a  woman  displaying a data visualization for teammates

What Makes a Good Data Visualization?

5 minute read
Pierre DeBois avatar
Effective data visualizations display complex information in a way that's easy to understand. Here's what goes into creating a good one.

Data visualization has grown more complex as the number of data points brought into the mix has increased. In the past, data visualizations appeared as ready-made graphs in the solution of choice. But visualization frameworks have introduced new options for illustrating data, especially in instances where more data points connect customers and clients to your brand.

So what does making a good marketing data visualization involve today?

What Makes a Visualization Good?

A good visualization should establish two aspects of the data being presented:

  • Show connections within the data that are too complex to explain with words.
  • Make it easier for the audience to quickly understand the information presented and consider the outcomes from that data.

The complexity comes from the number of observation categories — usually a variable and its data — along with the data type (numeric or categorical). A visualization for displaying statistical distribution of numeric data is usually a histogram or box plot if there is interest in its modes, and outliers. A visualization for categorical data should display frequency distribution and skew. All of this is typical for visualization data where the observations are for one variable — for example the screen size of a laptop would be a variable with observations containing numbers listed for a set of models.

Two or more variables creates a bit more complexity. The best visualizations depend on whether the comparison is numeric to numeric (e.g. scatterplot), numeric to categorical (e.g. multiple histograms), or categorical to categorical (e.g. side-by-side bar plot).

To help managers get a clear view of the data, a visualization should reduce the complexity inherent in the variable and data choices. The graph should convey a hierarchy of how the data relates to your business concerns, revealing underlying meanings that speak to the intended audience. So if you are presenting data to a marketing manager who is interested in online engagement, you would use a metric that is instantly recognizable to them. You should always be clear on who the audience is for any given visualization and what it should explain.

Interactivity introduces yet another layer of complexity. You can have interactivity over time, to give the rate in which a metric or result is occurring, or in a side by side comparison. The challenge lies in deciding what method best tells the story. Research data is often buried in a table, so selecting interactions for  comparison must be intuitive, especially if the visualization is developed from a programming framework. 

Related Article: How Data Visualizations Are Making Self-Service Analytics Easier

Learning Opportunities

A Few Rules of Thumb for Creating Powerful Visualizations

Overall, no matter the variables and observations available, a few rules of thumb will make your visualizations powerful mediums for illustrating the concept behind your data.

  • Consider how figures and background contrast. The standard rule is to have a gradient in similar shades for online observation, but you may want to have a version with different patterns if you know the audience will likely print the graph.
  • Identify accepted thresholds for each variable. Doing so provides a benchmark against past performance, and can provide guidance on how to avoid displaying misleading differences between categories. So for a line chart showing how a digital ad performed over time, you may want to show how previous campaigns fared to help the team decide if the campaign is worth repeating or if the results can be built up. 
  • Label bar charts with numbers, but not to the extent that the details become overwhelming. Long numbers are usually hard to view. Use a shorthand recognizable to the team when accuracy to the nth degree isn't important. For example, the value “10,523” can be shown as “10K” on a bar graph.
  • Where possible, sort data by value to emphasize scale, but be careful not to distort the difference between categories.

Related Article: Telling Stories Through Data Visualization

Data Outliers and How to Handle Them

Some data sources create unique complexity. You will have to review how that data is used to ensure  your choice of visualization will make sense to your intended audience.

Survey results, for example, usually need to be scored — coding a response into a category that can be recognizable for regression analysis or machine learning techniques. Thus you need to understand the survey and the data to decide how to explain categories and differences through graphs and charts.

The choice of visual patterns is another point where understanding the data used influences visualization choices. Clorapeth maps represent regions with different colors or patterns. These are perfect for regions or clustering, but make small regions hard to see for handouts or small screens. Heat maps can show gradient changes, but they can be a poor choice when knowing the distinct numeric differences between data points is needed.

At the end of the day, a visualization should unite the intended audience around the problem the data is describing. With the problem well described, your managers can go be the heroes and heroines who solve those challenges.

About the author

Pierre DeBois

Pierre DeBois is the founder of Zimana, a small business digital analytics consultancy. He reviews data from web analytics and social media dashboard solutions, then provides recommendations and web development action that improves marketing strategy and business profitability.