DataViz Remake: how visuals can mislead or reveal
September edition: on message, accessibility and axis illusions
Data visualization is one of the most powerful tools we have to communicate complex information quickly and effectively. A well-crafted graph can condense thousands of data points into a single, clear message that anyone can understand at a glance. But have you ever stopped to consider what happens when a visualization isn’t well-crafted? Can a visualization confuse more than it supports your message? The answer is a resounding yes.
Welcome to our new series called DataViz Remake, where we explore the same information with different approaches to better communicate a message. Let's dig in.
A case study: visualizing climate data in the Pantanal Region
Let’s examine three versions of the same dataset. The data tracks the relationship between annual mean temperatures and total annual rainfall in the Pantanal region (Brazil) over the last 45 years—crucial data shedding light on the ongoing changes in one of the world’s most important ecosystems.
The original graph
The original graph is titled:
“Precipitação anual acumulada versus média anual da temperatura máxima diárias nos últimos 45 anos: Gráfico mostra a tendência de seca e de aquecimento no Pantanal ao longo dos anos e o recorde em 2024”
or, in English:
“Total annual rainfall vs. annual mean of daily maximum temperatures over the last 45 years: The graph shows the trend of drought and warming in the Pantanal over the years and a record in 2024.”
Let's first read the graph properly:
We have two variables being shown over the years:
Rainfall (shown in the vertical axis / y); and
Temperature (represented horizontally - in x).
The years are represented through a categorical color system, where red represents the more recent years, yellow covers from 2000's to 2018, and green represents 1979 to the 90's. Honestly, this is the weirdest choice in our opinion, but moving on.
Despite the title's claim, the selected graph, and most importantly, how each variable is represented, create a misleading impression. With the choice of a scatter, it seems to show that temperatures are decreasing over time. Which is not true.
The confusion stems from the way the axes are arranged, combined with the timeline being represented as colors. Humans read a graph from left to right in a Z motion. Plus, we are used to seeing time being represented horizontally, from left to right. By placing temperature on the horizontal axis and rainfall on the vertical axis, the graph creates a visual effect that doesn’t accurately represent the true story of the data. Plus, by representing years in colors, we tend to assume they represent temperature, since red is most commonly associated with warmer temperatures. It is trying to fix the problem with the axis by adding colors and segmenting the timeline in zones. Which only adds to the confusion.
How long did it take you to understand this graph? Probably as long as it took us to explain it.
A clearer trend, but a new problem
Now, let’s reimagine the graph with new approaches, focusing on our story:
We want to say that in the same period over the years
The amount of rainfall is decreasing in Pantanal; and
The temperature is rising.
The choice for a correlation seems obvious for a scatterplot. But is it, though?
In this new version of the graph, we attempted at inverting the axes—temperature is now on the vertical axis and rainfall on the horizontal axis. This arrangement seems more intuitive at first glance. With this adjustment, the message becomes clearer: temperatures in the Pantanal are rising, while rainfall is steadily decreasing.
However, this change comes with a new problem: the years now appear out of order, making the timeline harder to follow. While the graph better visualizes the trends in temperature and rainfall, the lack of chronological order creates confusion about how these changes are unfolding over time.
This version of the graph does a better job of highlighting the alarming warming and drying of the region, especially in the most recent data from 2024, where we see a sharp spike in temperature and a corresponding drop in rainfall. Yet, without a clear time progression, it’s difficult to track the evolution of these trends accurately.
Why axis orientation still matters
Why does this small change in axis orientation make such a big difference? The human brain is wired to interpret visual data in specific ways. We naturally associate the vertical axis with variables that increase or decrease over time—like temperature—and we expect trends to flow from left to right, which typically represents the passage of time. When these conventions are flipped, even a technically accurate graph can lead to misunderstandings.
In this case, switching the axes clarified the trends, but the jumbled chronology hinders our ability to fully understand how these changes have occurred over time. It demonstrates that while axis orientation can improve clarity, it's not always a perfect solution when other aspects—like time—are compromised.
In the world of data visualization, these small design choices can make or break the accuracy of the message conveyed. A poorly designed graph doesn’t just fail to communicate the right information—it can actively mislead your audience, causing them to draw incorrect conclusions from the data.
The importance of selecting a dataviz that clearly represents your message
When we create data visualizations, two essential questions guide our choices: Who are our users? and What is the message we need to communicate? In this case, our users might be readers of a popular science magazine or a quick-scrolling social media post. They could be climate experts—or simply concerned citizens. What they all have in common is that they need a clear and compelling way to grasp the story behind the data.
That story? The climate crisis, which has been unfolding for decades, has led to a dramatic decrease in rainfall and a sharp increase in temperatures. These shifts are already triggering severe consequences across the globe—droughts, wildfires, and widespread damage to ecosystems.
Now, how do we make this information not only understandable but undeniable?
We’re working with three variables: the years (from 1979 to 2024), the annual mean of daily maximum temperatures (°C), and the total annual rainfall (mm). To show these two critical trends over time—temperature rising and rainfall shrinking—we chose a dual-axis combined graph. The years run along the x-axis, while the temperature trend takes the left y-axis and rainfall the right.
Why this format? It’s not just for the science-minded—it’s because it tells the story at a glance, to anyone. The message is accessible: rainfall is shrinking, and temperatures are climbing, year by year. The graph vividly highlights this contrast, with the blue curve for rainfall narrowing while the red for temperature swells.
We know that some people will say that representing data on two vertical axes is not the best solution as well. Our tests indicate it is. But if you're still not convinced, you can also choose to use two charts on top of each other, as the example below. Although they are correlated, they have different ranges, but still placing them in the same timeline can convey the same idea, even if separate.
Not everyone can read a scatterplot. Sometimes, you need to go back to simpler visualizations to convey your message.
Note: Since we do not have access to the full database, what we are showing in the graph is a simple simulation of the real data based on the graph we had.
The responsibility of data visualization
As designers and consumers of data visualizations, we have a responsibility to ensure that the information we present is accurate, clear and accessible. This responsibility extends beyond simply avoiding intentional misinformation. It also means being aware of how design choices—like axis orientation and the type of data visualization—can unintentionally distort the message or make it inaccessible.
In the case of the Pantanal scatterplot, the difference between a misleading and an accurate visualization was not just a matter of swapping the axes but also of choosing a better type of data visualization to represent the message, not just the two variables being represented over time. Instead of falsely suggesting that temperatures are cooling, our proposal reveals the worrying trend of increasing heat and decreasing rainfall. However, we must also ensure that other aspects, such as the chronological order of years, remain clear.
And the only way of understanding that is by iterating: testing, testing, testing. The more you test, the better you become at telling stories.
As the world becomes increasingly data-driven, the ability to create clear, accurate visualizations is more important than ever. Whether you’re a data scientist, a journalist, or just someone who loves a good data visualization, remember that how you present data is just as crucial as the data itself.
If you found this discussion on data visualization thought-provoking, there’s more to come! This is just one part of our ongoing Dataviz Remake series, where we explore potential challenges and overlooked issues in data visualization. Our goal is not to criticize but to spark meaningful conversations about how we can all enhance the way data is communicated.
→ Subscribe to our Substack for regular updates and insights.