5. Metrics Representation and Dashboards

There are four levels of information, according to Savoie (2012): data, information, knowledge, and wisdom. Each level makes the previous one richer and more informative. The fist level is made up of facts with no meaning and is just input data. As stated by Few (2006), collecting, processing, and storing data is well explored, but there is “little progress in using that information effectively.” As a result, to present data, it has to be turned into information, which is the second level of processing. To turn data into information, it must be organized and structured with a shared meaning. For example, we can organize the data into rows and columns, assigning titles and descriptions to them, and the data becomes meaningful as information. Furthermore, dashboards are the most effective approach to convey information (Savoie 2012).

EIS (Thierauf 1991), scorecards, and KPIs (Kaplan et al. 1997) have been studied since the 1990s. They are all used to support decision-making, measure performance, and monitor execution of activities, and according to Few (2006) dashboards appear to be synonymous or simply a new name for them. However, nowadays, the definition of dashboards by Few (2004) is the most used and common:

It is evident from the definition that the dashboard’s visual appearance, as well as its objectives, is very significant. It should deliver appropriate and reliable content in a straightforward and understandable manner to the end-user. According to Few (2006), most dashboards fail to communicate efficiently and effectively due to poorly designed implementation rather than a lack of technology.

Visualization of dashboards is essential not because of its beauty but because it can communicate with end-users with greater efficiency and richer meaning than plain text (Few 2006). As a result, visualization becomes more science than art because a successful dashboard is informed design, not just cute gauges, meters, and traffic lights (Few 2006).

Furthermore, while considering dashboard objectives, it is evident that different end-users would have varied objectives, requiring the development of different dashboards for each group of users. As a result, tailored dashboards are becoming increasingly useful (Few 2006; Johanssen et al. 2019), customized to meet individual needs to improve communication and suit the needs of users. In particular, faceted analytical displays give distinct views of the same data to different users for the purpose of analysis (Few et al. 2007); in other words, the same data is displayed in different ways.

According to Eckerson (2010), there are three types of dashboards: operational, tactical, and strategic. Each type, from operational to tactical, has more complex data, which means that for the first type, simple data is used, and for the strategic type, more complex data for purposes of analysis.

Each type differs in its objectives, functionality, and end-users, which is the reason for different data abstraction levels. But even if they have different usages, they overlap in some sense, which means that the same dashboard can be used for various purposes and by other end-users. For example, front-line workers can use both tactical and operation types of dashboards and managers can use all three of them. Thus, each dashboard type has its own features, but there is no clear boundary between them (Table 5.1).

Furthermore, there are two subtypes of the operational level: detect-and-respond and incent-and-motivate. The first subtype relates to monitoring activities or optimizing processes. The second subtype is a dashboard to increase workers’ productivity by presenting the workers’ or team’s performance metrics.

Most operational level dashboards display metrics of low-level processes; they contain detailed data or sometimes are slightly summarized. Furthermore, the dashboards often offer only one level of data with no drill-down functionality and provide the most up-to-date information. The dashboards look like automobile dashboards with alerts, dials, and gauges.

Interaction with charts and tables is a regular feature at the operational level. Drill-down, filters, sorting, changing views, pop-ups (Few 2006), and other features make it easier to interact with the dashboard and improve the user experience.

The dashboards used by mid-level managers and their appearance are somewhere between operational and strategic levels, allowing users to keep track of different processes and data in one spot.

Furthermore, some research includes a fourth type: analytical. It contains a large amount of complex data and what-if scenarios to assist executives in their analysis and planning (Nadj et al. 2020).

In addition, dashboards can also be classified as static (read-only) and interactive. Despite that, Few (2006) states that providing different varieties of data is meaningless, and nowadays static dashboards are not relevant anymore. Interactive dashboards can help to handle chunks of information effectively considering the growth of complexity and amount of data (Nadj et al. 2020).

There are four main purposes of dashboards according to Pauwels et al. (2009): consistency, monitoring, planning, and communication.

However, Rahman et al. (2017) propose a fifth purpose of dashboards—analysis, which is used to analyze personal and companies’ performance and is similar to the monitoring purpose of Pauwels et al. (2009) but more precise.

In the study by Yigitbasioglu et al. (2012), the authors underline the importance of interaction with a dashboard. Their research on dashboard graphical user interfaces distinguished two categories of design features: visual and functional.

“Data-ink ratio is a parameter that defines the relationship of ink used to illustrate data to the overall ink leveraged to represent the chart” (Nadj et al. 2020).

An example of poor visual features is inappropriate use of color or data-ink ratio, which will distract or confuse users, for example, using yellow and red too often may attract attention and also may lead to the impossibility of highlighting errors or critical information (Nadj et al. 2020).

Green, yellow, and red should represent acceptable, satisfactory, and poor performance/alarms, respectively (Few 2006).

Such features are a step toward interactive dashboards, in which the user is actively participating in the data analysis process. However, the user’s effort to interact can lengthen the analysis process or have a negative impact on decision-making (Nadj et al. 2020).

An example of a poor functional feature is dashboards which fail to provide needed functionality to a user who cannot gain enough information (e.g., to analyze or plan) from dashboards due to a lack of these features.

Few (2006) also includes standard practices in his study, such as visual and functional features:

Common features:

In addition, in the study by López et al. (2021), the authors provide different views and different kinds of charts to analyze the performance and distinguish different levels of abstraction to visualize the same data. According to the study, high-level strategic indicators are effective and improve decision-making processes.

Innometrics is a project that tracks, evaluates, and analyzes the software development process. This system provides a visualization of users’ working activity statistics.

The project collects, analyzes, and visualizes users’ activity data and developers’ code quality metrics:

This project aims to keep track of, assess, and evaluate developers and the development process in general. A robust dashboard, which is “such a powerful management tool” (Pauwels et al. 2009), must be designed to visualize data as metrics and communicate with end-users effectively and efficiently. It effectively communicates with the target user by presenting data as information. As noted in the review, we present data in dashboards since it is the most efficient and effective approach to convey data as information and communicate with users. However, to design such a dashboard, we examined its goals and who the target audience is.

There are two types of end-users in the project, each with a different purpose, managerial level, and request for different types of information from dashboards. The project’s primary audience are developers and managers (Fig. 5.2).

As a result, the challenge was to deliver information to both types of users in an effective and efficient manner. We created two unique dashboard visualizations for these purposes, each with a different level of data abstraction (López et al. 2021) and visualization elements (Yigitbasioglu et al. 2012). To put it another way, it will be a faceted dashboard. It offers a distinct view for developers that will satisfy front-line workers’ requirements. For managers, it has another visualization with different objectives, and it delivers another piece of information using the same data. Moreover, considering that the project is growing and will have more metrics and graphics in the future, extensible architecture was developed.

To decide how to visualize data properly, we first examined the objectives of both types of users, and then selected a type of dashboard and appropriate visual and functional features.

Developers are the first group of users. They are front-line employees that require the most up-to-date performance metrics for monitoring rather than historical metrics for analysis. As a result, the dashboard’s purposes are analysis and monitor, and its types are operational, with subtype incent-and-motivate. Considering that, there are less historical data and more Process Performance strategic indicators (gauges) (López et al. 2021) in the dashboard. Furthermore, it is more static than interactive because unnecessary interactive features can distract users (Nadj et al. 2020); yet, it does provide some interactivity choices to improve user experience.

The purpose, type, and features for developers’ dashboard may be found in the Creftab (Table 5.3).

The second group of users is mid-level managers—decision-makers who monitor and analyze employees’ performance and development processes. Thus, the dashboard has monitor and communication purposes and tactical type. It is also fully interactive, making it more analytical and assisting users in better comprehending complex data. In Table 5.4 type, purposes, and features for managers’ dashboards can be observed.

Since operational and tactical types that we determined for both dashboards sometimes overlap, some features are shared (Eckerson 2010), especially visual features, which affect how effectively and efficiently data is presented. In Table 5.5 these features can be observed.

Another problem was to develop a dashboard that was extensible and maintainable as the project progressed and more metrics and graphics were added. The goal was to provide front-end extensibility that allowed developers to delete, add, and change widgets, graphics, and other visualized data easily, without relying on other layers. Several design patterns and principles can be used to create a dashboard on the front-end layer that is easily expandable and maintainable. The Mediator Design Pattern can be utilized in object-oriented programming (OOP) (Gamma et al. 1995).

The ReactJS framework was used to build the user interface. Thus, we rely on Functional Programming (FP) since the framework relies on Functional components for modern React applications (Banks et al. 2017), and it was not possible to use the Mediator Design Pattern in its pure form.

Functional components were used, which means that components were defined as functions (not as classes). Functions have different design patterns and principles than OOP. We defined low-level functions and then applied composition (Banks et al. 2017). With the help of this, simple standalone functions are consolidated into one big dashboard system.

Moreover, we took the idea from the mediator pattern and created a mediator component that comprised low-level function components (Wieruch 2022). The mediator component took care of preprocessing data (turning data into metrics) and setting configuration data for low-level functions. After configuration components create cards with the following fields for each widget: The mediator component distributes data to low-level components based on the type field. As a result, low-level graphical representation components deal only with ready input data and can be reused (see Fig. 5.3). Thus, only configuration components can be updated to maintain the dashboard.

In Fig. 5.3 the structure of the dashboard can be observed:

As a result, updating existing widgets or adding new graphics can be accomplished in a few simple steps, which aligns with our goal of creating a dashboard that is easy to extend and manage.

Steps to edit or add new widgets:

As the software development process and its evaluation become more complex, understanding the purposes, types, and visualization approaches of tailored dashboards for managers and developers becomes an essential topic. Hence, it will improve the development of metrics presentations that provide end-users with more relevant data. After reviewing various works, it became evident that the dashboard requires specific representation for managers and developers, taking into account their respective goals and requirements. As a result, types, objectives, and aesthetic aspects were examined in order to design a dashboard that was suited to each type of end-user. On top of that, the dashboard’s architecture was designed to be easily maintainable and extensible since it was one of the research’s aims. A deep understanding of dashboard types, purposes, visuals, and architecture allows assessing the development process and presenting metrics to end-users more accurately and in a practical manner.