Cloud Data Viz and Analytics Health Check
Uncover the fitness of your Cloud Data Viz & Analytics
Get my free scoreTop 10 data dashboard errors (and tips to avoid them)
Do you create dashboards that communicate the right message clearly and steer people to the right action? Achieving a compelling visual data story often means you need to avoid a few common design pitfalls.
Letâs take a look at the Top 10 dashboard design no-noâs and tips to avoid them – from our Product Manager (and master dataviz maker), Piers Batchelor. đ
10. Too many colors!
Avoid using too much color. Stick to the fundamental ruleâââkeep it simple. Simplicity leads to understanding and keeps your audience on board. đ¨
A palette should consist of around 6â13 colors (odd numbers work best with gradients). With more than 14 colors, some of the colors become harder to distinguish. Twelve colors are ideal for visualizing months in different colors.
Though this isnât the broadest use case it is also wise to have enough colors to cover most cases. For example, there are many companies that have 10 product categories to distinguish between. Amazon has a lot more than this.
đĄ Tip: In most cases, you want to use between 4 and 6 colors.
If the palette is a diverging color palette, the colors need to be odd to consider a middle color, hence the maximum of 13 colors!
A great resource for tried and tested data viz color palettes is color brewer. To test color palettes, I use Viz Palette by Elijah Meeks and Susie Lu.
9. Homegrown gradients â
On the topic of color, gradients can introduce their own problems.
Humans understand simple gradients found in nature, try to avoid using a gradient made up of your brand colors. Gradients should follow a natural sequence commonly understood by all.
A rainbow isnât a simple gradient found in nature, because not all end-users will be able to remember which order the colors should appear in.
Examples of simple gradients are sunsets, ocean-themed, and in some cases, plant or forest themes. Hereâs an example of a sunset gradient which youâve probably seen in charts before.
Green/red doesnât always need to mean good or bad. Think of this in terms of high and low, and you’ll have more easily understood gradients.
đĄ Tip: To quickly generate clear gradients with a number of stops, use a tool like gradstop or chroma.js.
Gradients are best used in heatmaps and scatter charts, where gradients bring the chart to life and offer more meaning to the data by visualizing an additional metric. That would mean a scatter chart could show 1 dimension and 4 measures (x, y, z [size] and gradient color). A heatmap could show 2 dimensions (x,y) and 2 measures (size and color).
8. Unclear link to story/narrative
The chart below claims its purpose is to rank countries by alcohol consumption. Howeverâââ1L of Beer =5 units, 1L of Vodka/Spirit = 40 units.
If we converted to consumption of units of alcohol, then a country like Nigeria would most likely move further away from 5th place, telling us a very different story. đ
7. Transparent data
Donât try to embellish the data. Make sure it is transparent. Plus, also ensure the data is correct! And that the parts add up to the total figure.
In many cases filtering can be useful. Â For example, showing a rolling 28-day trend, you do want to ignore the past data.
đĄ Tip: Keep your story relevant and try to focus on what you want to share, without hiding aspects of the data.
6. Donât forget scale
đ Start from zero
As you can see, 28% of NBC charts have the correct scale đ
Donât truncate a graphâs axes as it may distort the data representation and minimize the impact of your message. A clear example of omitting Baselines and Truncating Scale is shown below, where the scale starts at 28%.
Interestingly, a colleague pointed out to me that 13% is greater than 28% – which then made me notice that 25% is also greater than 28%.
âď¸ Area vs radius
Source: Coolinfographics.com
If you create your own visualizations, remember circles are compared by area, not so much by radius. The ideal way to compare is âhow many yellow circles can fit inside the blue circleâ. If the answer is 9, then the blue circle is of course 9 times larger.
Equally a Bar chart would have been better here. This can be translated and more easily interpreted as Linear vs Quadratic change.
5. Too many charts
Too many charts indicate that the focus is on visualizing all possible data , rather than telling a meaningful story with data. If the chart does not produce insight, necessary context or better yet, actionable insight, it probably isnât needed.
Try to avoid:
- No breaking white space (between charts)
- Poor gradient used (thatâs right theyâre not all the same shade of green)
- Metrics not labeled (green box on the top right of each chart)
đĄ Tip: Good dashboards will make effective use of whitespace, breaking up charts and telling a meaningful story, with the intentional avoidance of metric overload.
With quality data and clear metrics, youâll be on track for a happy audience. Dashboard like AirBnBâs uses a tab-based container to focus on one metric at a time, not togetherâââreducing the need to show 3 charts alongside.
And I hear you say âBut Piers, I need to compare metricsââââaha!âââyou can do that in a Line chart that has a primary and secondary axis.
Airbnbâs performance dashboard. Source: AirBnB
4. Too many values in a chart
Change in popularity of genres over time
High cognitive load (the amount of info your brain can process at a time) can cause stress, and fast.âDistraction is a killer â your audience wonât understand what youâre trying to show and will move on.
If the chart settings feel overwhelming, so can a busy chart. If a user has to think about how they should best read it, the chart is not great.
The value of having âOtherâ is underrated. On a purposely small image of the visualization (a realistic scenario), I can clearly see 10 key Genres that span across time. The rest could be either grouped together or colored in grayscale, so that we draw focus to the change in popularity of top genres over time.
đĄ Tip: If youâre going to need lots of values, a helpful title can draw focus to the key values. Alternatively, consider highlighting key values in bolder colors.
3. Overly complex charts
đĄ Tip: Charts are like jokes, if you have to explain them, theyâre probably no good.
Like jokes, good charts are simple.
I love the creativity behind combining charts, however, I couldnât tell you where the top of the bar scale is! More frequently I see these as Bar charts. A general exception to this rule is the use of infographics, where many rules can be broken to look appealing and tell a visual story. An infographic usually serves the purpose of spending more time on a single visual or a single message.
2. Donât ignore sorting
When it comes to screen-reading, I (and I assume others) get impatient or frustrated when information is poorly laid out. There are many mistakes below:
đ Bad
- The bars are not sorted!
- The legend should be sorted in the reading order that the colors appear on the chart. For most of us, they should be ordered from top-left to bottom-right: blue, orange red, etc.
- Donât mix scales! Percentages are mixed with stacked numbers below, which is mixing normalized and non-normalized data! (see the legend)
- Number labels on the bar and scale? A bit of duplication but I could let this one go⌠For me, as a personal rule, bars are shown inside a stack, why not show them outside without an axis, saving space.
đĄ Tip: Labels look much neater when you use them only in one place.
Source: @_cingraham
One good about this chart is that âExcessive drinkingâ is the only variable that matters for the purposes of the rankingâ ââ@_cingraham.
1. (Static) 3D charts
It is 2022, not 2002!
A fun extra point to remind everybody that static 3D Bar charts add very little valueâââwithout changing perspective.
Change my mind in the comments if you so wish!
Artistâs impression of a chart.