Received May 14, 2021, accepted May 26, 2021, date of publication June 3, 2021, date of current version

June 11, 2021.

Digital Object Identifier 10.1109/ACCESS.2021.3085496

Visual Analytics for Decision Support: A Supply

Chain Perspective
ALIREZA KHAKPOUR 1 , RICARDO COLOMO-PALACIOS 1, (Senior Member, IEEE),
AND ANTONIO MARTINI 2
1 Faculty of Computer Sciences, Østfold University College, 1757 Halden, Norway
2 Department of Informatics, University of Oslo, 0373 Oslo, Norway
Corresponding author: Alireza Khakpour ([email protected])
This work was supported by DigiMat Project granted by Research Council of Norway under Project 296686.

ABSTRACT Supply chain (SC) activities generate huge amount of data that can be used in decision
making processes. However, proper data analytics techniques are required to combine, organize, and analyze
data from different sources and produce required insights available for decision makers. These techniques
promote analytical reasoning of the events and patterns hidden in the data using visualizations, so-called
Visual Analytics (VA). Although there is a large number of VA systems to facilitate the process of analysis
and decision making, there is a lack of an adequate overview of what already exists in this area for SC
management. To address that need, we conducted a systematic literature review to analyze the state of the art
in SC VA systems. Particularly, we focus on use cases, the type of the decisions that a VA system intended
to support, the type of visualizations employed, the type of analytics used, and the data that has been used
for analysis. The goal of this study is to provide SC and VA researchers with an overview of the works
carried out in the field of SC VA, helping them to observe latest trends and to recognize existing gaps that
need further investigation. Consequently, a mapping between decisions of various SC business processes and
their reciprocal visualization techniques and tactics have been provided. Adding to that, VA applications and
use cases in SC are identified based on the SC Operation Reference (SCOR) model and underlying decision
areas are recognized.

INDEX TERMS Decision support, information visualization, value chain analysis, visual analytics, visual-
ization techniques and methodologies.

I. INTRODUCTION paper, authors emphasized the role of VA in making effective

The massive and heterogenous amount of data produced by decisions, however, the corresponding decisions regarding
Supply Chain (SC) actors raises the need for developing data different SC processes are not identified. In our study, we aim
analytics solutions to support decision making activities [1]. to identify each of the decisions related to a particular SC
Visual Analytics (VA) plays a key role in analyzing the vast business activity supported by VA. The authors in [3] con-
amount of data collected by different SC actors as the result ducted a systematic literature review (SLR) regarding SC
of their day-to-day supply network operations, taking from analytic systems. They have focused on big data capabilities
suppliers and manufacturers to warehouses, logistics and of such systems, and despite the recognition of data visual-
retailers. ization as one of the main capabilities, it is not investigated
Many analytical approaches have been proposed to sup- in-depth. In this study, we identify the state of the art in the
port decision making for SC processes and activities [2], [3]. visualization techniques and tactics that have been used in SC
These studies are mainly focused on the big data capabili- analytics.
ties of companies. There, VA is defined as ‘‘the science of Authors in [3] provided insights regarding data visualiza-
analytical reasoning facilitated by interactive interfaces’’ [4]. tion and SC activities, and some of the specific visualization
The study presented in [2] included data visualization as techniques have been mentioned. However, the connection
part of their proposed architecture for SC analytics; in this between different visualization techniques and tactics and
SC business needs have not been addressed comprehensively.
The associate editor coordinating the review of this manuscript and Tactics are approaches addressing an analytical goal with the
approving it for publication was Gianmaria Silvello . human in the loop, such as, visualizing multi-dimensionality

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
81326 VOLUME 9, 2021
A. Khakpour et al.: VA for Decision Support: SC Perspective

by clustering, or providing multiple concurrent workspaces Additionally, given the interdisciplinarity of the topic that
to analysts. In contrast, techniques are the types of visual involves Human-Computer Interaction, Information Systems,
elements, such as, histograms, bar charts, scatter plots, and so and SC Management, it requires a systematic approach to
on. Therefore, in this study, we reviewed the literature with identify previous works published from different channels.
a special focus on the identification of various visualization Although there is an increasing body of works exploring the
tactics and visualization techniques. Moreover, the authors application of VA in SC, to the best of our knowledge, there
of the study presented in [5] analyzed the challenges and is not a comprehensive study characterizing the SC processes
opportunities of using VA in organizations, and outlined and decisions that can be supported with VA.
the collection, preparing and understanding of the data as Hence, the rationale for this study is twofold: First, the
the main challenges of data analysis that can be addressed necessity to identify the specific SC decisions that can be
with the help of VA. In this context, they have investigated supported with the help of VA, and second, the need to
companies from different industries. In our study, we focus consolidate the customization that is required for the spe-
on the SC activities, to specifically formulate the benefits cific needs of the corresponding domain decision. Therefore,
of each of the VA tactics according to their use in the cor- the requirements that need to be addressed by the relevant
responding SC process. On the other hand, the study in [6] discipline research community get highlighted. Moreover,
presented a literature review underlining the need for big data VA practitioners will find a clear and precise interpretation
analytics in SC management, and the necessity for visual- of research around VA for SC activities, and eventually, all
ization is highlighted in the process of data discovery from the particularities of the potential VA system can be easily
different SC activities. We also investigate into literature on identified and implemented.
the use of data visualization, but in particular we look into Therefore, in this study, we investigate the literature to
identifying the analytical capabilities of visualization in SC identify previous attempts concerning the application of
processes. VA in different SC related activities, and we provide a holistic
Furthermore, the authors of the study [7] explored the overview of VA techniques and data that are being used
big data technologies adopted in different industrial sectors. to support decisions associated with different SC processes.
They found that VA is the most used technology with an Consequently, both SC and VA researchers and practitioners
adoption rate of 40% within different sectors, especially can benefit from this study towards enhancing SC decision
in retail industry as part of the SC networks. In general, support systems.
VA can be used for three different types of analysis in SC The rest of this paper is structured as follows: First,
processes: 1) descriptive analysis, by providing the ability to the backgrounds of the study are presented in Section II along
identify the problems from visual presentation of bottlenecks, with some of the related works that emphasized more on
2) predictive analysis, by visualizing the prospective future the need for this study. Then, the research methodology is
events, and 3) prescriptive analysis, by providing the what-if introduced in Section III along with research questions that
scenarios [8]. are the main constructs of this study. Section IV-A and IV-
Therefore, we have identified following five gaps that are B provide the findings of the study in detail. Then, we syn-
supposed to be addressed by our study: thesize our findings in section V suggesting some potential
future research directions.
1. The lack of identifying specific SC business decisions
that can be supported by VA. II. THEORETICAL BACKGROUNDS AND RELATED WORK
2. The lack of exploring various analytical capabilities of A. SUPPLY CHAIN ANALYTICS
SC VA systems. The statement provided by Forrester about management dates
3. The lack of identifying the state of the art in visualiza- back to 1958, but it still holds true especially in the scope
tion techniques and tactics. of SC management. ‘‘Management is on the verge of a
4. The lack of clear mapping between business decisions major breakthrough in understanding how industrial com-
addressed by a specific visualization tactics and the pany success depends on the interactions between the flows of
visualization techniques that have been used for that information, materials, money, manpower, and capital equip-
purpose. ment.’’ [13]. Forrester emphasized the impact of information
5. The focus of the previous studies has not been mainly flow on company success. In recent years, companies realized
on SC VA. the importance of collaborating with their SC partners in
developing the information flow throughout the SC in order
In general, majority of the previous studies are either to compete with other SCs [14].
focusing on the application of VA in other topics [9]–[11] or On the other hand, companies involved in SC activities are
even if it is considered from a SC perspective, it is mainly facing the collection of an enormous amount of data as the
focused on VA itself [12] and the SC related activity has been result of their day-to-day operations. However, given the het-
mostly used only as a case study for the proposed VA, making erogeneity and high volume of the data, the interpretation and
it difficult to identify all SC specific VA studies from the analysis of such data remains a challenging task [15]. These
literature. data are the backbone of many decision making activities

VOLUME 9, 2021 81327

 A. Khakpour et al.: VA for Decision Support: SC Perspective

towards the companies short and long term strategies [16]. C. SC OPERATION REFERENCE MODEL
In this scenario, the impact of business analytics on the The results of this study are classified according to the Supply
SC management performance has shown to be notable [1]. Chain Operation Reference (SCOR) model developed by the
In simple words, SC analytics focuses on information and Association for Supply Chain Management (ASCM) [27].
analytical tools to make decisions that better match supply ASCM is a non-profit international organization that aims at
and demand [17]. informing companies around the world to optimize different
aspects of their SC. SCOR is a reference model that describes
B. VISUAL ANALYTICS the businesses activities related to meeting the demands of
VA is defined as ‘‘the formation of abstract visual metaphors the customers in a SC environment. This model was pre-
in combination with a human information discourse (inter- viously used by researchers to ground SC analytics studies
action) that enables detection of the expected and discov- based upon to enhance the mapping of the concepts and SC
ery of the unexpected within massive, dynamically changing activities [6], [17], [28]. The SCOR reference model is based
information spaces’’ [18]. VA contributes in the process of on three underlying aspects: business processes, metrics and
analyzing heterogeneous data spaces by providing insight best practices and technology, linking them into a unified
and knowledge generation ability for decision support sys- structure [29]. ASCM defines business processes as activities
tems [19]. While the advantages of VA have been investigated that are performed to meet predefined outcomes that a SC
in many fields in the last decade, its application in SC net- must execute to fulfill the requirements of its customers [29].
works has not been analyzed rigorously. VA provides decision These processes are then divided into 6 major categories
makers the ability to combine their knowledge and expertise corresponding to different SC phases: plan, source, make,
with the computer analytical capabilities in an interactive deliver, return and enable.
manner to gain more flexible, fit to the purpose, and reliable The plan processes are defined as the activities related to
insights from the complex systems such as a SC network [20]. planning for operating the SC. These activities include identi-
The increasing amount of data intensifies the desire to fying requirements, understanding the state of the resources,
make data-driven decisions. In this respect, data analytics has balancing requirements and resources, and planning the
been identified as a subfield of decision science and informa- actions to fulfill the requirements based on the resources.
tion systems [21]. In general, there has been an increasing Source processes are defined as the activities in connection
body of work exploring the advantages of data analytics for with the ordering, delivery, receipt and transfer of raw mate-
industries and in specific for the SC players. In this context, rial or services. These activities mainly include the placement
VA has shown promising results with respect to providing the of purchase orders, planning deliveries and receiving, and
analytical capabilities to firms. The authors of the study [22], validating and storing orders. The make processes are activi-
proposed a VA system that analyzes the structural aspects ties associated with the actual conversion of the raw material
of the SC network to identify risks and provides what-if to the finished products. These activities include but are not
analysis to predict future events. The application of VA in limited to, repair, recycling, refurbishment, and manufactur-
different use cases of the SC processes is also studied by ing of the products and services. The deliver processes are
various authors. For example, the study in [23] presented the defined as activities that lead to the creation, maintenance
application of VA for retail space management, or the study and fulfillment of customer orders. The most significant
in [24] that showed the use of VA for production performance activities in this category include receiving, packing, shipping
improvement. and invoicing the orders. The return processes are activities
Literature also investigated the viability of different visu- linked with the reverse flow of the product. These activi-
alization techniques in the context of SC processes. In this ties include identifying the goods that need to be returned,
regard, in [25], authors proposed the presentation of sales shipping and receiving the returned items, and the processes
data in a labelled tree structure to facilitate the interpretation for the disposal of the returned products if required. Finally,
of the hierarchical multi-dimension data. The study in [26] the enable processes are defined as activities that contribute
focused on the temporal characteristics of sales data that used to the overall management of the SC. These activities include
charts and maps along with some specific design consider- the managing and regulating the SC rules, data, resources,
ation in order to maximize the potential of the VA system. facilities and the network in general.
Although many studies are available with regards to pro-
viding VA proposals related to SC processes, to the best of III. RESEARCH METHODOLOGY
our knowledge, there have not been any literature reviews SLRs have been successfully used in many secondary studies
analyzing those studies while focusing on the following four to determine the state of the art in various areas of com-
important aspects: uses cases of VA in SC, types of VA used puter science [30]–[32]. In this study, we have employed the
in SC, types of decisions supported, and the type of data used. general guidelines for SLRs proposed by Kitchenham and
Therefore, this study provides a comprehensive literature Charters [33]. In general, systematic reviews are defined as a
review that aims to shed light into the aforementioned aspects means of identifying, assessing and interpreting all available
and to identify previous research gaps and future research research relevant to a particular research question, or topic
opportunities. area.

81328 VOLUME 9, 2021

A. Khakpour et al.: VA for Decision Support: SC Perspective

By means of a SLR, an appropriate overview of the state of TABLE 1. Search string terms.
the art is formulated in order to determine the gaps in existing
literature and opportunities for future research. Indeed, if the
outcomes of a systematic review are formulated in the form of
a string of beads created from the literature, the corresponding
research direction will be outlined. In this context, we carried
out the following steps: 1) Recognizing the need for study,
2) Outlining the review protocol, 3) Identifying the primary
studies, 4) Assessing the quality of the identified studies, and
5) Performing data extraction. Consequently, many signifi-
cant dimensions about different aspects of the field of study
will be provided, opening the doors to the future innovations
and evolutions.

A. RESEARCH QUESTIONS
Given the scope of this study we have formulated two main
research questions (RQ) and they are further parsed into namely, IEEE Xplore, ACM Digital Library, ScienceDirect,
multiple sub questions as follows: SpringerLink, and Wiley Online Library. Then, by conduct-
RQ1. How has VA been used to support the SC activities? ing an initial search process, we have reached to a conclusion
RQ1.1. What are the use cases of VA related to the SC about the final search string and a number of specific publi-
activities? cation channels, provided in next section, in order to dig deep
RQ1.2. Which decision areas of the SC activities are into literature systematically.
supported by VA?
1) SEARCH STRING
RQ1.3. How have the SC data been used for visualiza-
tion? A search string is used to fetch the related articles from each
RQ2. Which VA techniques or tactics have been used in of the databases. The search string is formulated in such a way
SC? that the results include studies that are related to three aspects
RQ2.1. What type of data visualizations have been used? of the research: 1) Visual Analytics, 2) SC processes, and 3)
RQ2.2. What type of data analytics have been used in Decision-making. The preliminary scanning of the databases
the SC VA? for each of the aspects leaded to generation of the search
The answer to the first research question is intended to string presented in Table. 1, where a Boolean AND is used
identify VA use cases that support decision making processes to connect the terms from different aspects.
by enhancing the information retrieval from the data collected
in different stages of SC decision phases, that is, strategy, 2) SEARCH RESOURCES
planning, and operation. Therefore, we have further broke the The publication channels have been selected based on two cri-
question down into the mentioned sub questions. By answer- teria: First, they have a special focus on the computer science
ing these two questions the first two gaps mentioned earlier fields, second, they publish peer-reviewed journals, books,
in introduction will be addressed. and conference proceedings, whereas the level of quality is
By answering the second research question, we aim to assured by the publishers.
identify 1) visualization types suitable for particular ana- We then selected the articles in an iterative manner. In the
lytical goals in each of the SC activities, i.e., visualization first round of the process, we went through all the article
techniques, and 2) analytical reasoning by analysts involving titles, abstracts, and keywords of 1554 search results to either
in the process of analysis, i.e., tactics. Therefore, we can include or exclude the article into our review based on a set
further break the question down into the sub questions. These of predefined inclusion and exclusion criteria as follows:
questions contribute towards covering the gaps number 3 to • The study must be related to SC activities.
5 mentioned earlier. • The study must employ a VA technique.
• The language of the study must be English.
B. SEARCH STRATEGY • The study must be peer-reviewed.
This literature review follows a systematic approach to iden- • The study must be accessible as a full text article.
tify the related previous studies and to select relevant papers Consequently, we selected 49 articles and in the second
for the review process. In this regard, we first needed to round, we went through the full text of each of them in order
conduct multiple rounds of explorations using various key- to verify the relevance to the research questions, as well as the
words in the main research database of Google Scholar. Later, appropriateness of the article. later, we assessed their quality
by interpreting the number of search results and consulting based on a set of questions corresponding to three aspects of
with some of the domain researchers we have identified the quality as follows: 1) Selection and measurement bias,
5 main publication databases to consider as our resources, 2) Validity of the paper, and 3) Generalizability. Eventually,

VOLUME 9, 2021 81329

 A. Khakpour et al.: VA for Decision Support: SC Perspective

TABLE 2. Number of papers retrieved in each round of the review. atic reviews’ findings are mainly based on the collected data
from the previous studies and not to be biasly positive or
negative about the studies is crucial. In order to assure this, we
used extraction forms to identify all the relevant information
from each of the selected studies. In addition, a checklist
for inclusion and exclusion criteria is developed to prevent
any biases in the selection process of the papers. However,
the limitation of databases regarding the possible number of
search terms and filtration features for the search results may
impose the possibility of missing some of the relevant studies.

2) THEORETICAL VALIDITY
Theoretical validity is to identify the confusing aspects of
the study and to make sure that we seize what we aim to
the final set of papers (23) is selected, after which a rigorous seize. That is, to make sure the data we collect from each
analysis of each article has been carried out to identify the of the studies are correctly answering our research questions,
answers to our research questions based on the contents therefore the reliability of the answers to each of the research
provided. questions are not subjected to any threats. This is rather an
In order to keep track of the findings and structuring the important check, normally there is an overwhelming number
results, a data extraction form was developed. The form con- of studies that can be referred to in a SLR, however, the qual-
tains the data collected from each article about the general ity of all those studies may not be acceptable [38], and may
information and answers to the research questions. Table. impose threats to the reliability of answers to our research
2 shows the number of papers retrieved in each round of the question. This, therefore, can be addressed by a proper quality
review along with the relevant databases used in the review assurance procedure of each of the studies.
process. Given that we have conducted a quality check on each of
the selected papers, we assessed the quality of each study
C. LIMITATIONS AND THREATS TO VALIDITY based on the common research method and techniques only,
Peterson and Gencel [34] provided a guideline to identify and and we beware of evaluating papers based on their technical
prevent various threats to validity for research. We adapted methods that can lead to missing some important aspect of a
this guideline to our study in order to outline and overcome particular study.
the limitations and possible threats. According to Petersen
and Gencel., similar to empirical studies, validity consider- 3) GENERALIZABILITY
ations are also applicable to SLR studies [34]. Similar to Generalizability deals with the extent to which the study is
how other researchers considered and mitigated the potential generalizable either internally, within the corresponding com-
threats to validity in their previous SLR studies [35]–[37], munity, or externally, across other communities. In the con-
we also considered and mitigated threats to validity in our text of a systematic review, it refers to the extent to which the
study. Petersen and Gencel categorized the threats to validity results of the review can be used by various disciplines. In this
based on two main phases of research, i.e., 1) data collection, study, the results can be used by both the VA community
and 2) data analysis. These two phases are the main phases and the SC management community. As mentioned earlier,
for a SLR where we collect the data regarding our research given the multidisciplinary nature of the topic, i.e., human-
questions in each of the related works and then we analyze computer interaction, information system, and SC manage-
the collected data to provide an interpretation of findings and ment, we tried to establish a link between different fields by
a research synthesis. dividing the main research questions to sub questions com-
In this section, we present the threats to validy of this prising of different aspects relevant for each of the disciplines.
study and our approaches to minimize these threats. Based Therefore, we made sure to formulate our research questions,
on Petersen and Gencel, we discuss four different threats to and consequently, categorize our findings in such a way that
validity: 1) descriptive validity, 2) theoretical validity, 3) gen- all the disciplines can gain from the results of this study. This
eralizability, and 4) interpretive validity. In what follows we also leads to the clarity and a broader understanding of the
briefly describe each category of the threats. topic from different perspectives.

1) DESCRIPTIVE VALIDITY 4) INTERPRETIVE VALIDITY

Descriptive validity is to make sure that observations are Interpretive validity is to confirm that the inferences of
carried out objectively and they are described accurately. the study are interpreted correctly and objectively. Which
Regarding which, we need to pay careful attention to the way is another important aspect to consider while conducting a
in which we collect our data from each of the articles in order systematic review study, since the result of such studies are
not to miss any important data and to mitigate bias. System- basically based on the interpretations from previous related

81330 VOLUME 9, 2021

A. Khakpour et al.: VA for Decision Support: SC Perspective

TABLE 3. Classification of the collected data from the selected studies based on the research questions.

works. In order to prevent any misinterpretation, the data ness process of the SC. Each of the process activities involves
collection procedure has been carried out in a systematic with some sort of decision making that can be enhanced with
manner by identifying the exact answer to the RQs from each the help of VA. In this case, we scrutinized the selected studies
of the studies. The conclusion of the study is made based on in order to identify these decisions that are either supported
the answers to the research questions. by a proposed visualization system or have been the case
studies presented in the literature. Thus, the identification of
IV. RESULTS AND ANALYSIS the SCOR process addressed by each of the studies helps
Based on the research questions and the background of our linking the application of VA to each of the SC processes,
study, we have used a data collection strategy to create a which in turn answers the first RQ.
building block for presenting the results of the study. In this Table. 3 summarizes the classification of the collected data
way, the interrelation between different studies can be iden- from each of the studies based on the RQs. Based on this
tified based on their data and a conceptual framework can table, in the next section, we present the VA applications,
be developed from the analysis of these studies. To do so, the decisions they support, and the data they use correspond-
the collected data from the results of the study are categorized ing to each of the SCOR processes, providing the answer to
based on the following dimensions to answer our research the first RQ. In the latter section, we present the visualiza-
questions: the use cases of VA in the SC, the decisions a tion and analytical methods employed in each category that
VA system aims at supporting, the data that is used for VA, delivers the answer to the second RQ.
the types of visualizations employed, and the type of analytics
implemented. Moreover, these dimensions are categorized A. VA APPLICATION AREAS, DECISIONS AND REQUIRED
based on the SC business process they are involved in. DATA
One of the aims of this review is to identify the decisions In what follows, we analyze related primary studies in detail
that are supported by a VA system. Therefore, we use the with respect to the different business processes and activities
SCOR business process model, explained earlier, to recognize they are involved in. In Figure 1 we provided a quantification
the underlying decisions that are related to a particular busi- of number of studies in the corresponding business processes

VOLUME 9, 2021 81331

 A. Khakpour et al.: VA for Decision Support: SC Perspective

the planning of SC operations have been employed in various

use cases, the most significant of which are sales and demand
analysis [23], [40]–[42].
The authors of the study [43], proposed a VA system
of market sales data of agricultural products. This system
investigates the dynamic information on the business circles
and identifies the trends in the consumer demand over a
particular period. The identification of trends in consumer
demands helps to achieve auxiliary market analysis and deci-
sion making. This is an activity that is useful for different
stakeholders in the SC, taking from the manufacturers, to dis-
tributers, warehouse, and retail. This study uses the market
sales data collected from different sources and in different
FIGURE 1. Quantification of studies in each of the SCOR activities.
formats, after which they have been collected and unified in
a data pre-processing step that includes the removing of the
incomplete data, missing value handling, and noise treatment.
Therefore, data processing is an important task and given
that data is collected from various heterogeneous sources, the
process of data collection requires an underlying strategy and
technology consideration that is not discussed in this study.
In a similar effort of sales data analysis, the same authors
conducted another study presented in [40], supporting the
same set of decisions with different visualization techniques
in a different application domain. This time the focus is on the
Fast-Moving Consumer goods (FMCG) industry and particu-
larly, the sales information of a tobacco retailer. The analysis
includes comparison of the sales volume and the percentage
of different brands of cigarette in different business circles.
FIGURE 2. Categorization of papers across different decision areas The result of this analysis helps decision makers in production
provided by Wanke et. al. and retail to identify the consumer preferences and adjusting
the next ordering round based on the location of the retail.
Another analysis has been done towards assessment of the
of the SCOR model. In this section, we focus on the appli- sales stability for a particular brand in a shopping district in
cation areas that the proposed visualization system or the order to adjust sales patterns. Furthermore, the analysis of
artefact in place has been used in, the decisions that are being the sales volume for various categories of products provides
supported, and the data that has been used for analysis. This the ability to the retailers and manufacturers to select more
section intends to answer the first RQ and its sub questions. efficient marketing strategies. In addition, identifying the
In the categorization of the decisions that VA systems consumers shopping habits may enable decision makers to
are intended to support, we have also used the SC decision consider the peak of the consumer purchases and adjust their
area categorization provided by Wanke et al. [39]. These marketing activities with respect to those timings. Although
authors provided the most relevant decision areas of the the sales data hold a lot of information about consumer shop-
SC management based on the literature. On the other hand, ping activities, there are many other sources of information
the SCOR model activities are encompassing the SC pro- such as seasonal events that can be considered in order to
cesses’ decisions and the presentation of our results based on analyze sales patterns.
this model provides a supplementary categorization for the The study presented in [41], focused on VA techniques and
decision areas. Therefore, the categorization of the decisions developed a prototype to investigate the benefits of having
can be observed from two dimensions, first, the SCOR model multiple workspace capability in a VA system. The authors
activities, and second, the decision area categorization pro- argued that having the ability to create multiple workspaces
vided by Wanke et al. [39]. Figure 2 shows the categoriza- in a tab-based interface, provides users the ability to follow
tion of the studies into different decision areas provided by multiple analysis paths and pursue different hypothesis simul-
Wanke et al. taneously. Regarding the decision support aspect of their
work, they provided a user scenario of an analyst investigating
1) PLAN PROCESSES the store sales data to create sales report of stores. It is based
A set of 9 studies out of 23 papers presented in the litera- on the factors affecting the profit of each store and identi-
ture are covering the use of VA systems in the plan process fying the bottlenecks of the sales that should be addressed.
activities of the SC. These visualizations that mainly concern However, this approach has not been tested in a real word

81332 VOLUME 9, 2021

A. Khakpour et al.: VA for Decision Support: SC Perspective

scenario, and instead a sample dataset is used for evaluation be good only for some particular products and not reliable for
of the systems. some other products. Hence, in this study authors investigated
The study presented in [44], provided a VA sensemaking the model selection based on a specific product selected by
framework that supports the prediction of water demand the analyst. Since many demand analysts and decision makers
based on the historical demands and weather data. The frame- in manufacturing and retail companies are not familiar with
work is designed based on the engineering system design the machine learning techniques, the approach of visualizing
thinking, where frame and data components are defined as the model performances can help them to select the demand
two main elements of the sensemaking process. A frame forecasting models with more confidence.
is the pattern recognition concept of data mining and is One of the aims of sales data analysis is to provide com-
defined as the mental map of the situation to support the petitive intelligence [47]. Competitive intelligence is related
decision-making process. The framework considers visual- to the ability to evaluate the growth, risk, and opportunities
ization, models, insights and knowledge as frames. The ana- in connection with the market share. The authors of the
lyst is supported by tools such as visualizations, data-mining, study [47] proposed a visualization tool that demonstrates the
and statistical analysis to frame from the data and reframe total sales, trends and growth rate of the focal company in
further. This process is based on the definition of VA process comparison to one of its competitors. Using the tool in a case
provided by Keim et al. [45], that is, to analyze the data first study, an analysis of the market with the aim of stepping into
(frame), demonstrate the importance (visualize), interact and the market with a new product is demonstrated. The analyst
filter, and analyze further (reframe). looked for the weaknesses of the leading competitor by look-
The study presented in [26] is an attempt for market seg- ing at the visuals showing the current sales volume of the best
mentation and demand trend forecasting, where clustering and worst-selling products, the trends of the companies, and
analysis is used to identify the patterns in data and mapping the rate of growth in different regions. In general, decision
them with trends. Authors proposed a piecewise rank-one ten- support systems require a risk analysis that can evaluate the
sor decomposition method that slices the data into homoge- implications of decision making. Analysts require some sort
nous partitions for a comparison visualization. Using this of risk demonstration using a what-if analysis setup to make
method, a VA system is proposed for identifying the cross more confident and reliable decisions.
selling of different product groups in different times and loca-
tions, known as product segmentation. Furthermore, based 2) MAKE PROCESSES
on the geographical location of the sales, the comparison of Make processes are mainly associated with manufacturing
sales over different products in different periods and market activities and efforts for optimizing productions. In this con-
segments are carried out. Consequently, the spatio-temporal text, the study presented in [24] provided a VA system for
trend patterns of the regional sales data are analyzed, and the optimizing the production of milk and maximizing the profit.
spatio-temporal events are identified in the data slices. This study investigated the analysis of animal test record data,
Identifying customer behaviors is another VA task that sup- specifically cows, in a dairy industry, to identify seasonal
ports decision making in the plan business processes. In the and monthly variations of milk productions. The authors
study presented in [42], the quantification of the tendency evaluated the viability of the tool by means of experts’ feed-
of customers towards buying green products is discussed. back. The overall user opinion about the system was positive.
This objective can be helpful both for the production and However, the level of interactivity with the system is low and
marketing planning, as well as for sustainability decision it does not provide the ability to conduct predictive analysis.
making. In this study, the demographic data of the customers
are captured from the retail loyalty card program database 3) DELIVER PROCESSES
along with their transactions in order to summarize customers The studies related to the delivery business process activities
purchasing behaviors. Upon this investigation, the level of of SC are commonly around the approaches concerning the
the so-called customer greenness is identified based on their identification of the best sale strategy in retail stores. This
incomes. application of VA can best serve retail industries to enhance
Demand forecasting is one of the main challenges of the their product deliveries to the hand of customers and improv-
SC industries as their operations are highly correlated to the ing customers purchasing experiences. Due to the dynamic
demand pattern of the consumers. However, demand forecast- market behavior, retailers should be able to dynamically adapt
ing models are facing the challenge of market uncertainty and to the day-to-day changes of the consumers requirements.
data complexity. In this regard, the study presented in [46] In this context, the authors of the study presented in [48],
focused on developing a VA system to demonstrate the perfor- proposed a VA system for new sales strategy creation. This
mance of various models in order to support decision makers utilizes the combination of the Point-of Sales data and the
in selecting the best performing model for their demand data from the historical movement of customers in the store
forecasting activities. Both manufacturers and retailers can area. This system supports the decisions about the location of
benefit from such models towards facilitating their model the products in store, sale time of the products, and adver-
selection tasks. Authors of this study argued that a forecasting tisements, all of which are based on the customers mov-
model that is developed based on the company sales data may ing and purchasing patterns. However, in this system, the

VOLUME 9, 2021 81333

 A. Khakpour et al.: VA for Decision Support: SC Perspective

interpretation of the visualized data is very much dependent In general, retail industries are currently operating in both
on the analyst opinion and its validity is not identifiable at the online and offline settings and identifying the right sales
analysis stage. strategy for each of the setting is a significant challenge.
In a similar study related to the retail space manage- The authors of the study presented in [52] used VA to opti-
ment, the study presented in [23], developed an interactive mize the profit regarding delivery in a food SC industry.
visualization tool that supports retail stores in conducting In particular, VA is used to come up with the suggestion
analytical reasonings for the product locations in the store. that in a dual channel SC setting, improving the logistic and
In this study, the retail sales data and the store floor plan operational activities together results in the increase of the
are used to visualize the amount of sales in each product profit in both the channels. They have utilized the historical
category corresponding to their locations. The application of transaction data of both online and offline sales to develop a
VA system in this scenario helps analysts to manage space location-based pricing model and later the VA is used to make
allocation for the products in different periods based on suggestions for the respective decision makers by means of
seasonality and forecasts. It is also possible to investigate charts, feature patterns, and graphs.
the bottlenecks and failed sales strategies and modify them As observed in [51], rule-based analysis technique is con-
accordingly. However, the sales performance of a particular sidered to be a noteworthy approach for SC data analysis.
product can be affected by multiple factors and it may not Another study that utilized this technique is presented in [53].
be addressed subsequently by the relocation of the product in This study proposed the use of RFID tags to analyze the
the store. SC transportation events during the delivery activities. The
Apart from the store space management, the location of the proposed method can be used to detect and locate SC inef-
store is also an important factor towards the successful deliv- ficiencies such as delays in shipments, inventory reduction,
ery of products. In this regard, the study presented in [49], robbery, and out-of-stock cases. This study presented a visu-
developed a visual store location recommender system that alization system that enables an analyst to identify inconsis-
makes use of the sales data, transit data, and store location tencies in the SC along with a set of performance metrics such
data for identifying the attractiveness of a business district. as dwell time, transportation time and product flow.
In this study, a customer flow analysis is combined with the Transportation of goods is influenced by the traffic con-
economic insights and expert knowledge to recognize the ditions of warehouse to retail routes in an extensive way.
candidate location for a store maximizing profit. The analysis Indeed, a factor that can impact in the successful delivery
of transit data that includes the records of fare payment for of goods and products is the right placement of warehouse
public transportations is used to identify the customer flow facilities. The study in [54] proposed a VA system towards
in each district. This approach results in identifying locations enabling the comparison of different candidate locations for
as attractive business districts that are more in the business warehouses corresponding to a business district. The authors
centers of the city. employed four types of data for such analysis: urban road
With the same context of customers purchasing activity network data, GPS trajectory data from a logistics com-
analysis from their flow in the stores, the customer purchasing pany, warehouse data including information about warehouse
behaviors have been analyzed as an another important factor characteristics, and business store data consisting of store
to enhance the deliver business activity [50]. In [50], the cus- locations and characteristics. The consideration of the traf-
tomers’ response to the sales campaigns is investigated using fic condition into the selection of the warehouse location
a VA system. The customers historical purchasing behavior is responds to the uncertainties in delivery imposed by traffic.
analyzed based on the customer loyalty card system data and The study presented in [55] corresponds to the shopping
is compared with their behavior during the sales campaign to analytics tasks related to providing VA as part of customers’
identify their reaction to the campaigns. This helps decision shopping experience. The authors of this study presented a
makers to reorganize the campaigns based on their loyal user study scenario of shopping activity that provides users an
customers purchasing behaviors, providing a dynamic sales analytical interaction capability. The combination of VA and
strategy. augmented reality is called situated analytics. It provides the
However, the customer behavior characteristics changes in users with the ability to visualize the analysis of data about
different time periods and resolutions. Therefore, it is impor- a physical object. Three different types of user interaction
tant to be able to characterize the customers with different with the analysis have been introduced: filtering—the process
granularity as per day or per month and also in different of visualizing only objects of interest, finding—showing the
time periods such as daily or weekly. This is something exact object of interest to the users, and ranking—the process
that is explored in the study presented in [51]. The authors of sorting the objects of interest based on a particular criterion
proposed a VA system for association rule mining in different and visualizing the results. Additionally, a set of abstract
granularities and time periods. They used sales data of an information is provided to the user by means of bar charts.
online retail to identify the relationships between product In general, immersive technologies such as augmented and
sales and sales strategies. However, the online and offline virtual realities enable users to engage with the analysis of
sale strategies should be different as they are having different physical objects. This is an innovative approach for develop-
means. ing customers shopping experience.

81334 VOLUME 9, 2021

A. Khakpour et al.: VA for Decision Support: SC Perspective

4) RETURN PROCESSES preparation, visual representation, interaction, and iteration

Return processes consist of activities related to identifying and refinement. Regarding data preparation, authors argued
the product disruptions and acting on product recalls. In this that the focus of the analysis determines which type of data
scope, the study presented in [56] proposed a VA frame- should be collected. That is, if the focus is one firm, the data
work that detects possible product recalls by analyzing social should include all direct and indirect partners of the firm,
media comments. The sentiment analysis investigates the if the focus is on the product flow, supply relationships should
negative comments that customers provide about a product be considered, and finally, if knowledge exchange is the focus
and visualizes the output to demonstrate the necessary time of the analysis, the R&D relationships should be the main
to organize the recall. The proposed approach helps decision sources of data. Moreover, the attributes of each of the firms
makers to monitor product consumptions and predict possi- in the SC and their connections are another source of data
ble disruptions in order to reduce company’s response time. that augments both the visualization and analytical potential
Despite the importance of return processes, we identified of the system.
only one study that supports the decisions connected to these
activities. B. VA TECHNIQUES AND TACTICS
In this section, we present the visualization techniques and
tactics used in different SC processes, as well as the analytical
5) ENABLE PROCESSES supports provided by the proposed VA systems. Overall, this
Decisions about SC network management as a whole are section contributes to answer the second research question
considered as the enable business activity in the SC refer- and its sub questions.
ence model. Some of these decisions are SC coordination,
risk management, and the detection of SC inconsistencies. 1) PLAN PROCESSES
Towards supporting these decisions, related studies focused Plan processes normally require the identification of sales
on understanding the structural aspects of the network. In this trends and recognizing the amount of demands in various
context, the authors of the study presented in [22] proposed business circles. A practice known as market analysis. There
an interactive VA system that is developed based upon data is a range of visualization techniques used for this purpose,
from a multi-echelon SC that consists of various SC day- such as maps, charts, and temporal views. The study pre-
to-day operation records of different SC stages, such as, sented in [43], proposed an interactive visualization approach
procurements, manufacturing, distribution, and retail. These in which two types of Macro and Micro visualization of data
data included for instance average processing time in each are provided. The Macro display uses map visualizations to
stage, daily demand rate at each stage, and orders that each display the business circle analysis view that is achieved ear-
stage plans to satisfy. The study is a proper endeavor towards lier with the help of clustering analysis over data and dividing
demonstrating the development process of a VA system for the market data into different business circles. The map view
SC complexity analysis. provides the users with the ability to easily select a business
Indeed, understanding the structural aspects of the SC net- circle on the map and conduct detailed multi-dimensional
work lies under a broader field known as business ecosystem data analysis. On the other hand, the Micro display provides
intelligence. The application of VA in this scope is explored chart analysis of product sales data in each of the business
in [57]. This study presented a representation of the global circles. The pie charts for example, can be used to demon-
SC that helps SC managers to gain benefits from the visibility strate and analyze the market share of each of the products
into their supply network structure and key performance indi- in each business circle. The interaction level provided is to
cators (KPIs). The proposed system offers three significant choose parameter setting for the cluster analysis, controls
insights to SC decision makers: recognizing the dependency for the map displays, and personalized settings for the chart
on a particular supplier known as SC breaking points, identi- views. Although the clustering algorithm is a powerful tool
fying SC bottlenecks, and distribution of SC risks. In general, for classification of market data, the optimization of this
these insights help decision makers to analyze the business algorithm has not been considered. In addition, the level of
ecosystem and gain more transparency and risk identification interaction with the system lacks the possibility to conduct
abilities. analysis in an iterative manner, which is the requirement for
SC managers are involved in organizing many aspects of every powerful VA system.
their firm’s relations with their SC partners. One of these The same authors conducted a similar study [40] with
aspects is the level of innovations employed in the SC pro- different visualization types in order to support other
cesses. Innovation refers to the interaction between elements decision-making activities. That study proposed five dif-
of process, technology and structure [58]. The VA system ferent visualization types: Density-based clustering results
provides insights that support decision makers in the iden- map view, Nested ring diagram, Temporal K view, Theme
tification of knowledge and technology flow trends and pat- Rivers map, and a Hotspot view. The use of the different
terns in the SC network. The authors of [58] investigated the visualization methods provides the ability to analyze the
development of a VA system for SC innovations. In this study, business districts from different angles. The choice of the
the development process involves design requirements, data visualization method is very much dependent on the number

VOLUME 9, 2021 81335

 A. Khakpour et al.: VA for Decision Support: SC Perspective

of the dimensions of the underlying data. As an instance, node consisting of all the data points in a dimension, while
Nested ring diagrams combine and visualize multi dimen- the lowest level contains one node for each of the data points.
sions of brand, categories sales, and percentages at the same Each dimension tree has multiple levels, representing dif-
visual representation. On the other hand, the Hotspot view ferent intervals of various dimensions. The three key design
is based on a two-dimensional Cartesian coordinate system aspects of the proposed method are: dimension value scales,
that shows the frequency of hotspots that occur in a certain dimension relations, and filter coordination. Dimension scale
period. The type of the data dimension is another factor that is the presentation of attribute frequency in a visualization
affects the selection of the visualization method. For example, element. This study proposed the proportional tree scale that
the Theme Rivers map is chosen because of the need for is a stacked bargram presentation, where the width of each
presentation of the time dimension. However, an analysis bar shows the relative frequency of that data attribute. The
about the effectiveness of that approach is not provided. analysis of different dimension relations is another important
In general, VA is an interactive iterative process in which task that these authors proposed, the coloring of the bargrams
analysts explore and analyze data in multiple rounds of ques- itself to demonstrate the relations. Filter Coordination is also
tions and answers. However, in the conventional visualiza- introduced to provide the ability of applying interactive filters
tion systems such as those presented in [43] and [40], data to the visualization to restrict the presentation in other dimen-
visualization is done once and any modification to the data sions. In this manner, the user can select one of the categories
filtering or dimension selection alters the overall visuals and in a dimension and gets the corresponding presentation of
previous analysis paths will be lost. This issue is addressed in that category in other dimensions. This approach is to solve
another study [41], where multiple concurrent workspaces are the issue of multi-dimensional data visualization, however,
proposed as a solution for back tracking of previous analysis the user study conducted in this paper revealed that their visu-
paths. In this way, users can see multiple outcomes of their alization generates difficulties in reading data distribution and
work and analyze different aspects of the analysis at the same reading time trends, two of the important data analysis tasks
time. According to the authors, the four main design consider- related to the sales data.
ations of this VA system are: separate analysis paths, a shared In another effort regarding the multi-dimensional VA using
context between different workspaces, the ability of the user tree views [26], authors proposed a VA system that mainly
to easily navigate between workspaces, and the ability to focus on partitioning or clustering data in order to identify
review, recall, and reuse previous steps and workflows. They underlying patterns. Such VA system should fulfill the fol-
have used histograms, charts, and scatterplots for different lowing design requirements: presentation of hidden patterns
dimension preferences of the users. on different dimensions, ability for comparison of different
In the previous section, we have seen the VA process clusters of data, analysts should be able to visually identify
as a sensemaking activity that was investigated in a water if a pattern is reliable or not, the clustering task should be
demand prediction case study [44]. In this study, JMP soft- interactive, and finally, the iterative process of partitioning
ware from SAS Inc. and Tableau desktop were used for the should be trackable by the analyst in order to refine and
VA process. Based on their proposed framework, they inves- repartition. The combination of different charts and maps
tigated the relationships between weather variables and water along with an automatic partitioning algorithm provide the
supply demands. To understand the correlation, data was ability to satisfy previously mentioned design requirements.
prepared and transformed first, then visualization took place In general, clustering of data is a method that helps iden-
to understand the relationships between variables. An inter- tifying hidden patterns in data. However, the authors of the
active visualization is provided for the analysis of top-level study presented in [42], decided to use a Self-Organizing
relationships. The analysis continued by correlation analysis, Time map (SOTM) to identify the different degrees of the
stepwise regression modeling and consequently the required so called customers greenness, instead of general partition-
insights and knowledge about the supply system is gained. ing of customers into groups of green and non-green. The
The matching of the visualization framework and the visu- SOTM algorithm indeed, is another method to analyze the
alization case study demonstrates a better understanding of multi-dimensional data by viewing the data from a single
the advantages of providing a conceptual framework for variable point-of-view. This study presented the results using
VA activities that are in line with system engineering tasks. colored feature planes. In this visualization method a dimen-
From another perspective, sales and demand data analy- sion is presented upon differing degrees of the target variable
sis are normally multi-dimensional historical data and the with the help of stacked colored circles, being the level of
required VA system should preferably have the ability to sup- greenness in this case.
port many dimensions. Another study [25] argued that sales Analyzing multi-dimensional data by viewing the data
data, as a typical component of On-Line Analytic Processing from a single variable point-of-view helps the analysts to
systems, consists of well-defined dimension hierarchies that capture the hidden uncertainties related to different data
divide dimensions into intervals. Authors then proposed an points in a specific dimension. In this context, the study
interface, a hierarchical labeled tree presentation of data that presented in [46], provided five components in their visu-
can be used for sales data exploratory analysis. In this type alization system for demand forecasting model selection: an
of tree visualization, the top-level node of the tree is a single overview of the analysis, a similarity view that allows users to

81336 VOLUME 9, 2021

A. Khakpour et al.: VA for Decision Support: SC Perspective

select clusters of data manually, an item view which displays data analysis. Regarding the store space management activity,
analysis specific to a particular data item, a detail view that the tool should have some set of design requirements such
demonstrates more information related to the analysis of that as: space and time awareness, large-scale data views, ability
specific data item, and a concluding view that helps analysts to manage multiple views, interactivity of the system, and
to come to some sort of a conclusion at the end of their facility to steer the high-dimensional and multivariate data.
analysis. These different views are provided in a dashboard The 3D view of the store space provides the store managers
with a mixture of different graph-based glyph designs. that are not data analysts the ability to interact with the system
These dashboards are commonly known as multiple coor- and understand the insights from the massive amount of sales
dinated views. Demonstrating the data in different views, pro- data more efficiently.
vides the ability to the user to consider different aspects of the With the same approach of combining the map view and
data. As we already mentioned, Keim et al. [45] described the statistical analysis view, the authors of the study presented
first step of VA as analysis and then visualizing the analysis. in [49] used the subway map to provide a geographic infor-
However, [47] argues that visualization of the whole data in mation visualization. Given that the study investigated the
a single view without prior analytics in turn can provide a customers flow to identify an attractive location for retail
great ability for analysts to get an overall insight about the store, different degrees of color in the map demonstrate the
data, especially in the case of sales data. Therefore, authors level of attractiveness in different areas. The VA system
of this study proposed a pixel-based visualization that shows proposed in this study also provides statistical analysis view
each datapoint as a pixel in a matrix. Although this technique using a multi-layer circular diagram that stacks different lev-
shows good outcomes as a response to the issue of scalability, els of information on top of each other in a circular design.
in cases when sales data consists of various trends occurring Moreover, a heat map provides the recommendations for the
in multiple durations can be extremely large to demonstrate store location and a bar chart presentation of different factors
in a single view. That is where an interactive filtration feature that affect the attractiveness of each location in a comparison
permits focusing on a subset of data at a time, making analysis view. The authors provided a dashboard that consists of dif-
feasible. ferent views in order to satisfy different design requirements
that are fulfilling the objectives of the application.
2) MAKE PROCESSES Another visual analysis of the customer behaviors is car-
Regarding the make process of the reference model, ried out in [50]. In this case, a self-organizing map (SOM)
the authors of the study presented in [24] utilized various is used to first cluster the customers based on their pur-
visualization techniques to provide a VA system for milk chasing habits. Then, the customers’ behavior migration
production optimization. This study integrated visualization pattern is visualized using a feature plan representation.
techniques that are particularly useful for time-dependent SOM converts the high-dimensional data into two dimensions
multivariate data. Three visuals are implemented to satisfy and helps the presentation of clusters visually. This study
the requirements of the application. First, a scatter plot is mentioned the motivations of using SOM over alternative
used to show milk attributes (fat, protein values. . . ). In this methods as the link between the dimension reduction and
way, the dependency of attributes to the production time clustering activity, the grid structure of the visualizations,
can be identified. Second visual is a mix of a line charts ability to handle missing data and outliers, and computational
and histograms to show the density of the data and enables efficiency.
the user to verify the data quality. In this way, the value In general SOM is a visualization technique that is used
of produced milk in each day as well as the relation of the to identify the multivariate temporal patterns in data. It is,
quantity of the produced milk and milk attributes are shown. in fact, detecting similar data records that are close to each
Finally, a multiple timeline view is used to demonstrate the other. With a similar objective, association rule mining is to
overall production performance. Although multiple views discover time-dependent association rules by finding the rela-
are developed in order to provide analysis, the system lacks tionships between data records. The study provided in [51]
interactive analytic capabilities. explored the VA design for association rule mining. They
have defined five design requirements for such a system:
3) DELIVER PROCESSES the support for temporal pattern analysis, ability to do the
Multi-dimensionality of the data in the SC, such as the spatio- task of rule mining based on an item view, possibility to
temporality, requires the combination of other visualization analyze the rules visually, clear demonstration of the rela-
techniques to obtain required results. In this vein, the study tionship between data and the rules, and the ability to save
presented in [23], developed a visualization framework that and compare rules in their course of analysis. So, this study
provides 3D graphics, parallel coordinates, choropleth map, provided a multi view dashboard that consists of multiple
scatter, ternary and some other visualization elements to panels of charts, graphs and feature planes to provide the
provide a layered component based visualization toolkit. design requirements as desired.
The proposed framework is used to develop a retail store The rule-based analysis is also explored in [53] by a
management visualization system that provides interactive map-based visualization method. It demonstrates the aggre-
3D views to store managers along with the spatio-temporal gation of the product flow throughout the SC and visualizes

VOLUME 9, 2021 81337

 A. Khakpour et al.: VA for Decision Support: SC Perspective

the problematic areas by identifying the potential business layout, Matrix layout, and Substrate-based layout to highlight
issues. The system also provides the ability to the users node compositions, visualizing edges, and network structure,
to analyze the data manually by filtering capabilities and respectively. The authors also provide a what-if analysis capa-
performance metrics. These authors use different rule-based bility covering both descriptive and predictive analysis tasks.
analysis to detect some of the SC inconsistencies such as The what-if analysis is based on scenario-based techniques
Velocity, Dwell-Time, and Life-Cycle inconsistencies. rather than the advanced machine learning based methods.
Map-based visualization methods have been used by many The scenario-based analysis provides sensitivity analysis on
studies [23], [48], [49], [53], [54] to analyze various spa- each node that can be used in different scenarios such as
tial parameters such as the location of the facilities and the monitoring the impact of a sale campaign on the network per-
flow of the products. In [54] authors use a radar chart-based formance. However, the study covers only cost and demand
to demonstrate the comparison between different candidate sensitivity analysis and other high-level analytics such as the
warehouse locations. Providing the ability to compare dif- impact of various uncertainties, e.g., a disease outbreak, are
ferent visuals gives the analysts a great ability to contrast left open for future studies.
relevant information and differences in a precise way. In a similar effort to determine the structural aspects of
the supply network, in [57], authors used the force-directed
4) RETURN PROCESSES layout to provide insight into complex interdependencies and
Regarding the return processes, in [56] authors used graphs to facilitate determining the suppliers at SC breaking point.
and control attribute p-charts visualizations to provide users This study used a circular concentric layout to demonstrate
the ability to identify product disruptions. In this method, risk distribution in the network. In this visualization method,
first, sentiment analysis is used to classify customers’ feed- suppliers are shown with colored dots. The size of the dots
backs and comments as either negative or positive, then, VA is represents the importance of each supplier and the color
provided. Three different visuals are provided for exploring represents the level of the risk as being either medium or high.
these sentiments. A graph demonstrates the daily net score The connection between different entities in the network are
which is the daily transition of customers’ sentiments about a also shown using colored lines. The authors explored using of
product along with a timeline of events showing the product different visualization techniques, but the proposed tool lacks
announcements to understand user’s reaction. This view pro- the power of interactivity and some of the required analytical
vides a general idea about user’s reaction to a product rather capabilities such as predictive analytics.
than showing the number of negative or positive comments. As mentioned before, one of the implications of structural
Therefore, a second visual is provided with the help of a aspect analysis is the ability to analyze the propagation of
heatmap that presents the distribution of the comments by innovation in the supply network. In this scenario, given
showing their score and day. Finally, an attribute p-chart is that the data for the innovation circulation in the network
used to show the variation in negative comments per day. is spatio-temporal, the corresponding visualization should
Although the proposed framework provides a good means of have the ability to present such multi-dimensionality. Con-
analyzing social media data for a product disruption identifi- sequently, in [58], authors proposed the utilization of five
cation, it does not provide the ability to interact with the data types of visualization as follows: Geographic visualization
and conduct predictive analysis. for Insights regarding the spatial distribution of innovations,
Circos-based visualization for identifying the concentration
5) ENABLE PROCESSES of the knowledge, Concentric visualization for flow of knowl-
Given that enable processes are connected with the man- edge, Force-directed visualization to cluster innovations, and
agement of the SC network as a whole, the visualization Matrix visualization for shared knowledge landscape presen-
techniques used should provide a holistic view of the SC tation. Thus, different visuals to the user along with an enough
network so that decision makers and analysts can get insight set of interactions, such as selecting, filtering, and navigation,
about working condition of different nodes of the network. provide users the ability to conduct reliable analysis.
The authors of the study presented in [22] utilized a sup- In order to sum up our interpretation of previous studies,
ply network view called Force-directed layout to distinguish we provided a mapping from each of the tactics taken for
between different clusters and modules. The authors visu- tackling a particular analytical goal to the visualizations tech-
alized the structure of a network with a set of connected niques that have been implemented. In Table 4, we present
nodes. Nodes demonstrate the different stages of the net- the techniques used towards implementing each of the tac-
work and edges are the connection between the stages. This tics along with the analytical goals and their advantages or
type of visualization facilitates the identification of the pri- applications. In fact, Table 4 presents the state of the art
mary sources at the heart of the network that are acting in deploying each of the tactics with the help of the corre-
as the lifeline of the SC. Moreover, the system made use sponding visualization techniques tailored for an analytical
of one version of the circular layout called chord diagram. goal. However, only 20 studies out of 23 primary studies
This view provides a comprehensive summary of the overall found deployed a particular tactic. In a nutshell, we found
flows between different types of the activities in network. 15 different tactics towards achieving 39 analytical goals with
Three other views integrated in the system are Tree map the help of 45 different visualization techniques. Moreover,

81338 VOLUME 9, 2021

A. Khakpour et al.: VA for Decision Support: SC Perspective

TABLE 4. Summary of visualization tactics and techniques used for different analytical goals.

VOLUME 9, 2021 81339

 A. Khakpour et al.: VA for Decision Support: SC Perspective

TABLE 4. (Continued.) Summary of visualization tactics and techniques used for different analytical goals.

from the table, it is also possible to identify the application transportation management, and operation management
or the advantages of each of the tactics paired with their (Table. 3). These decision areas can be used to identify the
corresponding techniques to address a possible analytical SC requirements for developing respective VA systems.
goal. We hope this summary serves as a guide for taking SC Furthermore, in order to answer the second research ques-
VA steps further in research and development. tion, a mapping between visualization tactics towards achiev-
ing specific analytical goals regarding SC business processes
V. DISCUSSION AND CONCLUSION and the reciprocal visualization techniques has been pro-
The overall procedure of this study, from formulating to vided. Adding to that, applications, uses cases, and advan-
research questions to identified context for grounding the tages of tactics and techniques pairs are identified.
results, is illustrated in Figure 3. Therefore, both VA researchers and developers can bene-
fit from the state-of-the-art VA systems, including the data
A. OVERVIEW they utilized and the decisions and use cases they intended
In this study, we have conducted a SLR to identify, evaluate to address. We also analyzed the literature to identify lat-
and interpret previous related works regarding the use of est trends in visualization techniques and tactics that have
VA for SC processes and decision making. In this review, been utilized for some specific analytical goals in SC.
we have scrutinized the literature and classified the findings We believe, this study highlights the significant aspects that
based on a SC reference model provided by the ASCM. VA researchers and developers need, in order to outline a
We have then mapped our findings towards identifying the customized VA system for a particular SC.
use cases of VA systems in SC, the decisions they intend Overall structure of the results of this review are towards
to support, the type of visualizations employed, the type of answering the main RQs on how VA is being used in SC
analytics used, and the data that is being used for analysis. activities and which VA techniques and tactics should be used
Finally, in order to answer the first research question and for SC related activities.
its sub-questions, we have categorized the studies found in
literature based SCOR model activities, as well as the deci-
sion areas categorization provided in the literature. Based B. RQ1: HOW HAS VA BEEN USED TO SUPPORT THE SC
on the literature, we have identified 8 different decision ACTIVITIES?
areas that have been previously addressed by VA systems, Considering the RQ1, we have found that 10 studies are
namely, sales management, network design, collaborative focused on the plan business processes. Plan processes
forecasting, demand management, network integration and mainly used sales and demand data to identify trends on
visibility, production management and distribution planning, consumer demands. However, the relationships between sales

81340 VOLUME 9, 2021

A. Khakpour et al.: VA for Decision Support: SC Perspective

FIGURE 3. Research framework.

and demand and the current operations of the company or company’s production potentials. Furthermore, delivering the
the external variables, e.g., the uncertainties that can impact products to customers is an important part of the SC activities
the sales, are not considered extensively. Likewise, given that that involves almost every partner in the chain. A notable
sales and demand patterns are influenced by many factors, percentage of the studies, i.e., 35% (10/23), inspected the
the VA system should either have an ability to be scaled with identification of sales strategies regarding the location of the
external data or the use of such data should be integrated stores, the location of the products in the store, the timings for
into the system. In this regard, one of the primary studies offering products to customers, and analyzing the reactions
included consumers’ demographic data into the system to of the customers to sales campaigns, all of which are parts
analyze customers purchasing behavior. of the delivery process. As high as 45% (4/9) of the papers
Plan processes mostly include sales management and col- within the delivery processes are also mainly address the
laborative forecasting, three studies for each. After which sales management decision area. Return processes are also an
demand management have two studies following with busi- important part of the SC covering the product disruption iden-
ness intelligence processes and network integration and visi- tification and recall. It is becoming more viable with the help
bility decision areas with one paper for each. of social media data by fetching consumers’ opinion about
In general, market and sales pattern analysis can help products. In this situation, the on-time prediction of disrup-
decision makers to identify factors affecting the sales and tions can prevent future losses, the visualization of trends in
conduct comparison inspection on different product sales consumers’ sentiments both spatially and temporally assists
against different timings and periods. Moreover, we found with identifying the disruptions and acting for recall. In this
that although source and make processes are important parts regard, operation management has been identified to be the
of the SC activities involved in procurement and preparation target decision area.
of raw materials and the actual production of goods, there Finally, the enable processes concerning the management
have not been many studies to address those aspects. Pro- of the SC as a whole have been addressed mainly by analyz-
duction activities are developing a lot of data consisting of ing the structural aspects of the supply network. In general,
product attributes in different parts of the manufacturing line. network design, and network integration and visibility have
Such data can be used as a basis for VA and improve the been identified as the decisions areas within which analysis

VOLUME 9, 2021 81341

 A. Khakpour et al.: VA for Decision Support: SC Perspective

have been carried out. This type of analytics can provide learning methods, such as classification and regression anal-
several benefits to SC managers. In this regard, VA helps a ysis, remained unexplored that can provide more advanced
rapid detection of risks and SC disruptions and, at the same analytical capabilities to a VA system. The main analytical
time, inspecting the bottlenecks of the chain by identifying capability is left to the interaction of the analysts with the
the sole dependency on a particular actor in the chain in order visuals with the help of filtering and selecting the subsets of
to react accordingly. However, the problem of information data in order to identify hidden patterns on different dimen-
flow within SC partners and the lack of required data form the sions of data.
overall SC operations, act as big barriers towards integrating
such systems to the decision-making processes. We believe, D. FUTURE WORK
the development of information flow within SC partners can For future research a number of directions can be mentioned.
enhance the overall SC intelligence that can benefit all the First, the application of VA in supporting of source and
partners. make processes of the SC may be explored, such as cost
modeling of different sources and production scheduling.
C. RQ2: WHICH VA TECHNIQUES OR TACTICS HAVE BEEN
Second, visualizing the impact of external variables on dif-
USED IN SC?
ferent SC operations such the effect of weather forecast on
Concerning the RQ2, we identified 15 visualization tactics sales may be investigated more extensively. Third, we suggest
as the main approaches to address several particular ana- improving the interaction level of the visualization systems
lytical goals within the SC business processes. These tac- in order to gain more advantage of the human in the loop
tics provide the way analysts decide to intervene towards of analytics. Finally, the application of different machine
some application areas. Accordingly, for each of the goals, learning methods based on different analysis tasks and pro-
a specific visualization technique with its characteristics is viding the users with the ability of choosing the underly-
required to implement the visualization tactic under consid- ing machine learning method for analysis may be studied
eration. Table 4 provides the list of all identified tactics and in future research. We hope this review could highlight the
techniques. Adding to that, the application or advantages of previous efforts in connection with the use of VA in SC and
the each of the pairs are also recognized. The difference shed light into future research opportunities.
between application and advantages is where the application
of a particular tactic and technique pair necessarily does not
REFERENCES
provide the proof for any advantages but have been used for
a specific application in SC. [1] B. Chae and D. L. Olson, ‘‘Business analytics for supply chain: A dynamic-
capabilities framework,’’ Int. J. Inf. Technol. Decis. Making, vol. 12, no. 1,
Moreover, many visualization techniques have been identi- pp. 9–26, Jan. 2013, doi: 10.1142/S0219622013500016.
fied. Majority of the works are using dashboards that include [2] S. Biswas and J. Sen, ‘‘A proposed architecture for big data driven supply
different visuals of charts and maps to provide some sort of chain analytics,’’ SSRN J., pp. 7–34, 2016, doi: 10.2139/ssrn.2795906.
analysis capability to the analysts. However, the gap here is [3] D. Arunachalam, N. Kumar, and J. P. Kawalek, ‘‘Understanding big data
analytics capabilities in supply chain management: Unravelling the issues,
the level of interaction of the analysts with the system. In fact, challenges and implications for practice,’’ Transp. Res. E, Logistics Transp.
significant boosters for a VA system are in the abilities of Rev., vol. 114, pp. 416–436, Jun. 2018, doi: 10.1016/j.tre.2017.04.001.
analyst to interact with the system and the options they get in [4] K. A. Cook and J. J. Thomas, ‘‘Illuminating the path: The research and
development agenda for visual analytics,’’ Pacific Northwest Nat. Lab.,
logging the interactions for further reviews and comparisons. Richland, WA, USA, Tech. Rep. PNNL-SA-45230 400904120; TRN:
On the other hand, given that most of the data used are sales US200801%%906, 2005.
data and spatio-temporality is one of the main characteristics [5] S. Kandel, A. Paepcke, J. M. Hellerstein, and J. Heer, ‘‘Enterprise
data analysis and visualization: An interview study,’’ IEEE Trans. Vis.
of such data, the visualization technique should have the
Comput. Graphics, vol. 18, no. 12, pp. 2917–2926, Dec. 2012, doi:
ability to provide the analysis of different attributes against 10.1109/TVCG.2012.219.
the time and location. In this manner, map-based visualiza- [6] S. Chehbi-Gamoura, R. Derrouiche, D. Damand, and M. Barth, ‘‘Insights
tion and time-line views provide space and time dependent from big data analytics in supply chain management: An all-inclusive
literature review using the SCOR model,’’ Prod. Planning Control, vol. 31,
views for the pattern extraction support. The gap here is no. 5, pp. 355–382, Apr. 2020, doi: 10.1080/09537287.2019.1639839.
the identification and visualization of the links between the [7] E. Raguseo, ‘‘Big data technologies: An empirical investigation on their
space and time dimensions that comprises of changes in space adoption, benefits and risks for companies,’’ Int. J. Inf. Manage., vol. 38,
no. 1, pp. 187–195, Feb. 2018, doi: 10.1016/j.ijinfomgt.2017.07.008.
dimension according to the time, leading to a significant
[8] A. Diez-Olivan, J. D. Ser, D. Galar, and B. Sierra, ‘‘Data fusion
analysis capability. In general, the choice of a given visu- and machine learning for industrial prognosis: Trends and perspectives
alization technique mainly depends on the analytical goals towards industry 4.0,’’ Inf. Fusion, vol. 50, pp. 92–111, Oct. 2019, doi:
that is supposed to be achieved. As an instance, theme rivers 10.1016/j.inffus.2018.10.005.
[9] C. Vieira, P. Parsons, and V. Byrd, ‘‘Visual learning analytics of educational
are used to explore time dimensions, histograms are used data: A systematic literature review and research agenda,’’ Comput. Educ.,
for visualizing a single dimension, or tree views are used to vol. 122, pp. 119–135, Jul. 2018, doi: 10.1016/j.compedu.2018.03.018.
compare the patterns across different subsets of data. [10] C. D. G. Linhares, J. R. Ponciano, F. S. F. Pereira, L. E. C. Rocha,
The main focus of the data analytic capabilities is the J. G. S. Paiva, and B. A. N. Travençolo, ‘‘Visual analysis for evaluation
of community detection algorithms,’’ Multimedia Tools Appl., vol. 79,
clustering analysis of data, namely, density clustering and nos. 25–26, pp. 17645–17667, Jul. 2020, doi: 10.1007/s11042-020-
self-organizing maps. Here, the application of other machine 08700-4.

81342 VOLUME 9, 2021

A. Khakpour et al.: VA for Decision Support: SC Perspective

[11] A. Benito-Santos and R. Therón Sánchez, ‘‘Defragmenting research areas [34] K. Petersen and C. Gencel, ‘‘Worldviews, research methods, and their
with knowledge visualization and visual text analytics,’’ Appl. Sci., vol. 10, relationship to validity in empirical software engineering research,’’ in
no. 20, p. 7248, Oct. 2020, doi: 10.3390/app10207248. Proc. Joint Conf. 23rd Int. Workshop Softw. Meas. 8th Int. Conf. Softw.
[12] L. G. Nonato and M. Aupetit, ‘‘Multidimensional projection for visual Process Product Meas., Oct. 2013, pp. 81–89, doi: 10.1109/IWSM-
analytics: Linking techniques with distortions, tasks, and layout enrich- Mensura.2013.22.
ment,’’ IEEE Trans. Vis. Comput. Graphics, vol. 25, no. 8, pp. 2650–2673, [35] T. Tahir, G. Rasool, and C. Gencel, ‘‘A systematic literature review on soft-
Aug. 2019, doi: 10.1109/TVCG.2018.2846735. ware measurement programs,’’ Inf. Softw. Technol., vol. 73, pp. 101–121,
[13] J. W. Forrester, ‘‘Industrial dynamics. A major breakthrough for decision May 2016, doi: 10.1016/j.infsof.2016.01.014.
makers,’’ Harvard Bus. Rev., vol. 36, no. 4, pp. 37–66, 1958. [36] G. White, V. Nallur, and S. Clarke, ‘‘Quality of service approaches in IoT:
[14] R. Patnayakuni, A. Rai, and N. Seth, ‘‘Relational antecedents of infor- A systematic mapping,’’ J. Syst. Softw., vol. 132, pp. 186–203, Oct. 2017,
mation flow integration for supply chain coordination,’’ J. Manage. doi: 10.1016/j.jss.2017.05.125.
Inf. Syst., vol. 23, no. 1, pp. 13–49, Jun. 2006, doi: 10.2753/MIS0742- [37] H. Munir, M. Moayyed, and K. Petersen, ‘‘Considering rigor and rele-
1222230101. vance when evaluating test driven development: A systematic review,’’
[15] R. Y. Zhong, S. T. Newman, G. Q. Huang, and S. Lan, ‘‘Big data for supply Inf. Softw. Technol., vol. 56, no. 4, pp. 375–394, Apr. 2014, doi:
chain management in the service and manufacturing sectors: Challenges, 10.1016/j.infsof.2014.01.002.
opportunities, and future perspectives,’’ Comput. Ind. Eng., vol. 101, [38] J. P. A. Ioannidis, ‘‘Why most published research findings are false,’’
pp. 572–591, Nov. 2016. PLoS Med., vol. 2, no. 8, p. e124, Aug. 2005, doi: 10.1371/journal.
[16] E. Hofmann, ‘‘Big data and supply chain decisions: The impact pmed.0020124.
of volume, variety and velocity properties on the bullwhip effect,’’ [39] P. Wanke, H. L. Correa, and M. F. Hijjar, ‘‘Establishing the relationship
Int. J. Prod. Res., vol. 55, no. 17, pp. 5108–5126, Sep. 2017, doi: between logistics complexity and supply chain objectives and decision
10.1080/00207543.2015.1061222. areas in large companies operating in Brazil,’’ J. Oper. Supply Chain
[17] G. C. Souza, ‘‘Supply chain analytics,’’ Bus. Horizons, vol. 57, no. 5, Manage., vol. 3, no. 1, pp. 34–54, 2010.
pp. 595–605, Sep. 2014, doi: 10.1016/j.bushor.2014.06.004. [40] Q. Wu, C. Deng, R. Sun, K. Zhao, and S. Li, ‘‘Multi-dimensional informa-
[18] P. C. Wong and J. Thomas, ‘‘Visual analytics,’’ IEEE Comput. Graph. tion visualization analysis of business circles and products based on density
Appl., vol. 24, no. 5, pp. 20–21, Sep. 2004, doi: 10.1109/MCG.2004.39. clustering,’’ IFAC-PapersOnLine, vol. 51, no. 17, pp. 811–816, 2018, doi:
[19] D. Sacha, A. Stoffel, F. Stoffel, B. C. Kwon, G. Ellis, and D. A. Keim, 10.1016/j.ifacol.2018.08.095.
‘‘Knowledge generation model for visual analytics,’’ IEEE Trans. Vis. [41] M. El Meseery, Y. Wu, and W. Stuerzlinger, ‘‘Multiple workspaces in visual
Comput. Graphics, vol. 20, no. 12, pp. 1604–1613, Dec. 2014, doi: analytics,’’ in Proc. Int. Symp. Big Data Vis. Immersive Anal. (BDVA),
10.1109/TVCG.2014.2346481. Oct. 2018, pp. 1–12, doi: 10.1109/BDVA.2018.8534019.
[20] D. A. Keim, F. Mansmann, and J. Thomas. (May 27, 2010). Visual
[42] A. H. Holmbom, S. Ronnqvist, P. Sarlin, T. Eklund, and B. Back, ‘‘Green
Analytics: How Much Visualization and how Much Analytics?.
vs. Non-green customer behavior: A self-organizing time map over green-
Association for Computing Machinery. Accessed: Apr. 27, 2020,
ness,’’ in Proc. IEEE 13th Int. Conf. Data Mining Workshops, Dec. 2013,
10.1145/1809400.1809403.
pp. 529–535, doi: 10.1109/ICDMW.2013.103.
[21] H. Chen, R. H. L. Chiang, and V. C. Storey, ‘‘Business intelligence and
[43] K. Zhao, R. Sun, C. Deng, L. Li, Q. Wu, and S. Li, ‘‘Visual analysis system
analytics: From big data to big impact,’’ MIS Quart., vol. 36, no. 4,
for market sales data of agricultural products,’’ IFAC-PapersOnLine,
pp. 1165–1188, Dec. 2012, doi: 10.2307/41703503.
vol. 51, no. 17, pp. 741–746, 2018, doi: 10.1016/j.ifacol.2018.
[22] H. Park, M. A. Bellamy, and R. C. Basole, ‘‘Visual analytics for supply
08.107.
network management: System design and evaluation,’’ Decis. Support
[44] P. F. Vitiello and R. S. Kalawsky, ‘‘Visual analytics: A sensemaking
Syst., vol. 91, pp. 89–102, Nov. 2016, doi: 10.1016/j.dss.2016.08.003.
framework for systems thinking in systems engineering,’’ in Proc. IEEE
[23] M. Jern, ‘‘An information visualization approach to retail space manage-
Int. Syst. Conf. SysCon, Mar. 2012, pp. 1–6, doi: 10.1109/SysCon.2012.
ment (VisMT),’’ in Proc. 11th Int. Conf. Inf. Visualizat. (IV), Jul. 2007,
6189489.
pp. 109–116, doi: 10.1109/IV.2007.24.
[45] D. A. Keim, F. Mansmann, J. Schneidewind, and H. Ziegler, ‘‘Challenges
[24] L. Di Silvestro, M. Burch, M. Caccamo, D. Weiskopf, F. Beck, and
in visual data analysis,’’ in Proc. 10th Int. Conf. Inf. Visualisation (IV),
G. Gallo, ‘‘Visual analysis of time-dependent multivariate data from dairy
Jul. 2006, pp. 9–16, doi: 10.1109/IV.2006.31.
farming industry,’’ in Proc. Int. Conf. Inf. Vis. Theory Appl. (IVAPP),
Jan. 2014, pp. 99–106. [46] D. Sun, Z. Feng, Y. Chen, Y. Wang, J. Zeng, M. Yuan, T.-C. Pong, and
[25] M. Sifer, ‘‘User interfaces for the exploration of hierarchical multi- H. Qu, ‘‘DFSeer: A visual analytics approach to facilitate model selection
dimensional data,’’ in Proc. IEEE Symp. Vis. Analytics Sci. Technol., for demand forecasting,’’ in Proc. CHI Conf. Hum. Factors Comput. Syst.,
Oct. 2006, pp. 175–182, doi: 10.1109/VAST.2006.261422. Apr. 2020, pp. 1–13, doi: 10.1145/3313831.3376866.
[26] D. Liu, P. Xu, and L. Ren, ‘‘TPFlow: Progressive partition and multidi- [47] S. Ko, R. Maciejewski, Y. Jang, and D. S. Ebert, ‘‘MarketAna-
mensional pattern extraction for large-scale spatio-temporal data analysis,’’ lyzer: An interactive visual analytics system for analyzing compet-
IEEE Trans. Vis. Comput. Graphics, vol. 25, no. 1, pp. 1–11, Jan. 2019, doi: itive advantage using point of sale data,’’ Comput. Graph. Forum,
10.1109/TVCG.2018.2865018. vol. 31, no. 3, pp. 1245–1254, Jun. 2012, doi: 10.1111/j.1467-8659.2012.
[27] (2020). Association for Supply Chain Management—ASCM. Accessed: 03117.x.
Dec. 3, 2020. [Online]. Available: https://www.ascm.org/ [48] H. Satonaka and W. Sunayama, ‘‘Sales strategy mining support with visu-
[28] P. Trkman, K. McCormack, M. P. V. de Oliveira, and M. B. Ladeira, alization of moving history,’’ Rev. Socionetwork Strategies, vol. 7, no. 1,
‘‘The impact of business analytics on supply chain performance,’’ pp. 15–29, Jun. 2013, doi: 10.1007/s12626-013-0030-8.
Decis. Support Syst., vol. 49, no. 3, pp. 318–327, Jun. 2010, doi: [49] K. Li, Y.-N. Li, H. Yin, Y. Hu, P. Ye, and C. Wang, ‘‘Visual analysis of
10.1016/j.dss.2010.03.007. retailing store location selection,’’ J. Vis., vol. 23, no. 6, pp. 1071–1086,
[29] (2020). SCOR Digital Standard. Accessed: Sep. 15, 2020. [Online]. Avail- Dec. 2020, doi: 10.1007/s12650-020-00681-8.
able: https://www.ascm.org/corporate-transformation/DCM/scor-ds/ [50] Z. Yao, P. Sarlin, T. Eklund, and B. Back, ‘‘Combining visual customer
[30] A. Khakpour and R. Colomo-Palacios, ‘‘Convergence of gamification segmentation and response modeling,’’ Neural Comput. Appl., vol. 25,
and machine learning: A systematic literature review,’’ Technol., Knowl. no. 1, pp. 123–134, Jul. 2014, doi: 10.1007/s00521-013-1454-3.
Learn., vol. 3956, pp. 1–40, Jul. 2019. [51] H. Zhang, J. Chen, Y. Qiang, J. Zhao, J. Xu, X. Fan, Y. Yang, and X. Zhang,
[31] M. Sánchez-Gordón and R. Colomo-Palacios, ‘‘Taking the emotional ‘‘DART: A visual analytics system for understanding dynamic association
pulse of software engineering—A systematic literature review of empir- rule mining,’’ Vis. Comput., vol. 37, no. 2, pp. 341–357, Feb. 2021, doi:
ical studies,’’ Inf. Softw. Technol., vol. 115, pp. 23–43, Nov. 2019, doi: 10.1007/s00371-020-01803-x.
10.1016/j.infsof.2019.08.002. [52] C. Wei, S. Asian, G. Ertek, and Z.-H. Hu, ‘‘Location-based pricing and
[32] P. Ivson, A. Moreira, F. Queiroz, W. Santos, and W. Celes, ‘‘A systematic channel selection in a supply chain: A case study from the food retail
review of visualization in building information modeling,’’ IEEE Trans. industry,’’ Ann. Oper. Res., vol. 291, nos. 1–2, pp. 959–984, Aug. 2020,
Vis. Comput. Graphics, vol. 26, no. 10, pp. 3109–3127, Oct. 2020, doi: doi: 10.1007/s10479-018-3040-7.
10.1109/TVCG.2019.2907583. [53] A. Ilic, T. Andersen, and F. Michahelles, ‘‘Increasing supply-chain visibil-
[33] B. Kitchenham, ‘‘Procedures for performing systematic reviews,’’ Keele ity with rule-based RFID data analysis,’’ IEEE Internet Comput., vol. 13,
Univ., Keele, U.K., 2004, vol. 33, no. 2004, pp. 1–26. no. 1, pp. 31–38, Jan. 2009, doi: 10.1109/MIC.2009.10.

VOLUME 9, 2021 81343

 A. Khakpour et al.: VA for Decision Support: SC Perspective

[54] Q. Li, Q. Q. Liu, C. F. Tang, Z. W. Li, S. C. Wei, X. R. Peng, RICARDO COLOMO-PALACIOS (Senior Mem-
M. H. Zheng, T. J. Chen, and Q. Yang, ‘‘Warehouse vis: A visual ana- ber, IEEE) received the M.B.A. degree from the
lytics approach to facilitating warehouse location selection for business Instituto de Empresa, in 2002, and the Ph.D.
districts,’’ Comput. Graph. Forum, vol. 39, no. 3, pp. 483–495, Jun. 2020, degree in computer science from the Universi-
doi: 10.1111/cgf.13996. dad Politécnica of Madrid, in 2005. He worked
[55] N. ElSayed, B. Thomas, K. Marriott, J. Piantadosi, and R. Smith, ‘‘Situated with the Universidad Carlos III de Madrid, Spain.
analytics,’’ in Proc. Big Data Vis. Analytics (BDVA), Sep. 2015, pp. 1–8, He is currently a Full Professor with the Computer
doi: 10.1109/BDVA.2015.7314302.
Science Department, Østfold University College,
[56] A. Zavala and J. E. Ramirez-Marquez, ‘‘Visual analytics for identifying
Norway. He has been working as a Software Engi-
product disruptions and effects via social media,’’ Int. J. Prod. Econ.,
vol. 208, pp. 544–559, Feb. 2019, doi: 10.1016/j.ijpe.2018.12.020. neer, a Project Manager, and a Software Engineer-
[57] R. C. Basole, ‘‘Visual business ecosystem intelligence: Lessons from the ing Consultant in several companies, including Spanish IT leader INDRA.
field,’’ IEEE Comput. Graph. Appl., vol. 34, no. 5, pp. 26–34, Sep. 2014, His research interests include applied research in information systems, soft-
doi: 10.1109/MCG.2014.104. ware project management, people in software projects, business software,
[58] R. C. Basole, M. A. Bellamy, and H. Park, ‘‘Visualization of innovation software and services process improvement, and web science. He is also an
in global supply chain networks,’’ Decis. Sci., vol. 48, no. 2, pp. 288–306, Editorial Board Member and an Associate Editor for several international
Apr. 2017, doi: 10.1111/deci.12213. journals and conferences and the Editor-in-Chief of International Journal of
Human Capital and Information Technology Professionals (Emeritus, since
August 2018).

ANTONIO MARTINI received the M.Sc. degree

from the University of Parma, Italy, in 2011, and
the Ph.D. degree from the Chalmers University of
Technology, Gothenburg, Sweden, in 2015. He is
currently an Associate Professor with the Uni-
versity of Oslo, Norway. He is also a part-time
Researcher with the Chalmers University of Tech-
nology. His experience covers software engineer-
ing and management in several contexts: large,
embedded software companies, small, web com-
panies, business to business companies, and startups. He has also received
several funds for the commercialization transfer to practice of research
results. He has worked as a Principal Strategic Researcher with CA Tech-
nologies (now Broadcom) for a technology transfer project in the context
ALIREZA KHAKPOUR received the bachelor’s of a Marie Curie EU project related to the TecnioSpring+ program. He has
degree in information system and Engineering led a 10-year project focus on managing technical debt, collaborating with
from MSRIT, the master’s degree in software engi- several large companies in north- and central-Europe. He is a scientific
neering from Azad University, and the master’s advisor for a startup and is the creator of a tool to manage technical debt.
degree in applied computer science from Øst- He has also worked as a software developer and as an independent consultant.
fold University College. He is currently pursu- He has published more than 50 contributions in top software engineering
ing the Ph.D. degree with the University of Oslo. conferences and journals. His research interests include technical debt,
He is also working as a Doctoral Research Fellow software architecture, technical leadership, and agile software development.
with Østfold University College. His Ph.D. title He is active in the research community as an editor and a Reviewer for
is ‘‘Information Sharing for Customized Dynamic top IEEE software engineering conferences and journals, such as IEEE
Visual Analytics.’’ As part of his Ph.D., he is working in a project called Digi- Software and IEEE Transactions on Software Engineering. He involved in
Mat which has its aim in making the Norwegian food supply chain, smart, the organization of IEEE conferences, such as TechDebt @ ICSE, ICSA,
transparent, and sustainable. His main research interests include machine and SEAA Euromicro.
learning, visual analytics, and information systems.

81344 VOLUME 9, 2021