Data visualisation lessons we’ve learned

There’s a point in almost every data visualisation project where you’re staring at a spreadsheet full of percentages and thinking: Okay… so what's the story?

The numbers are rarely the problem - it's deciding which numbers matter, how they relate to each other, and how to turn them into something a person can understand without having to study the page.

That’s something we’ve learned again and again while designing research communications at MYMAVINS. A good visual isn’t necessarily the fanciest design. Usually, it’s the one that makes the insight obvious without requiring too much effort.

You look at it and feel good: Right, I get it. 

So, what actually helps us get there?

Start with what the data is actually saying

It’s tempting to standardise charts and design treatments based on data structure and type. There is sense in that, but we have found before making treatment decisions it's useful to first ask: What is the data saying and what do I want the audience to take away from this?

A survey can contain hundreds of data points, but they’re not all equally important. Sometimes the story is a large percentage and sometimes it's a small percentage that was expected or hoped to be bigger. Sometimes it’s the gap (or lack of) between two groups. Sometimes it's a story about time. Sometimes nothing looks particularly dramatic until you put several pieces of data next to each other and a pattern appears.

If you can’t explain the insight in one or two sentences, it’s probably too early to design it. Instead of asking, “How do I visualise this table?”, start asking, “What does someone need to understand from this data insight?” - which are two very different questions. This also requires input from team members that have a deep understanding of the client/audience's contextual needs.

Not everything needs to shout

Once you know the main point, the next challenge is deciding what the viewer sees first.

This comes up a lot in research design because there can be a tendency to treat every statistic as equally important: Same font size, same colour, and same visual weight. Technically, everything is there. But visually? Nothing is really leading.

We’ve learned that a strong piece of data communication needs hierarchy. There should be a clear first read, then a second, then perhaps more detail for someone who wants to keep looking. Maybe the headline statistic is huge. Maybe one comparison gets the strongest colour. Maybe the supporting figures sit underneath in a quieter treatment.

The exact solution changes from project to project, but the principle stays the same:

If everything is important, nothing looks important.

Put the comparison in front of the reader

A percentage without context can be surprisingly unhelpful.

Is 55% high? Low? Expected? Surprising? Put it next to a comparative group or past figures and immediately, there’s a story.

A lot of the strongest findings we work with come from differences: sociodemographic groups, attitudinal and behaviour segments, expectations compared with reality, or the lack of difference where we expected there to be one.

If two numbers are meant to be compared, put them close together and use design to contrast differences while emphasising they can be compared - this is not a pick the difference puzzle. It sounds obvious when you say it out loud, but it’s an easy thing to lose once layouts start getting busy.

Good data visualisation reduces that mental effort. The comparison should happen on the page, not in the reader’s memory.

Sometimes the best thing you can add is… nothing

Designers like designing things. That probably isn’t a shocking revelation.

There is a temptation to demonstrate design effort. As we are going to discuss in an upcoming blog, there is an inherent power in perceived effort. A halo effect that implies effort rendered into generating the insights themselves.

There’s always a temptation to add an icon, a background shape, an illustration, a pattern, a callout, maybe another little graphic to “make it more visual” or “break things up” - sometimes those things genuinely help. Sometimes they just make the insights harder to see.

We’ve become more conscious of asking a fairly basic question: What is this design element really adding?

If an illustration helps explain an abstract idea, great. If a colour scheme helps separate groups, useful. If an icon makes a category quicker to recognise, keep it. But if something is only there because the layout felt a little empty, maybe empty wasn’t actually a problem.

This doesn’t mean every data visualisation needs to look clinical or stripped back. A lot of our work lives in carousels, videos and social content, where visual storytelling is important. People still need a reason to stop and look.

The trick is making sure the visual idea supports the data rather than becoming the thing you notice instead of the data. There’s a difference between making research engaging and decorating it until the research disappears. When data visualisation works really well, you don’t necessarily notice the design decisions behind it. You just understand the information more intuitively. The hierarchy feels natural. The comparison is obvious. There’s enough context to understand what the number means, but not so much that you have to hunt for it.

The goal is to make the reader think “that’s interesting” rather than “that's impressive looking".

Give people a way into complex information

Sometimes there really is a lot to show. You can’t always reduce a research story to one statistic, and you probably shouldn’t. In those cases, we’ve found it helps to think less about simplifying the information and more about sequencing it.

• What does the reader need to see first? 

• What does the reader need to see next? 

• Then what makes more sense once they already understand those first two things?

For example, you might start with a broad finding setting up the magnitude of the issue or problem. Then dive into details that structure a journey through the lived experience of the issue and then explore how groups differ. Only after that do you tie a bow on it with suggestions for improvements and advocacy support.

The detailed table hasn’t disappeared. The reader just has a way into it, and that’s an important distinction. Complex information doesn’t necessarily need to be made simpler. It needs to be made easier to enter and follow a journey through sequenced digestable 'insight bites'.

Format follows function

This might be one of the biggest habits we’ve had to unlearn - that all data needs a chart.

But what if a chart isn’t actually the best way to communicate the finding? Sometimes one large number and a short sentence are enough. Sometimes two prominent figures sitting next to each other communicate the point faster than a graph. 

If it’s about one standout result, maybe that result deserves to stand on its own. If it's a nested insight (such as the reasons for something), sometimes just listing the top three in supporting prose places the insights where it's most relevant and helps reduce charts bogging things down. Sometimes a well designed table tells the clearest story.

The format should follow the question the data is answering, not the other way around.

Accuracy isn’t finished when the numbers are correct

There’s another side to data design that’s less exciting to talk about, but probably more important than anything else in this article.

You can use the correct number and still communicate it inaccurately. A bar can exaggerate a small difference. A large icon can make one percentage feel much bigger than another. An inconsistent scale can turn an ordinary movement into something that looks dramatic. Even colour and visual weight can quietly imply that one finding matters more than the evidence actually supports. We all want our work to tell a story - even if sometimes it's just not there.

That’s why our QA process isn’t just about checking whether the artwork reflects the datacube. We also have to look at the design and ask: Does this feel like an honest representation of the data?

Researchers are responsible for the accuracy of the numbers. Designers carry responsibility for ensuring those numbers are communicated without misleading,

We weren’t the first people to try and figure this out

None of these ideas are particularly new but perhaps suprisingly are still overlooked. We found many of the same principles reflected in Edward Tufte’s The Visual Display of Quantitative Information, first published in the 1980s.

Tufte wrote extensively about reducing unnecessary visual noise, making comparisons easy, preserving the integrity of the data and making sure the information, not the decoration, remains at the centre of a graphic. What’s interesting is how relevant those ideas still feel.

The formats we design for now are very different. We’re working with scrolling websites, social carousels, short-form video, animated graphics and responsive reports.

The fundamental problem hasn’t really changed though. We still need to decide what matters, help people feel that ‘aha’ moment and remove what gets in the way.

Dataviz Gallery - The Early Years

I thought I would leave you with some of the earliest data visualisations to really make an impact!

William Playfair, 1786

One of the earliest widely recognised uses of line and area charts to show economic change over time.

John Snow, 1854

Famous early use of data-driven graphics to address a London cholera outbreak. Plotting individual cholera cases uncovered a clear pattern pointing directly to the origin of the epidemic: a public water pump on Broad Street. 

Florence Nightingale, 1858

Her polar-area diagrams made the causes of military deaths visible and helped communicate the case for sanitary reform.

Charles Joseph Minard, 1869

A famously information-dense visual combining the size and movement of Napoleon’s army with geography and temperature.

Author

Nico Villaneuva is a senior designer at MYMAVINS with many years experience leading diverse design projects from marketing to research insight communications.