Data-driven visualizations – design is half the battle

When designing visualizations, the same principles apply as with designing any website, mobile app, a product or a service whatsoever. What, to whom, why and how – these four basic questions are good guidelines for designing an effective visualization.

A visualization is usually considered as a process of transforming non-visual data into a visual form that can be perceived by the human visual system. The human visual system can perform a number of visual information processing tasks quickly and effectively, and that is because our brain can process multiple visual features in parallel. The processing happens automatically without the need for  conscious processing. Visualizations are a useful tool for analyzing and examining data, which helps the user to understand the information the data contains and thus enhance problem solving.

Designing of visualizations can be divided into four stages. During the first three stages, the aim is to define the goal of the visualization, and the fourth stage aims to answer to the question of how this goal can be achieved.

1. What information does the source data contain? What are the characteristics of it?
2. To whom is the visualization aimed for? What kind of context is the visualization meant to be used in?
3. Why is the visualization worth implementing? What is the purpose of it, what kind of needs it fulfils?
4. How can the goal for the visualization be achieved? What type of a visualization helps to achieve the goal best? What kind of visual coding is the most effective? And what kind of interaction supports the user to achieve his goals?

1. What?

At first, it is recommended to take a closer view into the source data at hand: what the data is about, where the data has been collected from and how and what information the data contains. It is essential to examine what types of variables the data contains and to identify the most important (or the most interesting) variables and the relations between them, in order to evaluate if the data is accurate enough for providing the visualization intended.  The source data often contains a lot of noise, such as divergent values, due to which the pre-processing of the data should also be planned carefully. It is also important to notice that if the data is oversimplified or cleansed too much something important can be lost.  

2. To whom?

Another important design task is to identify the target users to whom the visualization will be provided for. Which visualization type is the most appropriate depends on the purpose of the visualization and the target users: for example, whether the visualization is meant for a publication printed for a wide audience, for a group of experts to examine and understanding a phenomenon or for a limited target group to support decision making.  The way how the target users interpret visualizations depends on their unique personal characteristics like cognitive capacities, previous knowledge and experiences, motivational and cultural aspects and contextual factors, such as terms and practices typical for the domain. For developing understanding of the needs and characteristics of the target group, it is recommended to use end-user analysis.

3. Why?

The purpose of the visualization is the most important issue to keep in mind when designing visualizations. Pragmatic visualizations typically aim to support the user with examining and getting insight into the data and finding new information. However, the primary purpose of the visualization can also be to support the designer to transmit information, convince the user or to provoke discussion. In any case, the purpose of the visualization has to be described with as concrete actions as possible for defining what data should be represented in the visualization, which variables or relations between variables are the most important or interesting ones and, if needed, what kind of interaction supports the user to achieve his goals the best.

4. How?

It is much easier to select a proper visualization type and visual features to be used if there is a clear understanding of what kind of data is being visualized, who the target user is and what the purpose of the visualization is. When selecting the visual features, it is important to pay attention to the types of the variables displayed as well as to the general principles of visual information processing. For example, according to the laws of perception our brains automatically group together certain visual details and perceive them as a single unit. The visual contrasts, such as a deviant shape, size, color, position and orientation, are perceived quickly and automatically without the need for an active comparison of the details, which can be utilized for highlighting the essential information and drawing the user’s attention.

Designing is always worth investing: as a result of insufficient design the final visualization can be too complicated and thus uninterpretable or the information represented can be either too detailed or too general for decision makers to understand and use. An inadequate definition of the goal can lead to a visualization with wrong variables represented or wrong questions answered.

There is no one-size-fits-all solution to visualizing data. With every single case a separate end-user analysis has to be conducted. The four basic questions (what, to whom, why and how) represented above help the designer to find solutions to the design problems: what type of a visual representation is the most appropriate for transferring the intended information to the user or what kind of visual mapping or interaction techniques should be used. A good approach to validate the design solutions is to utilise user-centered design methods for getting feedback from the target users. It is important then that the visualization is based on real data.


Fig. 1. Both visualizations are based on exactly the same data. A bar chart is the most common graph type for comparing discrete data values within or across categories. However, line graphs can be less cluttered than a bar chart and thus be a better option for showing trends.

data-driven visualizationsdata-driven visualizations


Fig 2. When using a pie chart, the user has to estimate and compare pie wedge sizes, which is challenging. For comparing fractions with each other the standard bar graphs are preferred. The difference in length can be perceived quickly and automatically – much easier than with a pie chart.

data-driven visualizationsdata-driven visualizations

Source: Ministry of Economic Affairs and Employment, Employment Service Statistics 2018.