. Is this person an avid Excel user? Maybe you can design and style your tabular data display in a way that makes it intuitive for pivot table veterans to drill down into deeper depths of the numbers. Does this person have higher concern for certain department KPIs? Organize your dashboard to make those the most accessible and easiest to see.

The core idea here is to identify the user’s “existing mental models” as Lima put it, and take advantage of them. It’s like the coach building a game plan around the team’s strengths. Support the user’s creativity and curiosity by encouraging filtering, highlighting, and guiding them to the conclusions and insights living in the presentation.

3) Delight users.

This point is slightly funny to me, only because I feel like it is so broad and can be applied to any field of work. Nonetheless, it is an important rule to follow when visualizing data. As a data scientist it is your duty to take data and build an experience with it, to take rows and columns and develop a story. Don’t take this duty for granted.

As a data scientist, you should always go the extra mile to exceed your coworkers’ and bosses’ expectations. Show them everything they wanted to see, but on presentation day, come with an extra special perspective that they didn’t think of.

Choose design elements that elicit small moments of shock and delight. Trust me, bosses love to have things done quickly. But hear me out. When an employee comes to a presentation with work that took longer, but it’s thorough AND has insights that the boss may not have thought of, it’s worthwhile. That’s one of the perks of becoming increasingly specialized in your field – getting to surprise people with what you know.

4) Give clarity of focus.

Has this ever happened to you?

This pie chart (that quite frankly hurts my eyeballs) is a great example of how not to design visualizations. We should always avoid excessive categories, extraneous charts and labels, distracting animations, etc. Color should be applied in a meaningful, not overwhelming, manner.

A good rule of thumb could be that if the user has to squint, or lean in extremely close to the screen to understand the points you are trying to convey, maybe you should rethink your design.

Combine this principle with number two. Really step back and think about what the user needs to see. Once you have decided, direct the user to this as seamlessly as possible. As Lima said, “focus on the user’s task and all else should follow.”

5) Embrace scale.

This rule is especially important for the newer gals and guys to the field. Only because the longer you have been in the field, the more likely that you have gotten burned by forgetting or ignoring the potential for scale. (Mine was a divide by zero, and it’s ingrained in my brain forever.)

When you are designing a data visualization, always consider how and where this graph will be consumed. Will the public be viewing this? If so, make sure that you select clear color palettes, fonts, and labels that are accessible to users with all kinds of needs, as well as adding alt text. Will this be viewed on mobile? If so, make sure the visual is mobile friendly.

Also, try to imagine what this graph will look like in a month, a year, and three years. Is there a large amount of growth expected? Will this visualization still be appropriate then? Will the underlying queries to retrieve the data perform to scale? Is there potential for new categories or variables to be introduced? Always design your visualizations in a way that will let users explore possibilities and change perspective when needed.

6) Provide structure.

We all know that Starbucks doesn’t have the best coffee… and that it might be better to support local coffee shops than big business… and that the Starbucks drive thru can be a literal nightmare. So, why do we keep going back?

We always know what we’re going to get. All Starbucks look the same. The formulas for the drinks are all the same. The employees all wear the same uniform. Their brand provides structure.

We as data scientists can essentially brand our visualizations in a way that provides consistency, trust, and uniformity to our users. Pick a few colors for your color scheme and stick to them. Implement common typography across graphs to establish a feel of smooth transition and cohesiveness from visual to visual. Always provide a way for a user to go back after navigating through the chart.

These six principles for designing any chart are essential practices to generate an insightful, intuitive, and intriguing data visualization that will leave a lasting impact on your users.

One thought on “Reviewing Google’s Six Data Visualization Principles

Leave a Reply

Your email address will not be published. Required fields are marked *