Design and formatting choices for dashboards and visuals

Good design and formatting build trust in our dashboards and visuals. They shape how we use, understand and test our algorithmic processes.

Well-crafted visuals help us make accurate decisions based on reliable information, reducing confusion and misinterpretation. We want to ensure that the most relevant information, alerts and details are presented clearly. Dashboards that are too busy get ignored, or worse, they mislead.

This article shares 10 design and formatting practices to make visuals sharper.

Dynamic Dashboards

These let people filter and explore the data. They are designed flexibly to answer multiple questions.

Because these are dynamic, it’s not always that easy to get the formatting balance right. Still, there are simple ways to keep dashboards easy to read.

Here are some core practices to help make dashboards more effective:

1. Keep it simple

• Focus on usefulness over decoration
• Avoid 3D charts and exploding charts
• Use enough white space to avoid clutter

2. Chart type

• Pick chart types that readers can understand quickly
• Be extra careful with pie charts and donuts; they have limited uses
• The Financial Times visual vocabulary is a useful resource

3. Slicers and filters

• When items are selected, highlight them (use clear highlights for selected slicers and filters)

4. Number formats

• Don’t use the defaults
• Numbers don’t always need to be precise (“$2954.99” is hard to read, “$2,955” is easier, “$3k” could work)
• Unless really needed, drop the decimals in percentages

5. Colours, and fonts

• Limit the number of colours, and use contrasting colours where possible
• For internal use, corporate (brand) colours are often well known and can be easier to understand
• Use fonts that are easy to read (again, brand fonts may work best)

6. Legends, axes and labels

• Label directly, replacing legends, if possible
• If a legend is needed, minimise the range of potential values
• Start all axes at zero
• Use data labels, dropping axis titles/values if possible

7. Header, footer, and other metadata

• Include information about the graphs, data sources and dates
• Include known limitations, quality issues like gaps, if consequential

Static Reports, Graphics, Visuals

Present information without control over what’s displayed, but still with the ability to answer a few different questions. With static reports, what’s shown doesn’t change.

These are easier to control than dynamic dashboards. The 8 practices for dynamic dashboards above continue to apply, and we can add this:

8. Show pre-selected slicers/filters

• Include information on which filters have been applied
• Show both the inclusions (what has been selected) and exclusions

Data Stories

Quite different to the previous two, because here we are focusing attention on the story.

Extending on the items above, we now add:

9. Hero statements instead of headings

• A data story does not have a traditional heading
• Replace the “heading” with a bold sentence that describes the key takeaway
• Include notes, callouts or annotations for secondary points

10. Focus attention by boldly highlighting the core elements of the story

• Make the key part of the data story stand out visually
• For example, highlight the specific bar or line the story focuses on.

Disclaimer: The information in this article does not constitute legal advice. It may not be relevant to your circumstances. It was written for specific algorithmic contexts within banks and insurance companies, may not apply to other contexts, and may not be relevant to other types of organisations.