1.4 Dashboard Basics
Dashboards are a helpful way to communicate and report data. They are versatile in that they support multiple types of reporting. Dashboards are predominantly used in business intelligence contexts, but they are being used more frequently to communicate data and visualize analysis for non-business services also. Popular dashboarding platforms include Tableau, and Power BI, although there are other options, such as Excel, R + Shiny, Geckoboard, Matillion, JavaScript, etc.
These technologies aim to make creating data reports as simple and user-friendly as possible. They are intuitive and powerful; creating a dashboard with these programs is quite easy, and there are tons of how-to guides available online [10], [17], [18].
In spite of their ease of use, however, dashboards suffer from the same limitations as other forms of data communication, to wit: how can results be conveyed effectively and how can an insightful data story be relayed to the desired audience? Putting together a “good” dashboard is more complicated then simply learning to use a dashboarding application.
1.4.1 Dashboard Fundamentals
Effective dashboarding requires that the designers answer questions about the planned-for display:
who is the target audience?
what value does the dashboard bring?
what type of dashboard is being created?
Answering these questions can guide and inform the visualization choices that go into creating dashboards.
Selecting the target audience helps inform data decisions that meet the needs and abilities of the audience. When thinking of an audience, consider their role (what decisions do they make?), their workflow (will they use the dashboard on a daily basis or only once?), and data expertise level (what is their level of data understanding?).
When creating a dashboard, its important to understand (and keep in mind) why one is needed in the first place – does it find value in:
helping managers make decisions?
educating people?
setting goals/expectations?
evaluating and communicating progress?
Dashboards can be used to communicate numerous concepts, but not all of them can necessarily be displayed in the same space and at the same time so it becomes important to know where to direct the focus to meet individual dashboards goals. Dashboard decisions should also be informed by the scope, the time horizon, the required level of detail, and the dashboard’s point-of-view.In general,
the scope of the dashboard could be either broad or specific – an example of a broad score would be displaying information about an entire organization, whereas a specific scope could focus on a specific product or process;
the time horizon is important for data decisions – it could be either historical, real-time, snapshot, or predictive:
historical dashboards look at past data to evaluate previous trends;
real-time dashboards refresh and monitor activity as it happens;
snapshot dashboards show data from a single time point, and
predictive dashboards use analytical results and trend-tracking to predict future performances;
the level of detail in a dashboard can either be high level or drill-able – high level dashboards provide only the most critical numbers and data; drill-able dashboards provide the ability to “drill down” into the data in order to gain more context.
the dashboard point of view can be prescriptive or exploratory – a prescriptive dashboard prescribes a solution to an identified problem by using the data as proof; an exploratory dashboard uses data to explore the data and find possible issues to be tackled.
The foundation of good dashboards comes down to deciding what information is most important to the audience in the context of interest; such dashboards should have a core theme based on either a problem to solve or a data story to tell, while removing extraneous information from the process.
1.4.2 Dashboard Structure
The dashboard structure is informed by four main considerations:
form – format in which the dashboard is delivered;
layout – physical look of the dashboard
design principles – fundamental objectives to guide design
functionality – capabilities of the dashboard
Dashboards can be presented on paper, in a slide deck, in an online application, over email (messaging), on a large screen, on a mobile phone screen, etc.
Selecting a format that suits the dashboard needs is a necessity; various formats might need to be tried before arriving at a final format decision.
The structure of the dashboard itself is important because visuals that tell similar stories (or different aspects of the same story) should be kept close together, as physical proximity of interacting components is expected from the viewers and consumers. Poor structural choices can lead to important dashboard elements being undervalued.
The dashboard shown in Figure 1.14 provides an example of group visuals that tell similar stories (the corresponding Power BI file can be found on the Data Action Lab website).
Figure 1.14: An exploratory dashboard showing metrics about various cities ranked on the Global Cities Index. The dashboard goal is to allow a general audience to compare and contrast the various globally ranked cities – statistics that contribute to a ‘higher’ ranking immediately pop out. Viewers can also very easily make comparisons between high- and low-ranking cities. The background is kept neutral with a fair amount of blank space in order to keep the dashboard open and easy to read. The colours complement each other (via the use of a colour theme picker in Power BI) and are clearly indicative of ratings rather than comparative statistics (personal file).
Knowing which visual displays to use with the “right” data helps dashboards achieve structural integrity:
distributions can be displayed with bar charts and scatter plots;
compositions with pie charts, bar charts, and tree maps;
comparisons use bubble charts and bullet plots, and
trends are presented with line charts and area plots.
An interesting feature of dashboard structure is that it can be used to guide viewer attention; critical dashboard elements can be highlighted with the help of visual cues such as use of icons, colours, and fonts. Using filters is a good way to allow dashboard viewers of a dashboard to customize the dashboard scope (to some extent) and to investigate specific data categories more closely.
The dashboard shown in Figure 1.15 provides an example of a dashboard that makes use of an interactive filter to analyze data from specific categories.
Figure 1.15: An exploratory dashboard showing information about the National Hockey League draft class of 2015. The dashboard displays professional statistics (as of August 2019) of hockey players drafted into the NHL in 2015, as well as their overall draft position. This dashboard allows casual hockey fans to evaluate the performance of players drafted in 2015. It provides demographic information to give context to possible market deficiencies during this draft year (i.e. defence players were drafted more frequently than any other position). This dashboard is designed to be interactive; the filter tool at the top allows dashboard viewers to drill-down on specific teams (personal file).
1.4.3 Dashboard Design
An understanding of design improves dashboards; dissonant designs typically make for poor data communication. Design principles are discussed in [1]–[3], [6], [8]. For dashboards, the crucial principles relate to the use of grids, white space, colour, and visuals.
When laying out a dashboard, gridding helps direct viewer attention and makes the space easier to parse; note, in Figure 1.14, how the various visuals are aligned in a grid format to lay the data out in a clean, readable manner.
In order to help viewers avoid becoming overwhelmed by clutter or information overload, consider leaving a enough blank space around and within the various charts; note, in Figure 1.15, that while the dashboard displays a lot of information, there is a lot of blank/white space between the various visuals, which provides viewers with space to breathe, so to speak. In general, clutter shuts down the communication process (see Figure 1.16 for two impressive examples of data communication breakdown).
Colour provides meaning to data visualizations – bright colours, for instance, should be used as alarm indicators as they immediately draw the viewer’s attention. Colour themes create cohesiveness, which improves the overall readability of a dashboard. There are no perfect dashboards – no collection of charts will ever suit everyone who encounters it.
That being said, dashboards that are elegant (as well as truthful and functional) will deliver a bigger bang for their buck [6], [7]. In the same vein, keep in mind that all dashboards are by necessity incomplete. A good dashboards may still lead to dead ends, but it should allow its users to ask: “Why? What is the root cause of the problem?”
Finally, designers and viewers alike must remember that a dashboard can only be as good as the data it uses; a dashboard with badly processed or unrepresentative data, or which is showing the results of poor analyses, cannot be an effective communication tool, independently of design.
1.4.4 Examples
Dashboards are used in varied contexts, such as:
interactive displays that allows people to explore motor insurance claims by city, province, driver age, etc.;
a PDF file showing key audit metrics that gets e-mailed to a Department’s DG on a weekly basis;
a wall-mounted screen that shows call centre statistics in real-time;
a mobile app that allows hospital administrators to review wait times on an hourly- and daily-basis for the current year and the previous year; etc.
The Ugly
While the previous dashboards all have some strong elements, it is a little bit harder to be generous for the two examples provided in Figure 1.16. Is it easy to figure out, at a glance, who their audience is meant to be? What are their strengths (do they have any)? What are their limitations? How could they be improved?
The first of these is simply “un-glanceable” and the overuse of colour makes it unpleasant to look at; the second one features 3D visualizations (rarely a good idea), distracting borders and background, lack of filtered data, insufficient labels and context, among others.
The Good
Good dashboards, on the other hand, simply breathe. The number of charts on each page is small, boxes are eschewed, simple colour schemes are preferred, the canvas is quiet, as in the example below, fro the University of Cincinnati.
Figure 1.17: A ‘zen’ dashboard; course evaluations at the University of Cincinnati [10].
More details on data storytelling and design principles will be presented in chapters (PDV-EVD) and 4.
Golden Rules
In a (since-deleted) blog article, N. Smith posted his 6 Golden Rules:
consider the audience (who are you trying to inform?does the DG really need to know that the servers are operating at 88% capacity?);
select the right type of dashboard (operational, strategic/executive, analytical);
group data logically, use space wisely (split functional areas: product, sales/marketing, finance, people, etc.);
make the data relevant to the audience (scope and reach of data, different dashboards for different departments, etc.);
avoid cluttering the dashboard (present the most important metrics only), and
refresh your data at the right frequency (real-time, daily, weekly, monthly, etc.).
With dashboards, as with data analysis and data visualization in general, there is no substitute for practice: the best way to become a proficient builder of dashboards is to … well, to go out and build dashboards, try things out, and, frequently, to stumble and learn from the mistakes.
We will revisit dashboards in chapter 5.
