Information Dashboards and Tailoring Capabilities - A Systematic Literature Review

Received July 28, 2019, accepted August 1, 2019, date of publication August 6, 2019, date of current version August
21, 2019.
Digital Object Identifier 10.1109/ACCESS.2019.2933472
Information Dashboards and Tailoring

Capabilities - A Systematic Literature Review
ANDREA VÁZQUEZ-INGELMO 1, FRANCISCO J. GARCÍA-PEÑALVO 1, (Member, IEEE),
AND ROBERTO THERÓN 1,2
1 GRIAL Research Group, Computer Science Department, Research Institute for Educational Sciences, University of Salamanca, 37008 Salamanca, Spain
2 VisUSAL Research Group, University of Salamanca, 37008 Salamanca, Spain
Corresponding author: Andrea Vázquez-Ingelmo ([email protected])
This work was supported in part by the Spanish Ministry of Education and Vocational Training through the FPU Fellowship under Grant
FPU17/03276, in part by the Spanish Government Ministry of Economy and Competitiveness throughout the DEFINES Project under
Grant TIN2016-80172-R, in part by the PROVIDEDH Project, funded within the CHIST-ERA Programme, MINECO, Spain, through the
National Grant Agreement, under Grant PCIN-2017-064, and in part by the Ministry of Education of the Junta de Castilla y León, Spain,
throughout the T-CUIDA Project under Grant SA061P17.
ABSTRACT The design and development of information dashboards are not trivial. Several factors
must be accounted; from the data to be displayed to the audience that will use the dashboard. However,
the increase in popularity of these tools has extended their use in several and very different contexts
among very different user profiles. This popularization has increased the necessity of building tailored
displays focused on specific requirements, goals, user roles, situations, domains, etc. Requirements are
more sophisticated and varying; thus, dashboards need to match them to enhance knowledge generation
and support more complex decision-making processes. This sophistication has led to the proposal of new
approaches to address personal requirements and foster individualization regarding dashboards without
involving high quantities of resources and long development processes. The goal of this work is to present
a systematic review of the literature to analyze and classify the existing dashboard solutions that support
tailoring capabilities and the methodologies used to achieve them. The methodology follows the guidelines
proposed by Kitchenham and other authors in the field of software engineering. As results, 23 papers about
tailored dashboards were retrieved. Three main approaches were identified regarding tailored solutions:
customization, personalization, and adaptation. However, there is a wide variety of employed paradigms and
features to develop tailored dashboards. The present systematic literature review analyzes challenges and
issues regarding the existing solutions. It also identifies new research paths to enhance tailoring capabilities
and thus, to improve user experience and insight delivery when it comes to visual analysis.
INDEX TERMS SLR, systematic literature review, tailoring, custom, personalized, adaptive, information
dashboards.
I. INTRODUCTION Although identifying what is and what is not an informa-

Information dashboards are nowadays key tools for under- tion dashboard can be confusing in some cases, an informa-
standing and extracting knowledge from large datasets, but tion dashboard can be defined as a set of (visual) resources
they can take many forms. Information dashboards can be that enable its audience to understand and/or reach insights
employed for different goals, to analyze different datasets regarding the data being displayed [1]–[3].
(framed within different domains), to explain concepts, Their capabilities not only try to cover the exploitation of
to generate knowledge, to confirm hypotheses, etc., [1]. The datasets but also to provide a proper user experience to ease
spread of dashboards and their use in different contexts makes knowledge discovery. However, user experience, as the name
their definition a complex task. suggests, depends on each user, there is no ‘‘one size fits all’’
in this domain. Although a ‘‘one size’’ dashboard, valid and
The associate editor coordinating the review of this manuscript and useful for every possible user profile would be ideal, it is
approving it for publication was Osama Sohaib. utopic; not every user is driven by the same goals, not every
VOLUME 7, 2019 This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see http://creativecommons.org/licenses/by/4.0/ 109673
A. Vázquez-Ingelmo et al.: Information Dashboards and Tailoring Capabilities - A SLR
user is interested in the same data, not every user has the same exactly know which configuration is the best to accomplish
visualization literacy, and so on. These aspects include not their goals [7].
only personal preferences, but also social factors, like biases, It is clear that dashboards are valuable but sophisti-
beliefs, or past experiences [4], [5]. cated tools, and their potential benefits when supporting
The support that technology provides to our everyday decision-making processes has increased their popularity in
life has led to an exponential growth of data, making it several fields (business intelligence, learning analytics, ser-
necessary and crucial to take advantage of information to vices monitoring, etc.) and activities. Sarikaya et al. shown
perform informed decision-making processes. Data are more in a recent survey the relevance of researching on these tools
accessible, and thus, not only specific profiles are in charge and the relevance of users’ goals, their characteristics, and
of visual analyses. Some users might need solutions that context for designing useful dashboards [1]. However, before
not only let them configure or develop dashboards given tackling how a tailored dashboard can be efficiently delivered
their requirements, but also that assist them in choosing a to a specific user, it is necessary to understand and explore
proper configuration if they don’t have enough experience existing research lines and solutions regarding this domain.
with visual analyses or enough visual literacy. Users should Laying a foundation on tailored dashboards can help to design
be provided with tailored dashboards that fulfill their require- better solutions based on case studies found in literature,
ments and foster insight delivery to enhance the outcomes of analyzing their strengths and weaknesses.
the decisions made. A systematic literature review of existing tailoring meth-
Given these facts, it is essential to take into account final ods regarding information dashboards has been carried out
users when developing information dashboards, to improve to clarify this matter. Through this review, the authors aim
the user experience and subsequently provide a dashboard at providing a comprehensive view of this domain’s solu-
that promotes knowledge generation. The user-centered tions, to examine new research paths and opportunities for
design paradigm tries to address these issues by focusing on delivering effective tailored dashboards, and to learn about
the user needs and requirements during all the development the trends and methods regarding the problem of finding a
phases [6]. Involving the end-user into the design processes suitable dashboard configuration given a concrete user. Also,
supports the development of better systems, which are useful this systematic literature review can help to identify caveats or
for them and match their needs. research opportunities to improve tailoring processes and to
While necessary, this paradigm still lacks individualism obtain more practical, usable, and individualized dashboards
when providing a solution, as not every potential user of subsequently.
the system can be involved in a development process. These The term ‘‘tailored dashboard’’ is used throughout this
potential users can present very different characteristics, men- work to enclose any dashboard solution that can vary its
tal schemas, and goals and therefore can demand very dif- appearance and functionalities to match the users’, data’s
ferent features, especially in the dashboards domain that is and context’s requirements, be them explicit requirements or
faced in this work, given its complexity, so each person should implicit requirements. A general term is necessary, because
be taken into account. However, is it efficient to develop an using ‘‘customizable,’’ ‘‘personalized’’ or ‘‘adaptive’’ indis-
individual dashboard for each user? Should several quantities tinctly to refer to these solutions, could lead to misconcep-
of resources be involved for the benefit of individualism? tions around these last terms, which, in the end, have different
There exist any other approaches for designing and for build- nuances.
ing information dashboards for several and different user As it will be exposed, tailored dashboards can be catego-
profiles? rized taking into account a series of factors like the stage at
Personalization and customization approaches try to which the tailoring process is performed, the driver of the
address these individualization issues by tailoring products tailoring process, the targets of the tailoring process, etc. This
through different mechanisms. These mechanisms aim at sup- categorization shows that although the outcomes are ‘‘the
porting developers to configure products by reusing compo- same’’ (tailored dashboards), the methods to provide them
nents and consequently, by decreasing the development time can differ from each other (customization, personalization,
(even by assisting users in configuring their own products adaptation, etc.).
driven by their own needs). In the case of dashboards, there The rest of this work is organized as follows. Section two
exist user-friendly tools that enable users to create and cus- (Methodology) describes the methodology, and the steps fol-
tomize their dashboards without requiring any programming lowed to perform the review. Section three (Review Plan-
skills, like Tableau1 or Grafana2 . This kind of approaches ning) details the SLR planning phase. Section four (Review
give freedom to the users to configure their tools, but in such a Process) presents the review and data extraction steps.
complex domain that is visual analytics, some users might not Section five (Results) presents the results obtained from the
analysis of the selected works to answer the research ques-
tions. Section six (Discussion) discusses the results, followed
1 https://www.tableau.com/ by section seven (Threats to Validity), in which the threats to
2 https://grafana.com/ the validity of the review are outlined. Finally, section eight
109674 VOLUME 7, 2019

(Conclusions) includes some conclusions and future research •RQ6. How mature are tailored dashboards regarding
lines. their evaluation?
The first RQs block aims at answering questions regarding
II. METHODOLOGY how tailoring capabilities have been materialized in tangible
A systematic process has been followed to conduct the dashboard solutions (methods, requirements management,
present literature review; specifically, the systematic liter- domain transferability). The goal of answering RQ5 is to
ature review (SLR) methodology by Kitchenham [8] and identify research opportunities in terms of the application
Kitchenham and Charters [9]. The SLR has been comple- of AI mechanisms to support the dashboards’ tailoring pro-
mented with a systematic mapping of the literature following cesses automatically. The last question’s purpose is to under-
the methodology proposed in [10]. The mapping results can stand if the solutions found have been tested with end-users
be consulted in [11]. In this section, the protocol followed and if the tailoring capabilities have been useful for enhancing
in carrying out the SLR is described, providing all the nec- insight delivery and knowledge generation.
essary information to trace the subsequent results. Following As mentioned before, the SLR has been complemented
the [8], [9] guidelines, the SLR is composed of three main with a literature mapping to perform a quantitative analysis
phases: planning, conducting, and reporting the study. These of the domain and to obtain a broad view of the research area.
phases are detailed through the following sections. The following mapping questions (MQs) were posed, but the
Before planning the present SLR, a preliminary search was outcomes of the mapping are out of the scope of this paper
made to verify that no recent SLRs about tailored dashboards and can be consulted at [11]:
were carried out. If that were the case, there would not be any • MQ1. How many studies were published over the years?
necessity to conduct a new one. This verification was per- • MQ2. Who are the most active authors in the area?
formed by searching through different electronic databases • MQ3. What type of papers are published?
(Scopus, Web of Science (WoS), IEEE Xplore and Springer) • MQ4. To which contexts have been the variability pro-
terms related to the methodology (‘‘SLR’’, ‘‘systematic liter- cesses applied? (BI, learning analytics, etc.)
ature review’’, etc.) and the target of the review (‘‘tailored’’, • MQ5. Which are the factors that condition the dash-
‘‘customizable’’, ‘‘personalized’’, etc., along with the term boards’ variability process?
‘‘dashboards’’). The outcomes of these queries confirmed that • MQ6. What is the target of the variability process?
currently, there are no previous systematic literature reviews (visual components, KPIs, interaction, the dashboard as
about the thematic addressed in this work, justifying the a whole, etc.)
execution of this SLR. • MQ7. At which development stage is the variability
achieved?
III. REVIEW PLANNING • MQ8. Which methods have been used for enabling vari-
The review planning process involves the identification and ability?
definition of different aspects to lay the foundations of • MQ9. How many studies have tested their proposed
the review execution, such as posing the questions to be solutions in real environments?
answered, detailing the protocol followed, and any other The systematic mapping performed at [11] employs the
relevant information to make the review traceable. These same approach as in the present SLR. However, the mapping
different aspects are described in this section. provides an overview of the research area by identifying and
classifying the available evidence, while the following SLR
A. RESEARCH QUESTIONS results involve the analysis and interpretation of the evidence
First, a series of research questions have been raised. These found [12] to answer the specific research questions posed at
questions can be classified into three main blocks: technical the beginning of this subsection.
aspects (RQ1-RQ4), artificial intelligence (AI) application Given the previous research questions, the PICOC method
(RQ5), evaluation of the solutions (RQ6). proposed by Petticrew and Roberts [13] has been followed to
• RQ1. How have existing dashboard solutions tackled the define the review scope.
necessity of tailoring capabilities? • Population (P): Software solutions
• RQ2. Which methods have been applied to support • Intervention (I): Provide support to tailor (information)
tailoring capabilities within the dashboards’ domain? dashboards
• RQ3. How the proposed solutions manage the dash- • Comparison (C): No comparison intervention in this
board’s requirements? study, as the primary goal of the present SLR is to ana-
• RQ4. Can the proposed solutions be transferred to dif- lyze existing approaches regarding tailoring capabilities
ferent domains? and gain knowledge about them.
• RQ5. Has any artificial intelligence approach been • Outcomes (O): Information dashboard proposals
applied to the dashboards’ tailoring processes and, • Context (C): Environments related to data visualization
if applicable, how these approaches have been involved and (or) decision making (in the academia, industry,
in the dashboards’ tailoring processes? etc.)
VOLUME 7, 2019 109675

B. INCLUSION AND EXCLUSION CRITERIA of Science (WoS), IEEE Xplore, and SpringerLink. These
Once the scope of the review has been established, a series databases were chosen according to the following criteria:
of inclusion (IC) and exclusion criteria (EC) are defined to • It is a reference database in the research scope
select relevant works for answering the identified research • It is a relevant database in the research context of this
questions. If a work does not meet the whole set of inclu- literature review
sion criteria or does meet any exclusion criterion, it will be • It allows using similar search strings to the rest of the
excluded from the review. selected databases as well as using Boolean operators to
• IC1.The paper describes a dashboard solution (proposal, enhance the outcomes of the retrieval process
architecture, software design, model, tool, etc.) AND Regarding the search concepts employed to build the
• IC2. The solution is applied to information dashboards search query, the following terms were included:
AND • The ‘‘meta-dashboard’’ concept to search for solutions
• IC3. The solution supports or addresses tailoring capa- that employ a meta-modeling approach to extract com-
bilities (customization, personalization, adaptation, vari- mon and abstract features from dashboards that can be
ation) regarding information dashboards AND applied for tailoring processes.
• IC4. The tailoring capabilities of the dashboard are • Related terms to tailoring capabilities: tailored, cus-
related to its design, components or KPIs AND tomized, personalized, adaptive, flexible, configurable,
• IC5. The papers are written in English or Spanish AND context-aware, etc., along with the word ‘‘dashboard,’’
• IC6. The papers are published in peer-reviewed Jour- which is the main target of the review.
nals, Books or Conferences AND • Other terms like ‘‘selection,’’ ‘‘composition,’’ or ‘‘gen-
• IC7. The publication is the most recent or complete of eration’’ to search for generative solutions that provide
the set of related publications regarding the same study dashboards as a result of a generation, composition or
The exclusion criteria are derived from the inclusion crite- selection process of suitable visualizations and features.
ria as their opposite. • The term ‘‘template’’ to retrieve works that use dash-
board templates that can be configured to fit specific
• EC1.The paper does not describe a dashboard solution requirements (this term can also be related to generative
(proposal, architecture, software design, model, tool, processes)
etc.) OR • The term ‘‘driven’’ to enclose works that use context-
• EC2. The solution is not applied to information dash- driven, data-driven, user-driven, etc., approaches, thus
boards OR being necessary to take into account these factors to
• EC3. The solution does not support or address tailoring develop the dashboards
capabilities (customization, personalization, adaptation, • Additional terms related to heterogeneous requirements
variation) regarding information dashboards OR and diverse necessities to retrieve works that do not men-
• EC4. The tailoring capabilities of the dashboard are not tion directly any of the above terms, but do implicitly
related to its design, components or KPIs OR refer to them by calling upon the heterogeneity of dash-
• EC5. The papers are not written in English or Spanish boards requirements and the involved user profiles, thus
OR potentially addressing these issues by tailoring mecha-
• EC6. The papers are not published in peer-reviewed nisms.
Journals, Books or Conferences OR
• EC7. The publication is not the most recent or complete Finally, given the fact that the word ‘‘dashboard’’ is also
of the set of related publications regarding the same employed for referring to cars’ control panels, words related
study to the automotive area (‘‘car,’’ ‘‘vehicle,’’ ‘‘automotive’’)
were excluded to avoid irrelevant papers outside the scope
The IC5 includes the Spanish language, because the main of information dashboards.
research terms, as it will be seen in the next subsection, are
compatible with their Spanish equivalent terms (dashboard∗ D. QUERY STRINGS
along with custom∗ , personal∗ , adapt∗ , flexib∗ and config∗ ). The search strings for each chosen source were built using
As a consequence, works written in Spanish could be relevant search terms derived from the PICOC methodology
retrieved through the search string and could be potentially outcomes, connected by Boolean AND / OR / NEAR opera-
included in the review given the authors’ comprehension of tors. Moreover, the wildcard (∗ ) was used to enclose both the
this language. singular and plural of each term.
The NEAR operator enables the user to retrieve works
C. SEARCH STRATEGY where the terms joined by this operator are separated by an
It is necessary to identify the most important databases interval of words explicitly specified. This operator is handy
regarding the research context in which the queries will be in the context of the present research, as the terms ‘‘customiz-
performed to obtain relevant outcomes from the search. In this able,’’ ‘‘personalized,’’ ‘‘adaptive,’’ etc. should only refer to
case, four electronic databases were selected: Scopus, Web the dashboard term, to avoid works that are not explicitly
109676 VOLUME 7, 2019

focused on the tailoring capabilities of dashboards. However, (requirement∗ OR stakeholder∗ OR user∗ OR need∗ OR task∗
the drawback of using this operator is the necessity to explic- OR necess∗ ))))
itly define the maximum number of words that can separate In case of SpringerLink, this query string was comple-
the target terms. mented with an additional restriction, given SpringerLink
In this case, the chosen number was 10 (i.e., the ‘‘dash- policy of searching the query terms along with the papers’
board’’ term and the rest of the terms will be within 10 number full-text (which includes huge amounts of noise to the review
of words of each other). This number was selected after process). Through the advanced search tool, the query results
performing a ‘‘simulation’’ by executing the same search with were limited to those that have the term ‘‘dashboard∗ ’’ in
different proximity values (5, 7, 10, and 12). Examining the their titles, additionally to the search string terms in their
titles, abstracts and keywords of the additional records found full-texts to ensure that the main focus of the retrieved works
after incrementing this value, it was concluded that the ten is information dashboards.
value would retrieve relevant works without adding noise
(i.e., irrelevant works), meaning that the terms affected by the E. QUALITY CRITERIA
NEAR operator are potentially in the same sentence, given Although the inclusion and exclusion criteria are useful for
average sentence length guidelines and evidence [14], [15]. including in the review relevant works in terms of the scope
Once the NEAR operator value was selected, the specific of the literature review, they don’t address the quality of the
query strings for each chosen database were specified using retrieved papers regarding their capacity to answer the posed
their query syntax. research questions. A new set of criteria has been defined
to check the works’ quality before including them into the
1) SCOPUS final literature review. Each criterion can be scored with
TITLE-ABS-KEY ((meta-dashboard∗ ) OR ((dashboard∗ ) three values: 1 (the paper meets the criterion), 0.5 (the paper
W/10 (custom∗ OR personal∗ OR adapt∗ OR flexib∗ partially meets the criterion) and 0 (the paper does not meet
OR config∗ OR tailor∗ OR context-aware OR generat∗ the criterion).
OR compos∗ OR select∗ OR template∗ OR driven)) OR 1. The research goals of the work are focused on address-
((dashboard∗ ) AND ( (heterogeneous OR different OR diverse ing the variability, adaptability, customization or per-
OR dynamic) W/0 (requirement∗ OR stakeholder∗ OR user∗ sonalization of an information dashboard to improve
OR need∗ OR task∗ OR necess∗ )) )) AND NOT TITLE-ABS- individual user experience (UX)
KEY ( car OR vehicle OR automo∗ ) AND NOT DOCTYPE(cr)
• Partial: not every research goal tries to address UX
through tailoring capabilities
2) WEB OF SCIENCE
TS=((meta-dashboard∗ ) OR ((dashboard∗ ) NEAR/10 2. A software solution that supports the variability of the
(custom∗ OR personal∗ OR adapt∗ OR flexib∗ OR config∗ dashboard components is presented
OR tailor∗ OR context-aware OR generat∗ OR compos∗ • Partial: the software supports customization of the
OR select∗ OR template∗ OR driven)) OR ((dashboard∗ ) dashboard but is not the focus
AND ((heterogeneous OR different OR diverse OR dynamic) 3. A model, framework, architecture or any software engi-
NEAR/0 (requirement∗ OR stakeholder∗ OR user∗ OR need∗ neering artifact that address the variation of the dash-
OR task∗ OR necess∗ )))) NOT TS= (car OR vehicle OR board components and interaction methods are prop-
automo∗ ) erly exposed
• Partial: a model, framework, architecture or any
3) IEEE XPLORE software engineering artifact is exposed but not
(((meta-dashboard) OR ((dashboard) NEAR/10 (custom∗ OR detailed, i.e., the nature of the referred elements is
personal∗ OR adapt∗ OR flexib∗ OR tailor OR tailored OR mentioned, but their internal structures and details
configurable OR context-aware OR generation OR generated are not further explained.
OR generative OR composed OR composition OR selection
4. The employed methods or paradigms to achieve tailor-
OR selecting OR template OR driven)) OR ((dashboard)
ing capabilities are properly described
AND ((heterogeneous OR different OR diverse OR dynamic)
• Partial: the employed methods or paradigms
NEAR/0 (requirement OR stakeholder OR user OR need
OR task OR necessities)))) AND NOT (car OR vehicle OR to achieve tailoring capabilities are partially
automo∗ )) described, i.e., the methodology is mentioned,
but how the methodology has been particu-
4) SPRINGERLINK larly used within the application context is not
((meta-dashboard∗ ) OR ((dashboard∗ ) NEAR/10 (custom∗ detailed.
OR personal∗ OR adapt∗ OR flexib∗ OR config∗ OR tailor∗ 5. The context or domain of application of the dashboard
OR context-aware OR generat∗ OR compos∗ OR select∗ is described
OR template∗ OR driven)) OR ((dashboard∗) AND ((het- • Partial: the context or domain of application is
erogeneous OR different OR diverse OR dynamic) NEAR/0 mentioned but not detailed.
VOLUME 7, 2019 109677

FIGURE 1. Phases and outcomes of the review process using the PRISMA flow diagram.
6. The proposed solution has been tested with real users IV. REVIEW PROCESS
The data gathering process to conduct the present SLR has
• Partial: real users have used it and tested its func-
been divided into different phases in which various activities
tionality, but no further testing has been performed
are carried out. The PRISMA flow diagram [16] has been
7. Issues or limitations regarding the proposed solution employed to detail the actions performed during the data
are identified extraction (Figure 1).
Once the search was performed (on January 22, 2019),
• Partial: issues or limitations are mentioned but not
the paper selection process was carried out through the fol-
detailed
lowing process:
Each paper can obtain a maximum of 7 points regarding 1. The raw results (i.e., the records obtained from each
its quality following this methodology. This 0-to-7 score was selected database) were gathered in a GIT reposi-
transformed into a 0-to-10 scale, and the seven value was tory3 [17] and arranged into a spreadsheet4 . A total
chosen as the threshold for including a paper into the final of 1034 papers were retrieved: 254 from Web of Sci-
synthesis. If in a 0-to-10 scale, a paper obtains a score of fewer ence, 501 from Scopus, 97 from IEEE Xplore and
than seven points, it will be dismissed from the review as it 183 from SpringerLink.
did not meet a minimum quality to answer the stated research 2. After organizing the records, duplicate works were
questions. removed. Specifically, 348 records were removed,
The chosen threshold ensures that the works have obtained retaining 686 works (66.34% of the raw records) for
the maximum score in some criteria, without neglecting the the next phase.
rest of the quality statements. With this threshold, a paper is
limited to a maximum of two criteria with a 0 score to reach
the next phase, ensuring that the majority of the criteria is 3 https://github.com/AndVazquez/slr-tailored-dashboards
4 http://bit.ly/2wRCU5w
always fully or partially met.
109678 VOLUME 7, 2019

3. The maintained papers were analyzed by reading their

titles, abstracts, and keywords and by applying the
inclusion and exclusion criteria. A total of 613 papers
were discarded as they didn’t meet the criteria, retain-
ing 73 papers (10.79% of the unique papers retrieved)
for the next phase.
4. The selected 73 papers were read in detail and further
analyzed. The papers were scored regarding their qual-
ity to answer the research questions using the quality
assessment checklist described in the previous section.
One paper was added after checking the references of
the assessed works, leaving 74 records for this quality
assessment phase.
5. After applying the quality criteria, a total of 23 papers
(3.35% of the unique papers retrieved and 31.08%
of the full-text assessed papers) were selected for the FIGURE 2. Classification of the retrieved solutions in terms of their
present review. tailoring method. Source: [11], elaborated by the authors.
Two records were finally discarded. The reason for this
exclusion was that the two works were previous versions of
other studies found within the retrieved records. The decision involve generative or automatized approaches through the
was to keep the more complete and/or more recent work. specification of configuration files [21], [22], [24], mod-
els [23], [25] or pre-defined templates [20]. Although techni-
V. RESULTS cally the tailoring process is indeed made by the system (not
A. HOW HAVE EXISTING DASHBOARD SOLUTIONS involving direct user actions to modify the dashboard appear-
TACKLED THE NECESSITY OF TAILORING CAPABILITIES? ance), which is a characteristic of personalization approaches,
The first research question tries to answer, which are the the data contained within the configuration files or model
trends when it comes to tailoring an information dashboard. instances does involve explicit user requirements, so the
As stated in the introduction of this work, some terms are dashboard is tailored according to the users by means of
misleading or not being appropriately used, given their for- their requirements. In the end, these generative approaches
mal meaning. ‘‘Custom’’ and ‘personalized’’ are often used add an abstraction layer which helps users to configure their
as interchangeable terms with the same connotations. It is dashboards without requiring programming skills. For these
important to make distinctions among these terms, as they reasons, these solutions are also classified as customizable
have entirely different meanings regarding their mechanisms. dashboards.
The selected works were categorized in terms of their Despite the previous distinction about customizable
tailoring process. Each paper was analyzed to answer the dashboards, personalized solutions have also been identi-
questions that would frame the tailoring process employed fied. In this case, personalized solutions infer a suitable
(i.e., at which stage is the tailoring process performed? Who configuration based on implicit data about users, tasks,
performs the tailoring process?). As shown in Figure 2, or goals [28]–[31]. In [28], the methodology takes as input a
the majority of the selected works are framed in the category model of the business process and goals to describe and gen-
of ‘‘customizable,’’ meaning that the tailoring process of the erate a dashboard, so the authors use implicit data (goals) to
dashboard is driven by explicit user requirements [18]–[27]. build a concrete dashboard that would help to reach the input
Most customizable solutions identified involve manual goals. User-roles are also added to this methodology in [29],
approaches (which will be detailed in RQ2), meaning that to include more information to the dashboard personalization
users need to perform a set of explicit actions to tailor their process. A similar solution is presented in [30], which also
dashboard according to their needs. takes into account user-roles and business’ KPIs to generate
In [18], a customizable dashboard display for monitoring a dashboard that fits the business goals. Finally, in [31], the
mobile energy is presented; users can build their dashboards focus is on personalizing the display taking into account the
by selecting pre-defined widgets and data streams from dif- user abilities through an initial questionnaire that ask users
ferent sources (sensors, government agencies, social media, if they have eye diseases or any tremor in hands, making
and generic services). Manual approaches like those above the dashboard accessible if necessary. Once generated, these
are also used in [19], [26], [27], in which the customizability dashboards cannot be adapted at run-time, being essential to
capacity is based on the possibility of arranging the compo- re-generate them.
nents of the dashboard through end-user interaction, and even Adaptive solutions, on the other side, can adapt themselves
the ability to craft custom indicators, as described in [26]. at run-time based on environmental changes. Belo et al. [32]
However, not only manual user interactions are employed present an adaptive dashboard that restructures itself given
for arranging the tool, some of these customizable dashboards user-profiles and behaviors extracted from the dashboards’
VOLUME 7, 2019 109679

analytical sessions (i.e., through the analysis of the user

queries). Another adaptive solution presented in [33] uses
a dashboard generator fed with user, data and visualization
models, thus generating information dashboards according
to different contexts, and, in theory (as the proof of concept
is not fully adaptive at the time of publishing the paper),
adapting themselves given their users’ interaction history.
Other two kinds of tailored solutions have been identified,
as they cannot be framed on the last categories (customiz-
able, personalized, or adaptive). On the one hand, solutions
identified as ‘‘hybrid’’ are mainly personalized or adaptive
dashboards that allow the user to have the last word regarding
the dashboard configuration, or need user actions to com-
plete the tailoring process. In [34], a device cloud platform FIGURE 3. Customizable dashboard workflow.
dashboard is built based on the data model of the remote
devices being monitored, but users can also customize it
manually. Van Hoecke et al. [35] use a semantic reasoner
to personalize indicators from available data sources, but the
dashboard construction is still a user task. Santos et al. [36]
also proposes personalized dashboards based on knowledge
graphs and indicator ontologies, but allow the users to modify
the dashboard recommendation to her or his preferences.
On the other hand, there are customizable solutions that can
assist and help the users to build their dashboards according
to a series of factors. The four papers identified in this cate-
gory [37]–[40] use visual mapping to help users to determine
the best visualization types for the data to be visualized while
building and designing their dashboards. These solutions are
mainly customizable dashboards with mechanisms that help FIGURE 4. Customizable dashboard with system assistance workflow.
users with the selection of a suitable dashboard configuration.

Classifying these tools regarding their tailoring capabilities
is complex, as the selected papers present too many different Configuration wizards are also the preferred method for
solutions implemented through various methods with differ- customizable dashboards with system assistance, in conjunc-
ent goals, so this classification of tailored dashboards should tion with visual mapping methods that ease the selection of
be seen as a spectrum, allowing the existence of dashboards visualization types given the data types or structure [37]–[40].
that mix features of different approaches. However, framing Visual mapping is a transformation that matches data proper-
them in distinct categories, allow better understanding regard- ties with visual marks or visual elements to obtain a suitable
ing existing solutions as well as regarding the current state of visualization for the selected data [41]. Figure 4 shows a
the present field. generic workflow of how this approach work; users config-
ure their dashboards based on their needs, and the system
provides feedback to support the customization process and
B. WHICH METHODS HAVE BEEN APPLIED TO SUPPORT to obtain more effective dashboards potentially.
TAILORING CAPABILITIES WITHIN THE Another common method to customize dashboards
DASHBOARDS’ DOMAIN? is to configure them by using structured configuration
This research question is tightly related to RQ1. In the end, files [21], [22], [24], which also allow users to tailor their
the selected tailoring process narrows the potential methods dashboards with a higher level of abstraction (through
to accomplish it. As shown in the first research question, JSON files, XML files, etc.) through richer and more
the most common type of tailored dashboards are customiz- domain-specific syntaxes than programming languages.
able dashboards, within the scope of this systematic literature Figure 5 shows the workflow of these configuration processes
review. Regarding these mechanisms, the preferred method using configuration files, where a series of parameters are set
for customizing dashboards is the use of configuration wiz- to render a concrete and functional dashboard.
ards that supports the users’ decisions when building her or Some works also take advantage of the Software Product
his customized dashboards without requiring programming Line (SPL) paradigm [23], [25] or Model-Driven Devel-
skills. For example, [18], [19], [26], [27] use graphical user opment (MDD) [28]–[30]. In the case of SPL approaches
interfaces that ease the selection of widgets and the data to be applied to dashboards, they are based on the conception that
displayed, following the workflow shown in Figure 3. dashboards are sets of components with optional, alternative,
109680 VOLUME 7, 2019

FIGURE 7. The MDD workflow applied to dashboard development.
FIGURE 5. Dashboard configuration process involving files. propose a framework for creating different templates with
different KPIs and goals for small and medium enterprises.
A similar MDD approach is followed in [33], although
authors don’t explicitly indicate that they followed this
paradigm. In this case, to generate the dashboard, a context-
aware generator with user, data and visualization models as
inputs is in charge of generating the dashboard instances,
but the internal features of the dashboard generator are not
detailed.
Regarding adaptive solutions, agents are a common
method for managing changing requirements [32], [34].
In [34], device cloud platform dashboards are adjusted
through cloud agents that adapt themselves to the devices’
data models, thus generating remote user interfaces based
on the characteristics of the monitored devices. In [32], an
analytical system is guided by agents that are present in five
FIGURE 6. The software product line paradigm applied to dashboards.
communities (gatherers, conciliators, providers, visualizers,
and restructurers) to log user interactions with the system and
reconfigure the dashboard accordingly.
or mandatory features. These paradigms are used to finally Other methods found in the selected papers enclose inclu-
generate a dashboard that fits the previously defined feature sive user modeling for adapting the dashboard interface to
model, as shown in Figure 6. Regarding the solution pre- the user abilities [31], semantic reasoners for selecting appro-
sented in [23], it is worth to mention that an extended version priate data sources and compositions [35] and knowledge
of this work can be found at [42]. This last work did not graphs and ontologies to adapt the dashboards to the target
appear within the selected papers because it was published data domain [36].
after the execution of the present SLR, but in subsequent
updates, it would replace the previous paper, keeping the most C. HOW THE PROPOSED SOLUTIONS MANAGE THE
complete and recent version of the study. DASHBOARD’S REQUIREMENTS?
In the case of MDD approaches, the logic is similar; code As introduced before, the necessity of tailored dashboards
generators are fed with a series of models that describe the lies in a large number of existing user profiles that could
dashboard at high-level, for example, as described in [28]. potentially use these tools. Generic or ‘‘one size fits all’’
With a set of transformations and mappings, high-level dashboards are relatively easier to implement than a specific
models are transformed into concrete dashboards, through dashboard for each end-user, because the latter approach is
specific description files [28] or by using pre-defined not scalable at all, as the number of users could increment
or custom-made templates [30]. Figure 7 illustrates this and their requirements evolve. However, ‘‘one size fits all’’
approach. dashboards lack of flexibility, and would only be effective
In [30], the authors point out the necessity of having pre- and efficient for specific user profiles [43], because data that
defined templates in conjunction with the models, to mate- is relevant for one user could be irrelevant for another user,
rialize and generate the dashboards. The approach of using and vice versa, and could play different roles in their decision-
pre-defined templates is also present in [20], where authors making processes.
VOLUME 7, 2019 109681

These are the reasons why tailored dashboards should be through the solution’s web application. As the paper exposes,
considered; to fulfill the requirements and necessities of each the devices’ data models are held in XML format, so the
user profile simultaneously. But managing this high volume requirements management is made through these device mod-
of requirements at once (that can even evolve) is not a trivial els, and therefore falling in the same category as these previ-
task. ous works based in models and structured files.
That is why this question is to be answered; to learn how The other agent-based solution presented in [32] stores
these solutions manage the requirements associated with each and modifies settings according to the users’ behavior and
user and how they provide a tailored dashboard accordingly. their events, thus needing also authentication and account
The second research question shown that configuration management services to work correctly, as discussed at the
wizards are popular methods to manage these require- beginning of this research question. In this case system’s
ments by giving the user the responsibility of building agents are the drivers of the dashboard modifications.
their own dashboard based on their necessities. These The remaining solutions proposed, on the one hand,
solutions allow users to customize their displays while a semantic reasoner to infer potentially interesting composi-
using their dashboards freely, thus performing the tailor- tions of data streams in the context of the Internet of Things
ing process at user-configuration time (i.e., at run-time, (IoT) [35]. This solution personalizes the presented informa-
but with the intervention of the user through explicit tion by composing semantically annotated data and visualiza-
actions). All solutions found that use a configuration wiz- tion services. However, as specified in the research question
ard approach [18], [19], [26], [27], [37]–[40] manage indi- RQ1, the solution is classified as hybrid, as the paper states
vidual user requirements by implementing authentication that ‘‘sensor and data compositions need to be dynamically
and account management services, associating each user visualized, thereby limiting the user input to selecting the
to his/her dashboard configuration persistently. Even some preferred visualization method from a system-generated list
solutions let users build visualizations without logging in to of meaningful options, taking into account the preferences
the system [40], in case the users do not need or do not and characteristics of the current user profile’’ [35].
want to save their own configuration for the future. This So the dashboard is personalized given the available sensor
user management approach is also applied to other solutions data (the dashboards’ requirements are managed through rea-
found, like in [31], where a user creates an account and fills soning processes and knowledge bases), but in the end, users
a questionnaire about her or his abilities to finally access her need to select the widgets that will compose their dashboard,
or his personalized view based on the previous information. although this aspect of the system is no further discussed in
However, these works do no further discuss the storage the paper.
method nor the possibility of storing different versions of On the other hand, Santos et al. presented in [36] a knowl-
a user dashboard over time, which could be very useful to edge graph and indicator ontology approach to automatically
collect the evolution of the preferences or user behavior. generate dashboards in the context of smart cities. The pro-
On the other hand, 10 of the selected works take advantage posed dashboard generator takes as input serialized knowl-
of structured files or models to hold individual dashboard edge graphs and offers different dashboard configurations
requirements that finally serve as inputs of generators that accordingly. The application allows the customization of the
provide the configured dashboard instance meeting the orig- automatically proposed dashboard, given the user freedom to
inal specifications. In this category fall those solutions based change the configuration before generating it. The dashboard
on configuration files [21], [22], [24], context models [33], information requirements are managed through knowledge
software product lines [23], [25] or model-driven develop- graphs, but the user requirements management when man-
ment [28]–[30]. In this case, user requirements are managed ually customizing the dashboard is not further discussed.
‘‘outside’’ the dashboard systems, before their exploitation,
and stored within individual files or models.
In the case of [20], no requirement management is explic- D. CAN THE PROPOSED SOLUTIONS BE TRANSFERRED
itly performed, as the pre-defined templates enclose general TO DIFFERENT DOMAINS?
requirements collected from the gathering and analysis phase, Dashboards are used to exploit datasets that are usually large,
and subsequently, users select the template that fits better but also these datasets come from different domains. This
their needs. This management method allows better require- research question tries to answer the flexibility of the solution
ment traceability, as requirements are parsed and mapped found regarding their transfer capabilities to another domain.
from the specification to the concrete system features. Also, In other words, can the solutions fully support the visualiza-
it allows an easier version control of each file or model, tion of data from other domains without significant changes
keeping the evolution of individual dashboard requirements. in the original code?
Also, in the agent-based solution found [34], the system’s When a solution is focused on a particular data domain,
cloud agent adapts to each device data-model and adds addi- it could be challenging to reuse that same solution for other
tional cloud agent information markers which act as a user data domains if the source code is coupled to the original
interface description language. These markers are initially goals, allowing tailoring capabilities, but only within the
provided by the devices’ data model but can be modified domain’s frontiers.
109682 VOLUME 7, 2019

Some solutions based on configuration wizards, like The remaining works, on the one hand, use agents to
[38], [39], support the exploitation of datasets from different restructure dynamically a dashboard based on user behavior.
domains by allowing the users to upload or specify their This adaptive process is not coupled to the data domain,
concrete data sources. These solutions are robust as they as there are a specific community of agents (called gatherers)
can be reutilized for different goals depending on the data that are responsible for collecting the data from different
domain. The rest of solutions using configuration wizards sources [32], suggesting that their task is to gather data no
allow freedom when configuring the dashboards, but only matter its domain. Finally, in [33], a dashboard generator fed
within the original domain (environmental performance [18], with the user, visualization, and analysis scenario models is
micro-services monitoring [19], emergency situations [37], presented in the context of learning analytics. The user and
learning analytics [26], physics [27] or economics [40]). visualization models are more generic and focused on the
On the other hand, although the solution presented in [22] users’ preferences, experience, goals, visualization purposes,
can be adapted to different monitoring scenarios, it cannot be etc., allowing their reuse on other data domains. However,
transferred to other data domains as it relies on API endpoints the analysis scenario model is more coupled to the learning
to monitor resources. In the case of [21], the configuration analytics domain, mentioning learning objectives, pedagogi-
files allow the specification of the data sources, which can be cal context, fields of education, etc., so it could not be reused
local as well as remote, and their associated elements, mean- for domains outside the learning analytics context.
ing that the solution can be applied to other data domains As a clarification, it is worth to state that every method-
besides web analytics (which is the domain of the presented ology employed in the selected papers could be applied
prototype). The dashboard generator detailed in [24] also to develop dashboards in different data domains. However,
allows the specification of the dataset to be represented, but the purpose of this research question is to identify the most
this approach is mainly employed to develop studies regard- flexible and powerful solutions regarding their abstraction
ing usability guidelines, so the data domain’s transferability and, therefore, their potential reuse to other domains in an
is not significantly relevant in this case. automatized manner (i.e., avoiding to develop the same solu-
Other solutions that take part in the configuration files tion for new domains manually).
approach, as discussed in the second research question, are
the ones using MDD or SPL approaches. These solutions, E. HAS ANY ARTIFICIAL INTELLIGENCE APPROACH BEEN
which are based on meta-modeling [28]–[30], take advan- APPLIED TO THE DASHBOARDS’ TAILORING PROCESSES
tage of high-abstraction levels and commonalities among the AND, IF APPLICABLE, HOW THESE APPROACHES HAVE
potential products [23], [25] to address the generation of dif- BEEN INVOLVED IN THE DASHBOARDS’
ferent dashboards. Meta-modelling and domain engineering TAILORING PROCESSES?
allow the abstraction of the dashboards’ features, improving Again, dashboards deal with lots of data and requirements,
reusability of core assets and thus, making it possible to trans- and even generative approaches based on configuration files
fer the solutions to other data domains without significant or generators still need from manual configuration through
efforts. high-level languages or domain-specific languages. A similar
The dashboard presented in [31] is also tightly coupled issue arises from configuration wizards; in the end, users
to its original domain, as the adaptation of the dashboard need, through actions, to specify requirements that are not
is focused on the user physical abilities. The same is true always clear for themselves.
for [20]; although alternative templates can be chosen to Using methods that involve artificial intelligence (AI)
visualize different data aspects, they are always related to algorithms to manage the dashboards requirements could
business intelligence (sales, human resources, overall equip- lead to more accurate dashboard configurations and decrease
ment effectiveness, etc.). The template approach can be taken, the consumed resources during the requirement elicitation
of course, to address other data domains, but, new templates phases, as requirements could be automatically inferred by
should be developed for each target domain to accomplish the AI algorithm. With AI, systems can use algorithms to
this ‘‘domain transfer.’’ In the case of [36], the solution learn patterns from data and apply inference to predict future
employs a knowledge graph and indicator ontologies to gen- values. This approach would be potentially beneficial in
erate personalized dashboards; the ontologies used are related the domain of tailored dashboards because user preferences
to the Smart City context, so, to transfer this methodol- could be inferred from behavioral data, context, or any other
ogy to other domains, ontologies related to them should be factor.
employed. Only a few works have applied or mentioned AI when
Some of the works are focused on sensor monitoring [35] presenting their dashboard solutions. In [32], the Apriori
and device clouds [34]. The methodologies employed in the algorithm [44] is used to compute association rules, which
papers mentioned above (semantic reasoners and multi-agent is a technique from the data mining field. This solution
systems, respectively) could be reused for other domains, but takes advantage of ‘‘pairs of events that have happened
in the end, the dashboard solutions would need to be built in sequence’’ that fed the Apriori algorithm to obtain a
from zero to adapt them to new domains. set of if-then rules that will be used to restructure the
VOLUME 7, 2019 109683

dashboard in terms of the presented data and visualization some types of graph or chart and changing some layouts and
types employed. In a study referencing those mentioned colors’’, possibly needing more customizable elements.
above [45], the same authors specify that their solution also The two model-driven solutions regarding semantic
supports the restructuration of the dashboards through other approaches [28], [29] also mentioned user testing and claim-
methods, like Markov chains or top-k queries, but they don’t ing that the feedback shown the relevance of the semantic
detail these processes. description language to adapt their dashboards easily. On the
Also, in [35], a semantic EYE reasoner is employed to other hand, in [29], two users were considered for the assess-
discover potentially interesting data compositions through ment to test the role-based dashboard generation and provided
a knowledge base and semantically annotated visualization their KPIs requirements; however, no further details about
and data services. The use of semantic reasoners allows the this evaluations were given.
inference of consequences from facts, enabling in this case, In [34], the unique evaluation mentioned that addresses
‘‘the detection of complex events that previously would have user experience shows that it takes less time for a user to
remained undetected’’ [35]. However, no details about the find device commands through the custom user interfaces
implementation of the reasoner are addressed in this work. described in that work. However, no details about the eval-
Other papers mention the possibility of introducing AI uation sample nor methodologies are provided.
techniques, like [24], to rate the generated dashboards The dashboard solution focused on novices presented
through classification algorithms, but authors state that is out in [38] is complemented with a detailed usability study to
of the scope of the paper and refer to [46] as an inspiration. validate their approach and to examine how novice users
There is also a work that mentions inference [33] to provide create dashboards. Fifteen users (7 novice users and 8 BI
a suitable dashboard given the context, user description, and dashboards experts) participated in the study at a usability lab,
analysis scenario, although no further details are given nor where they were interviewed, recorded and asked to complete
the inference method named a series of tasks with the dashboard solution following the
think-aloud protocol. The findings shown that novices ranked
F. HOW MATURE ARE TAILORED DASHBOARDS better the dashboard regarding utility, functionality, ease of
REGARDING THEIR EVALUATION? use, and overall satisfaction, and expressed the intuitiveness
The proposed dashboard solutions are functional regarding of the dashboard. Experts, on the other hand, requested extra
their tailoring capabilities, but the maturity of these dash- functionalities. The authors then provide a series of additional
boards regarding their usability is essential to demonstrate if guidelines when designing visualization systems for novice
tailoring the dashboards is beneficial for the final users. users based on the results, which can be found in [38].
In [18], focus groups, pre- and post-study interviews were In [39], an insight-based evaluation is employed to test the
employed to test the perceived usability and impact of the validity of the presented model. Six participants not skilled
customizable dashboards. Five experts were in charge of in visual data exploration were asked to do an unguided
testing the prototype, and subsequently, 13 participants tested exploration of a dataset. Participants were asked to complete a
the final prototype. Issues regarding the solution involved survey focusing on insight-based metrics [48] complemented
the configuration process, the interface, and the diversity with a demographic survey. The study shown that ‘‘non-
of available widgets to include in each dashboard, meaning default, less familiar settings for expressive richness are more
that users needed more components to satisfy their concrete likely to lead to incorrect statements.’’
requirements. The solution described in [26] performed a usability test
In the case of [19], a combined survey and interview was through the SUS questionnaire [49] and other open questions.
performed to obtain requirements regarding micro-service Twelve participants were asked to complete a series of tasks
monitoring. A total of 15 participants were involved, and the using the DDART system (7 users performed the experiment
gathered information was used to define the main require- remotely and 5 with the presence of a researcher). The general
ments of the customizable dashboard, but no further usabil- usability using the SUS scale was 53.93 for the remote group
ity testing was performed regarding the finally implemented and 54.50 for the assisted group. They also tested the ease of
dashboard solution. use of the dashboard by analyzing the success ratio, average
In [37], it is mentioned that ‘‘first evaluations with users time, efficiency and average invalid operations ratio when
from the domain have already shown that this solution could crafting indicators to gain insights about the difficulty of this
successfully address the problem of information overload,’’ feature; results shown that the assisted group performed better
but no details about the evaluation methods nor detailed than the remote group.
results are given. Finally, the dashboard presented in [40] was also tested
The pre-defined templates-based solution presented in [20] to obtain information about the usefulness of the solution
was evaluated during 6-9 months in 40 different small and through users’ feedback. In this case, more than 60 users
medium enterprises (SMEs) and evaluated the implemented tested the solution through half-open scenarios, where users
dashboards’ capabilities through [47], obtaining good results can ask questions and directions are given only under their
regarding dashboard layout, design, presentation, alerting, demand. Satisfaction scores were collected regarding usabil-
analysis, KPIs, etc. Also, ‘‘25% of SMEs suggested to change ity, information output, and functionalities like search, detail
109684 VOLUME 7, 2019

view of the results, visualizations, and maps; obtaining high adapted the dashboards in real-time based on user behav-
percentages of satisfied users. ior [32], [33].
The solutions presented in [21]–[25], [27], Customizable solutions address individual requirements by
[30]–[33], [35], [36], did not mention any formal testing directly asking the user to design their own dashboard with-
regarding end-users’ perceptions about the dashboard solu- out requiring programming skills; by either using graphical
tions, mentioning these evaluations as future work. Some of user interfaces [18], [19], [26], [27], [37]–[40] or high-level
these proposed tools were tested in real-world scenarios to configuration files [21]–[25], [28]–[30] that can abstract the
prove their applicability or functionalities, but this research technical and complex details of the dashboard implementa-
question is focused on user perceptions and experiences on tion. This approach partially delegates the dashboard design
the solutions and composition responsibility to the users, which lead to a
decrease in development time. But these solutions come with
a significant disadvantage: users do not always know what
VI. DISCUSSION is good for them [7], so they can build ineffective dashboard
A variety of works have been retrieved regarding tailoring solutions unwillingly.
capabilities within the dashboards’ context. Through this lit- This issue can be addressed through assisted customiza-
erature review, it has been possible to answer questions about tion processes, as found in [37]–[40]. These approaches give
the capacities and approaches taken to build tailored dash- freedom to the users regarding their dashboards composition,
boards through the existing solutions found in the literature. but also help them with design decisions and charts selection
It is clear that dashboards can be extremely powerful tools through methods like visual mapping. Another approach is to
if leveraged; they can support decision-making processes, personalize dashboards by extracting dashboard requirements
motivate, persuade, and even make data memorable if prop- implicitly from the users [28]–[31].
erly designed [50]. However, as introduced, a large number of Hybrid solutions also seem to address these caveats by tak-
potential users and their large number of individual require- ing into account that users do not always know what the best
ments makes ‘‘one size fits all’’ approaches only effective for for them or their goals is, thus requiring to add a degree of per-
some profiles, primarily because one size does not fit all for sonalization that can materialize implicit requirements. But
tasks involving cognitive processes; users are influenced by forcing users to stick to a ‘‘personalized’’ solution that makes
their experiences, their biases, their individual preferences, them uncomfortable is counterproductive, so customizability
etc., [43]. Different people could see the same dashboard with options should be available if a user feels that she needs to
the same data and reach different insights as they could be change something. Also, it is interesting to take into account
driven by different goals. This cannot be overlooked because an adaptive dimension to consider the users’ behavior evolu-
people could be missing relevant data for their decision- tion because what could seem the best configuration at some
making processes if the dashboard is not correctly configured point of time could become ineffective over time, as users are
for them. What also comes into play are users’ abilities: a involved in new experiences that could change their goals and
colorful dashboard could be a pleasant, aesthetic and effective even improve their visualization literacy and knowledge.
dashboard for one person, but could be a nightmare for a However, relevant questions regarding self-adaptive solu-
person with eye diseases [51], [52]. tions involve how often should be the dashboard updated to
For all these reasons, dashboards should include mech- display a new configuration or when it is considered that
anisms to allow tailor-made solutions for individual users a requirement has evolved enough and, therefore, the dash-
without requiring large amounts of resources (making a dashboard needs a new configuration. Works addressing adaptive
board from scratch for every potential user is not an efficient solutions have almost no allusions regarding this con-
option). The existing literature has been analyzed to find how cern [32], [33]. In [33], authors state that the adaptation
these tailoring processes have been addressed before and to is made on-demand on their proof of concept dashboard,
understand the current research context in this area. but they don’t mention this issue. In [32], [45], the authors
Accurately, 23 solutions that addressed the necessity of suggest that the restructuring period is previously set in each
customizing, personalizing, adapting, etc. dashboards were restructurer agents’ agendas, but they don’t identify or test the
retrieved. The retrieved solutions address this challenge implications of these restructuration periods. Adaptation time
through very different approaches. The first three research or adaptation triggers are relevant factors to address when
questions had the goal of identifying the technical features using this approach, because continually changing the user
of the retrieved solutions. These research questions allowed interface could annoy users [53] and be counterproductive,
to distinguish between tailored dashboards by classifying despite the potential benefits of the adaptation.
them through technical dimensions. While some solutions One of the limitations of the retrieved solutions is that they
let users customize their displays manually [18]–[27] or are very specific to the domain to which they were designed
with assistance [37]–[40], other personalized dashboards for, as exposed in RQ4. An ideal solution would be valid
through implicit requirements like goals, roles, target data, for every data domain and context, but the vast number of
etc., [28]–[31], in some cases letting users customize the varying features these tools can have increased the complex-
personalized display on demand [34]–[36], the remaining ity of creating a suitable dashboard for every domain and
VOLUME 7, 2019 109685

context. There are interesting approaches that tackle tailored or even a personalized display already designed to match their
dashboards from an abstract point of view, proposing generic requirements. Some users could request contextual informa-
solutions that are instantiated to fit into concrete require- tion about the presented data if their knowledge about the data
ments. For instance, model-driven development [28]–[30] domain is limited. And so on. What is clear after conducting
and domain engineering paradigms [23], [25] aim at extract- the present SLR is that one size does not fit all when talking
ing commonalities and shared properties from instances of about dashboards, but pursuing generic solutions that can be
dashboard systems to foster reusability and decrease the derived to match different contexts might be a proper path to
development time. Dealing with domain transferability seems follow.
to be easier when a generic point of view of the problem has
been established, as it can be seen in [23], [25], [28]–[30], VII. THREATS TO VALIDITY
where instantiated dashboards can be adapted to fit into other This kind of reviews can be influenced by a series of limita-
requirements, domains and contexts without modifying high- tions. One of these limitations is the authors’ bias regarding
level, abstract models. the whole data extraction. As exposed in previous sections,
The answer to RQ5 revealed that artificial intelligence (AI) quality criteria were employed to reduce the effects of bias
mechanisms are not leveraged, which could be useful to infer in the inclusion phase of the SLR. The three authors were
user requirements. Only one work exposed the application of involved in the review planning to identify and avoid any
a data mining technique to their dashboard proposal. How- early issues regarding the study design. Moreover, the first
ever, although potentially beneficial, AI approaches present author was the primary reviewer, while the last two authors
challenges, like the gathering process of significant data to reproduced each SLR phase to ensure the validity of the
build models. It is crucial to implement collecting mecha- results taking into account different perspectives.
nisms to store user behavior and their implicit/explicit prefer- Also, different resources with the outcomes of each step
ences. But before implementing these mechanisms, relevant are provided to make the whole process reproducible.
users’ aspects and factors that can influence their user expe- Although following a systematic, well-defined protocol,
rience must be identified and backed up by previous studies it is not guaranteed that all the relevant works about this field
about perception, to avoid an arbitrary selection of factors that are retrieved. Regarding the search medium, the most relevant
could skew the AI models’ outcomes. electronic databases in the field of computer science were
Regarding this application, solutions based in struc- included. The exclusion of Google Scholar from this review
tured files to describe a dashboard’s features, like is justified by the necessity of considering only databases
in [21]–[25], [28]–[30], [33], provide a good basis for imple- that index quality contrasted contents. Also, to include the
menting AI solutions because AI algorithms are easier to maximum quantity of representative terms about the tailored
handle if their inputs are already structured. AI models could dashboards, synonyms, and related terms were identified, and
be useful to detect patterns or clusters of users regarding the results of preliminary search strings were evaluated to
dashboard configurations [54]. Artificial intelligence is cur- analyze if the retrieved data were relevant for the scope of
rently identified as a potentially beneficial method to recom- this literature review. Through this iterative process, the query
mend suitable settings of single information visualizations string was refined to ensure useful and precise data extraction.
given different factors, like the target data, user behavior, One of the main limitations regarding the field in which this
etc., [55], [56]; but dashboards, where several information SLR is framed is that several dashboard solutions are com-
visualizations can be displayed and can even hold linked mercial solutions. Thus no public research works regarding
views, are not mentioned in these works. their technical features or the methodologies used are avail-
Finally, there is a lack of user testing regarding the devel- able. Despite this issue, enough relevant works for answering
oped solutions. As exposed in RQ6, the solutions were tested the considered research questions were retrieved.
regarding their functionality, but few works also included Lastly, to ensure that the whole process is traceable and
testing regarding user experience and insight delivery. For- reproducible, all the materials, partial results, checklists, etc.,
mal testing is essential in this kind of solutions because it have been made available through public repositories.
can expose the actual usefulness of tailoring capabilities in
terms of usability and knowledge generation, thus helping VIII. CONCLUSION
to improve the tailoring mechanisms and enhance decision- A systematic literature review (SLR) has been conducted to
making processes. analyze previous works that address tailoring capabilities and
This systematic literature review exposed a wide range of mechanisms regarding information dashboards. This SLR
available approaches to tackle tailored dashboards. Choosing addresses relevant aspects regarding these solutions, such as
a proper approach depends, of course, on the application con- the applied methodologies, dashboard requirements manage-
text, the audience, the data, and the available resources for the ment, domain transferability, artificial intelligence applica-
development of tailored dashboards. Experts could demand tions, and user experience testing, to identify current issues
fine-grained features, but they can build their own dashboards and challenges, as well as new research paths to enhance tai-
without assistance given their visualization literacy. Novices loring capabilities and consequently, knowledge generation
could require system support for composing a dashboard, through individualized dashboards.
109686 VOLUME 7, 2019

By performing an SLR, research questions about these [9] B. Kitchenham and S. Charters, ‘‘Guidelines for performing systematic
dashboard solutions have been answered, providing a com- literature reviews in software engineering,’’ School Comput. Sci. Math.,
Keele Univ., Keele, U.K., Tech. Rep. EBSE-2007-01, 2007. [Online].
prehensive view of this research field’s current state. During Available: https://goo.gl/L1VHcw.
the review process, 1034 papers were retrieved from 4 differ- [10] B. A. Kitchenham, D. Budgen, and O. P. Brereton, ‘‘Using mapping studies
ent electronic databases. The number of papers was reduced as the basis for further research—A participant-observer case study,’’ Inf.
Softw. Technol., vol. 53, no. 6, pp. 638–651, 2011.
to 23 after applying an inclusion and exclusion criteria, and [11] A. Vázquez-Ingelmo, F. J. García-Peñalvo, and R. Therón, ‘‘Tailored infor-
a quality assessment to keep only relevant works for the mation dashboards: A systematic mapping of the literature,’’ presented at
scope of the research. The analysis of the selected papers the Interacción, Donostia, Spain, 2019.
exposed that tailored dashboards have been tackled through [12] B. Napoleão, K. R. Felizardo, É. F. de Souza, and N. L. Vijaykumar, ‘‘Prac-
tical similarities and differences between systematic literature reviews and
diverse methodologies and mechanisms that enable support systematic mappings: A tertiary study,’’ in Proc. SEKE, 2017, pp. 85–90.
for different dashboard configurations without consuming [13] M. Petticrew and H. Roberts, Systematic Reviews in the Social Sciences:
loads of resources and without requiring long development A Practical Guide. Hoboken, NJ, USA: Wiley, 2008.
[14] M. Cutts, Oxford Guide to Plain English. Oxford, U.K.: OUP Oxford,
processes as compared with the design and implementation
2013.
of individual dashboards from scratch. [15] R. R. Prasath, ‘‘Learning age and gender using co-occurrence of non-
This SLR provides a foundation in terms of existing dictionary words from stylistic variations,’’ in Proc. Int. Conf. Rough Sets
approaches for tailoring information dashboards. Information Current Trends Comput. Berlin, Germany: Springer, 2010, pp. 544–550.
[16] D. Moher, A. Liberati, J. Tetzlaff, D. G. Altman, and T. P. Group,
dashboards have become key tools when dealing with data, ‘‘Preferred reporting items for systematic reviews and meta-
and there are a lot of challenges regarding their development analyses: The PRISMA statement,’’ PLoS Med., vol. 6, no. 7, 2009,
and design; one of these challenges is their adaptation to Art. no. e1000097.
[17] A. V. Ingelmo. (2019). Resources for the Paper, Information Dashboards
different contexts, domains and users [1]. This review can and Tailoring Capabilities—A Systematic Literature Review. [Online].
support researchers and developers in choosing a proper Available: https://zenodo.org/record/3246043-.XTl15ZMzZ24
mechanism to develop tailored dashboards. Also, the iden- [18] D. Filonik, R. Medland, M. Foth, and M. Rittenbruch, ‘‘A customisable
tified challenges can open new research paths. Moreover, dashboard display for environmental performance visualisations,’’ in Per-
suasive Technology. Berlin, Germany: Springer, 2013, pp. 51–62.
the obtained results can be applied to improve dashboard [19] B. Mayer and R. Weinreich, ‘‘A dashboard for microservice monitoring and
solutions lacking flexibility. management,’’ in Proc. IEEE Int. Conf. Softw. Archit. Workshops (ICSAW),
Future work will involve the application of the gained Gothenburg, Sweden, Apr. 2017, pp. 66–69.
[20] W. Noonpakdee, T. Khunkornsiri, A. Phothichai, and K. Danaisawat,
knowledge to propose new tailored dashboard solutions that ‘‘A framework for analyzing and developing dashboard templates for small
address the challenges and issues found through this review. and medium enterprises,’’ in Proc. IEEE 5th Int. Conf. Ind. Eng. Appl.
Based on the presented analysis, one of the most promising (ICIEA), Apr. 2018, pp. 479–483.
research avenues is the application of AI paradigms to autom- [21] K. Kumar, J. Bose, and S. K. Soni, ‘‘A generic visualization framework
based on a data driven approach for the analytics data,’’ in Proc. 14th IEEE
atize the design and development of dashboards. This appli- India Council Int. Conf. (INDICON), Dec. 2017, pp. 1–6.
cation will not only involve the selection and development of [22] A. Cardoso, C. J. V. Teixeira, and J. S. Pinto, ‘‘Architecture for highly
tailoring mechanisms but also the study and classification of configurable dashboards for operations monitoring and support,’’ Stud.
Inform. Control, vol. 27, no. 3, pp. 319–330, 2018.
users to make these automatic tailoring processes useful and [23] A. Vázquez-Ingelmo, F. J. García-Peñalvo, and R. Therón, ‘‘Domain engi-
effective. neering for generating dashboards to analyze employment and employa-
bility in the academic context,’’ presented at the 6th Int. Conf. Technol.
Ecosyst. Enhancing Multiculturality, Salamanca, Spain, 2018.
REFERENCES
[24] O. Pastushenko, J. Hynek, and T. Hruška, ‘‘Generation of test samples for
[1] A. Sarikaya, M. Correll, L. Bartram, M. Tory, and D. Fisher, ‘‘What do construction of dashboard design guidelines: Impact of color on layout
we talk about when we talk about dashboards?’’ IEEE Trans. Vis. Comput. balance,’’ in Proc. World Conf. Inf. Syst. Technol. Cham, Switzerland:
Graph., vol. 25, no. 1, pp. 682–692, Jan. 2019. Springer, 2018, pp. 980–990.
[2] S. Few, Information Dashboard Design. Sebastopol, CA, USA: O’Reilly
[25] I. Logre, S. Mosser, P. Collet, and M. Riveill, ‘‘Sensor data visualisa-
Media, 2006.
tion: A composition-based approach to support domain variability,’’ in
[3] S. Wexler, J. Shaffer, and A. Cotgreave, The Big Book of Dashboards:
Proc. Eur. Conf. Model. Found. Appl. Cham, Switzerland: Springer, 2014,
Visualizing Your Data Using Real-World Business Scenarios. Hoboken,
pp. 101–116.
NJ, USA: Wiley, 2017.
[26] C. Michel, E. Lavoué, S. George, and M. Ji, ‘‘Supporting awareness
[4] J. Hullman, E. Adar, and P. Shah, ‘‘The impact of social information on
and self-regulation in project-based learning through personalized dash-
visual judgments,’’ in Proc. ACM SIGCHI Conf. Hum. Factors Comput.
boards,’’ Int. J. Technol. Enhanced Learn., vol. 9, nos. 2–3, pp. 204–226,
Syst., 2011, pp. 1461–1470.
2017.
[5] Y.-S. Kim, K. Reinecke, and J. Hullman, ‘‘Data through others ‘eyes:
The impact of visualizing others’ expectations on visualization interpre- [27] G. L. Miotto, L. Magnoni, and J. E. Sloper, ‘‘The TDAQ Analytics
tation,’’ IEEE Trans. Vis. Comput. Graphics, vol. 24, no. 1, pp. 760–769, Dashboard: A real-time Web application for the ATLAS TDAQ control
Jan. 2018. infrastructure,’’ J. Phys., Conf. Ser., vol. 331, no. 2, 2011, Art. no. 022019.
[6] D. A. Norman and S. W. Draper, User Centered System Design: New [28] M. Kintz, ‘‘A semantic dashboard description language for a process-
Perspectives on Human-computer Interaction. Boca Raton, FL, USA: CRC oriented dashboard design methodology,’’ in Proc. 2nd Int. Workshop
Press, 1986. Model-Based Interact. Ubiquitous Syst. (MODIQUITOUS), Copenhagen,
[7] L. Padilla. (2018). How do we Know When a Visualization is Denmark, vol. 947, 2012, pp. 31–36.
Good? Perspectives From a Cognitive Scientist. [Online]. Available: [29] M. Kintz, M. Kochanowski, and F. Koetter, ‘‘Creating user-specific busi-
https://medium.com/multiple-views-visualization-research-explained/ ness process monitoring dashboards with a model-driven approach,’’ in
how-do-we-know-when-a-visualization-is-good-c894b5194b62 Proc. MODELSWARD, 2017, pp. 353–361.
[8] B. Kitchenham, ‘‘Procedures for performing systematic reviews,’’ Dept. [30] T. Palpanas, P. Chowdhary, G. Mihaila, and F. Pinel, ‘‘Integrated model-
Comput. Sci., Keele Univ., Keele, U.K., Tech. Rep. 0400011T.1, 2004, driven dashboard development,’’ Inf. Syst. Frontiers, vol. 9, nos. 2–3,
pp. 1–26, vol. 33. pp. 195–208, 2007.
VOLUME 7, 2019 109687

[31] S. Arjun, ‘‘Personalizing data visualization and interaction,’’ in Proc. ACM [53] M. Zavyiboroda. (2018). Why User Interface Changes Annoy People &
Adjunct Publication 26th Conf. User Modeling, Adaptation Personaliza- How to Handle it. [Online]. Available: https://speckyboy.com/annoying-
tion, 2018, pp. 199–202. ui-changes/
[32] O. Belo, P. Rodrigues, R. Barros, and H. Correia, ‘‘Restructuring dynam- [54] J. Cruz-Benito, A. Vázquez-Ingelmo, J. C. Sánchez-Prieto, R. Therón,
ically analytical dashboards based on usage profiles,’’ in Proc. Int. Symp. F. J. García-Peñalvo, and M. Martín-González, ‘‘Enabling adaptability in
Methodol. Intell. Syst. Cham, Switzerland: Springer, 2014, pp. 445–455. Web forms based on user characteristics detection through A/B testing and
[33] I. Dabbebi, S. Iksal, J.-M. Gilliot, M. May, and S. Garlatti, ‘‘Towards machine learning,’’ IEEE Access, vol. 6, pp. 2251–2265, 2018.
adaptive dashboards for learning analytic: An approach for conceptual [55] M. Vartak, S. Huang, T. Siddiqui, S. Madden, and A. Parameswaran,
design and implementation,’’ in Proc. 9th Int. Conf. Comput. Supported ‘‘Towards visualization recommendation systems,’’ ACM SIGMOD Rec.,
Edu. (CSEDU), Porto, Portugal, vol. 1, 2017, pp. 120–131. vol. 45, no. 4, pp. 34–39, 2017.
[34] S. Radovanović, B. Majstorović, S. Kukolj, and M. Z. Bjelica, ‘‘Device [56] P. Kaur and M. Owonibi, ‘‘A review on visualization recommendation
cloud platform with customizable remote user interfaces,’’ in Proc. strategies,’’ in Proc. 12th Int. Joint Conf. Comput. Vis., Imag. Comput.
IEEE 4th Int. Conf. Consum. Electron. Berlin (ICCE-Berlin), Sep. 2014, Graph. Theory Appl. (VISIGRAPP), vol. 3, 2017, pp. 266–273.
pp. 202–204.
[35] S. Van Hoecke, C. Huys, O. Janssens, R. Verborgh, and ANDREA VÁZQUEZ-INGELMO received the
R. Van de Walle, ‘‘Dynamic monitoring dashboards through composition bachelor’s and master’s degrees in computer engi-
of Web and visualization services,’’ in Internet of Things. IoT neering from the University of Salamanca, Sala-
Infrastructures. Cham, Switzerland: Springer, 2015, pp. 465–474. manca, in 2016 and 2018, respectively. She is
[36] H. Santos, V. Dantas, V. Furtado, P. Pinheiro, and D. L. McGuinness, currently pursuing the Ph.D. degree in com-
‘‘From data to city indicators: A knowledge graph for supporting auto- puter sciences with the Research Group of Inter-
matic generation of dashboards,’’ in Proc. Eur. Semantic Web Conf. Cham,
action and eLearning (GRIAL), where she is
Switzerland: Springer, 2017, pp. 94–108.
also a member. Her research interests include
[37] B. S. Nascimento, A. S. Vivacqua, and M. R. S. Borges, ‘‘A flexible
related to human–computer interaction, soft-
architecture for selection and visualization of information in emergency
situations,’’ in Proc. IEEE Int. Conf. Syst., Man, Cybern. (SMC), Oct. 2016, ware engineering, information visualization, and
pp. 3317–3322. machine learning applications.
[38] M. Elias and A. Bezerianos, ‘‘Exploration views: Understanding dashboard
FRANCISCO J. GARCÍA-PEÑALVO received the
creation and customization for visualization novices,’’ in Proc. IFIP Conf.
Hum.-Comput. Interact. Berlin, Germany: Springer, 2011, pp. 274–291. degrees in computing from the University of Sala-
[39] M. A. Yalçın, N. Elmqvist, and B. B. Bederson, ‘‘Keshif: Rapid and manca and the University of Valladolid, and the
expressive tabular data exploration for novices,’’ IEEE Trans. Vis. Comput. Ph.D. degree from the University of Salamanca.
Graphics, vol. 24, no. 8, pp. 2339–2352, Aug. 2018. He was the Vice Chancellor of innovation with
[40] G. Petasis, A. Triantafillou, and E. Karstens, ‘‘YourDataStories: Trans- the University of Salamanca, from 2007 to 2009.
parency and corruption fighting through data interlinking and visual explo- He is currently the Head of the Research Group
ration,’’ in Proc. Int. Conf. Internet Sci. Cham, Switzerland: Springer, 2017, Interaction and eLearning and a Full Professor
pp. 95–108. with the Computer Science Department, Univer-
[41] E. H.-H. Chi, ‘‘A taxonomy of visualization techniques using the data state sity of Salamanca. He has led and participated in
reference model,’’ in Proc. IEEE Symp. Inf. Vis. (INFOVIS), Oct. 2000, over 50 research and innovation projects. He has authored over 300 arti-
pp. 69–75. cles in international journals and conferences. His main research interests
[42] A. Vázquez-Ingelmo, F. J. García-Peñalvo, and R. Therón, ‘‘Taking advan- include eLearning, computers and education, adaptive systems, web engi-
tage of the software product line paradigm to generate customized user neering, semantic web, and software reuse. He is also a member of the
interfaces for decision-making processes: A case study on university Program Committee of several international conferences and a Reviewer of
employability,’’ PeerJ Comput. Sci., vol. 5, p. e203, Jul. 2019.
several international journals. He was a Guest Editor of several special issues
[43] S. D. Teasley, ‘‘Student facing dashboards: One size fits all?’’ Technol.,
of international journals, such as the Online Information Review, the Com-
Knowl. Learn., vol. 22, no. 3, pp. 377–384, 2017.
[44] R. Agrawal and R. Srikant, ‘‘Fast algorithms for mining association rules,’’ puters in Human Behavior, and the Interactive Learning Environments. He is
in Proc. 20th Int. Conf. Very Large Data Bases, VLDB, vol. 1215, 1994, currently the Editor-in-Chief of the International Journal of Information
pp. 487–499. Technology Research and the Education in the Knowledge Society Journal.
[45] O. Belo, H. Correia, P. Rodrigues, and R. Barros, ‘‘A personalization Besides, he is the Coordinator of the multidisciplinary Ph.D. Programme on
system for data visualization platforms,’’ in Proc. IEEE 6th Int. Conf. Education in the Knowledge Society.
Innov. Comput. Technol. (INTECH), Aug. 2016, pp. 162–167.
[46] K. Reinecke, T. Yeh, L. Miratrix, R. Mardiko, Y. Zhao, J. Liu, and ROBERTO THERÓN received the Diploma
K. Z. Gajos, ‘‘Predicting users’ first impressions of website aesthetics with degree in computer science from the University of
a quantification of perceived visual complexity and colorfulness,’’ in Proc. Salamanca, the B.A. degree from the Universidade
ACM SIGCHI Conf. Hum. Factors Comput. Syst., 2013, pp. 2049–2058. da Coruña, the bachelor’s degree in communica-
[47] M. Dyczkowski, J. Korczak, and H. Dudycz, ‘‘Multi-criteria evaluation of tion studies and humanities from the University
the intelligent dashboard for SME managers based on scorecard frame- of Salamanca, and the Ph.D. degree from the
work,’’ in Proc. IEEE Federated Conf. Comput. Sci. Inf. Syst., Sep. 2014, Research Group Robotics, University of Sala-
pp. 1147–1155. manca. His Ph.D. thesis was on parallel calculation
[48] P. Saraiya, C. North, and K. Duca, ‘‘An insight-based methodology configuration space for redundant robots. He is
for evaluating bioinformatics visualizations,’’ IEEE Trans. Vis. Comput. currently the Manager of the VisUSAL Group
Graphics, vol. 11, no. 4, pp. 443–456, Jul. 2005. (within the Recognized Research Group GRIAL), University of Salamanca,
[49] J. Sauro, A Practical Guide to the System Usability Scale: Background, which focuses on the combination of approaches from computer science,
Benchmarks & Best Practices. Scotts Valley, CA, USA: Createspace
statistics, graphic design, and information visualization to obtain an adequate
Independent, 2011.
understanding of complex data sets. He has authored over 100 articles in
[50] S. Berinato, Good Charts: The HBR Guide to Making Smarter, More
Persuasive Data Visualizations. Brighton, MA, USA: Harvard Business
international journals and conferences. In recent years, he has been involved
Review, 2016. in developing advanced visualization tools for multidimensional data, such
[51] S. Silva, B. S. Santos, and J. Madeira, ‘‘Using color in visualization: as genetics or paleo-climate data. In the field of visual analytics, he develops
A survey,’’ Comput. Graph., vol. 35, no. 2, pp. 320–333, 2011. productive collaborations with groups and institutions internationally recog-
[52] T. Zuk, L. Schlesier, P. Neumann, M. S. Hancock, and S. Carpendale, nized as the Laboratory of Climate Sciences and the Environment, France,
‘‘Heuristics for information visualization evaluation,’’ in Proc. ACM AVI or the Austrian Academy of Sciences, Austria. He received the Extraordinary
Workshop BEyond Time Errors, Novel Eval. Methods Inf. Vis., 2006, Doctoral Award for his Ph.D. thesis.
pp. 1–6.
109688 VOLUME 7, 2019

Information Dashboards and Tailoring Capabilities - A Systematic Literature Review

Uploaded by

Copyright:

Available Formats

Information Dashboards and Tailoring Capabilities - A Systematic Literature Review

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Information Dashboards and Tailoring Capabilities - A Systematic Literature Review

Uploaded by

Copyright:

Available Formats

Received July 28, 2019, accepted August 1, 2019, date of publication August 6, 2019, date of current version August

Information Dashboards and Tailoring

I. INTRODUCTION Although identifying what is and what is not an informa-

109674 VOLUME 7, 2019

VOLUME 7, 2019 109675

109676 VOLUME 7, 2019

VOLUME 7, 2019 109677

109678 VOLUME 7, 2019

3. The maintained papers were analyzed by reading their

VOLUME 7, 2019 109679

analytical sessions (i.e., through the analysis of the user

users with the selection of a suitable dashboard configuration.

109680 VOLUME 7, 2019

FIGURE 7. The MDD workflow applied to dashboard development.

VOLUME 7, 2019 109681

109682 VOLUME 7, 2019

VOLUME 7, 2019 109683

109684 VOLUME 7, 2019

VOLUME 7, 2019 109685

109686 VOLUME 7, 2019

VOLUME 7, 2019 109687

109688 VOLUME 7, 2019

You might also like