WHAT IS DATA VISUALIZATION?
A DEFINITION

Data visualization is the art and science of transforming data into a visual form suitable for human interpretation to support decision-making. Data visualization provides meaningful context and understanding to often complex data where identifying patterns, trends or outliers may otherwise not be easily achieved.

There are other terms closely related to the field of data visualization such as visual analytics, infographics, information visualization and data journalism.

Data visualization pulls in knowledge from several fields in order to be effective. This includes an understanding of the human brain and cognition processes all the way to data structure and data modelling. Add in aspects of user interface design and user experience and creating effective data visualizations becomes possible.

Data visualization can be employed for complex and basic data sets. Indeed, the focus should be on the outcomes that the visualization drives and not as much on the complexity of the data. 

It is part of many analytics and business intelligence processes including applications in business dashboards, reports, data science and statistical models.

Why is data visualization important?

When done right, data visualization can help transform organizations by delivering key information to decision-makers at the right time in an easily consumable format.

Effective data visualization enables people to take action based on the state or conditions of their environment. Data visualization makes it easy for people to interpret and understand their data, in an easily consumable way.

While data visualization is often the end of most data pipelines, it is arguably one of the most important and can be considered the last, but difficult, mile. Even the most advanced data science project will need to be communicated to an audience in order for value to be realized so it behooves most data professionals to at least have a foundational awareness of data visualization approaches.

2: Understand the audience that will consume the chart and graph tYPES

Without understanding who the audience is, their motivations and what challenges they have, greatly increases the chances of not picking the right chart of graph. 

Perhaps the audience is a group of financial auditors. If you know auditors, they absolutely have a preference for certain chart types such as tables and waterfall charts. Without this knowledge, picking the right chart for this audience is doomed to failure. 

The field of user experience has much to say on this area. A persona design approach should be followed at this stage in order to narrow down the focus of who is being designed for to understand preferences and design criteria.

3: Identify the business questions that they need answered with data

With the previous steps taken, picking the right charts and graphs has become considerably easier. But it is not enough. 

Aligned to both the goals from step one, the personas in step two, the set of questions that the data needs to answer must be identified. The questions must be of sufficient detail to assist both in the selection of the data but also in helping to pick the right chart.

For example "How many widgets have we sold so far this year in millions of units?", is a well-formed question that would lead to picking a certain chart type specific to that question.

4: Assess and gather the data for what questions can be answered

The previous steps assumed there were no constraints on what data was available. But now, the data reality must be addressed. Based on a data assessment, the set of questions to be answered are narrowed down to what is feasible in the short term. That is, what can be visualized now versus in later iterations.

Naturally, this must be communicated back to the audience to ensure awareness and alignment with the approach. This is covered extensively in the book Delivering Data Analytics.

5: Select the RIGHT chart or graph TYPE to answer the questions

By following the previous steps, the risk of incorrect chart selection has been greatly reduced, but it is not zero. There are two approaches to selecting the right chart from here.

Firstly, leverage the list of primary and secondary chart types for a simplified set of 12 visualizations to use for 90% of situations.

Alternatively, a chart selector could be used to break up the chart types by:

• Composition
• Comparison
• Distribution
• Relationship

Beyond the above, sometimes it is not possible to keep things simple and, rather than picking very complicated chart types, a combination of charts might be optimal.

6: Get feedback before starting development

Even when taking every precaution in selecting the right charts and graphs, it is still possible to get it wrong. That is why having a failsafe step is vital. The solution? Wireframing. 

Whether it is a low-fidelity wireframe such as sketches or using cards and PowerPoint templates or high-fidelity with Adobe XD or a whiteboarding tool, the idea is to get the visualization in a visual form so the audience can give their feedback.

This final step will ensure a tight alignment in picking the right charts and graphs for the intended audience.

KEEP THE CHART AND GRAPH TYPES SIMPLE!

When picking charts and graphs, place simplicity above all is. Is there an easier way to answer a given question? Is it broken down into the smallest parts? Where is the "So what?" and can it be answered more directly?

In data visualization, simplicity is king. That is the reason why limiting the number of charts to select from makes the task of picking the right chart that much easier.

THE 12 CHART TYPES THAT KEEP IT SIMPLE

The 3 Foundational Charts

There are three charts that form the basis of the 12 best chart and graph types. Indeed, the additional nine visualizations are simply variations of these foundational three. The foundation charts are:

1. Vertical bar graph or column chart
2. Horizontal bar graph
3. Line graph

By themselves, these charts and graphs should cover many situations and should be considered first. They are widely understood and people know how to interpret them, i.e. keep it simple.

THE 9 VARIATIONS OF THE FOUNDATIONAL 3

However, they can easily be extended and scaled as needed for representing more nuanced data and insights. Together, these 12 charts and graphs cover 90% of use cases when following the steps above.

These 12 primary charts form the basis of picking the right data visualization. While not as simple as the three foundational ones, they offer additional options for displaying and representing context and data with an acceptable reduction in simplicity.

However, what about the remaining 10% of situations where more advanced charts are a must and all attempts have already been made to simplify the message? For those situations, leverage the six secondary chart and graph types. A note here, depending on the domain and industry, there may be a need for greater use of these secondary charts.

The 6 secondary charts and graphs

For special cases, certain industries and domains, there are some more sophisticated data visualizations that need to be called upon. These are the treemap, bubble chart or bubble plot, bubble map, table, scatter plot and a variation of the horizontal bar for population, the butterfly chart.

CHART PATTERNS FOR WHEN ONE CHART IS NOT ENOUGH

Chart patterns take in the wider business context and focus on chart interactions and relationships to get to a sufficient level of detail in order to take action on. It aligns with more human-centric thinking and answering questions rather than considering chart categories.

Chart patterns are more accessible and do not require technical understanding of chart best practices as less emphasis is placed on getting a single chart type exactly right. The point of entry is through business context. For example, imagine this scenario:

EXAMPLE

THE CHART SELECTOR APPROACH

An alternative approach to picking the right chart type is via a chart selector, popularized by Andrew Abela, which breaks charts into several categories. Taking this approach, visualizations can be grouped by how the question needs to be answered. This list is limited to the charts recommended above as other chart types, while can be valuable, have limited application and require additional user training, in general.

BEST CHART AND GRAPH TYPES FOR Comparing Values 

While most types of data visualizations will allow comparison of two or more trends or data sets, there are certain graphs or charts that will make the message all the more powerful.

If the main goal is to show a direct comparison between two or more sets of information, the best choices would be:

• Column bar charts
• Horizontal bar graphs
• Line graphs
• Bubble charts
• Scatter plots

BEST CHART AND GRAPH TYPES FOR Composition OF VALUES

If the primary aim is to showcase the composition of the data – in other words, show how individual segments of data make up the whole of something – choosing the right types of data visualizations is crucial in preventing the message from becoming lost or diluted.

For parts of the whole and composition questions then consider:

• Column bar charts
• Stacked column charts
• 100% stacked column charts
• Horizontal bar graphs
• Stacked horizontal bar graphs
• 100% stacked horizontal bar graphs
• Treemaps

BEST CHART AND GRAPH TYPES FOR DISTRIBUTIONS

When working with a large number of "entities" such as products, customers, employees, distribution charts can understand the range of the measures in the data set as well as trends and patterns.

Effective data visualizations for distributions include:

• Scatter plots
• Bubble charts
• Column bar charts
• Line graphs

BEST CHART AND GRAPH TYPES  FOR RELATIONSHIPS ACROSS VALUES

Relationship charts and graphs can can help to identify outliers and assist in highlighting any areas that may need more urgent attention. Though use with caution as these charts can be a bit harder for untrained users to grapple with. 

In these cases, the most effective visualizations include:

• Scatter plots
• Bubble charts

BEST CHART AND GRAPH TYPES  FOR TRENDS OVER TIME

Trend charts are useful whenever a value or several values need to be measured over time. Should a question require historical context then a trend visualization is a solid option to consider.

For trends over time, the ideal chart types are:

• Line graphs
• Column bar charts
• Area charts

IN SUMMARY, THERE ARE THREE DIFFERENT APPROACHES TO PICKING THE BEST CHART AND GRAPH TYPES

The good news is that there are options, as covering in this post they are:

• Leverage the 12 primary charts and 6 secondary
• Use chart patterns for certain scenarios
• Follow the chart selector approach

Regardless of the framework used, none of them will be truly the "best" until the business context and requirements are properly understood. Check out this video on How to Pick the Right Chart Types.