International May Conference on Strategic Management – IMCSM20

September 25 - 27, 2020, Bor, Serbia

International May Conference

Volume XVI, Issue (1) (2020) 42-51
on Strategic Management

GRAPH DATABASE MODELING OF A 360-DEGREE E-CUSTOMER

VIEW IN B2C E-COMMERCE
Ilija Hristoski1*, Tome Dimovski2
1
“St. Kliment Ohridski” University – Bitola, Faculty of Economics – Prilep, N. Macedonia;
2
“St. Kliment Ohridski” University – Bitola, Faculty of ICTs – Bitola, N. Macedonia

Abstract: For every B2C e-Commerce company, one of the major hurdles is the challenge of
tracking the digital footprints of each e-Customer's activities during their online shopping
sessions. As online competition becomes fiercer over time, online retailers face increasingly
more sophisticated e-Customers. Knowing their buying habits and online shopping behaviors,
which is a basic premise for building any strategies vis-à-vis retaining current and attracting
new e-Customers, creates great opportunities for those who are capable of following and
capturing relevant data about their e-Customers' digital trails. Usually part of contemporary
CRM systems, the digital profile of an e-Customer, also known as 'a 360-degree e-Customer
view', represents a collection of all e-Customers' data in one place. In this paper, a graph
database modeling framework for constructing a 360-degree e-Customer view is proposed,
with a single aim of exploring the possibilities of using NoSQL graph databases in storing
highly relational data reflecting the complex interactions between e-Customers and a
particular B2C e-Commerce website during online shopping sessions. The modeling
framework is based on the utilization of a Customer Behavior Model Graph (CBMG) and is
being implemented in Neo4j. The resulting graph database model represents a solid basis for
answering a plethora of CRM-related questions.

Keywords: 360-degree view, e-Customers, graph databases, B2C e-Commerce, Customer

Behavior Model Graph (CBMG), Neo4j

1. INTRODUCTION

The rapid growth and use of ICTs, particularly the broad proliferation of the Web 2.0
paradigm in the last two decades, have dramatically changed conventional ways of doing
business throughout the world. The e-Commerce model has become a deciding factor both for
the survival and prosperity of businesses in the contemporary global surrounding. Going
online has provided businesses with plenty of new possibilities, but several possible obstacles
to new market approaches have also been revealed. At the same time, the way how people
shop has witnessed a seismic shift, too. The advent of the B2C e-Commerce paradigm has
fundamentally changed the shopping behavior of millions of people worldwide since the
modern ICTs have transformed every aspect of the sales process, including sourcing,
browsing, searching, recommending, choosing, comparing, checking, ordering, receiving…

*
Corresponding author: [email protected]

42
 International May Conference on Strategic Management – IMCSM20
September 25 - 27, 2020, Bor, Serbia

This has contributed to a great deal of continual transformation of e-Customers, who became
highly sophisticated. Though e-Commerce has provided more opportunities than ever before,
it provides a very individual form of shopping for e-Customers. Each online shopper is
different, displaying his/her patterns of behavior. Having minded this notion, it is of the
utmost importance for online retailers to meet various e-Customers' expectations, and to
induce positive online shopping experience and satisfaction. For online businesses, the best
business strategy of all is, without question, the production of content e-Customers who might
eventually produce even more e-Customers. Satisfied e-Customers are the only prerequisite to
attract and then retain new e-Customers in the long run.
According to Roy H. Williams, a famous American businessman, "the first step in
exceeding customer's expectations, is to know those expectations." In the current, so-called
'digital economy', making organizations more customer-centric becomes a crucial task,
because the more they know about their customers, the less they need to worry about losing
them. The more they know about their customers, the more they can provide to them the
information that is increasingly useful, relevant, and persuasive. The so-called 'age of the
customer', representing a transition in power from institutions and organizations towards
customers, has changed and continues to change, the rules of business (Forrester.com, 2011).
The first step in managing the value, experience, and expectations of customers successfully
is to know who the customers are. Therefore, the design and construction of a 360-degree
view of customers and their behavior must be prioritized, since "companies that make
extensive use of customer analytics are more likely to have a considerable impact on
corporate performance, outperforming its competitors" (Fiedler et al., 2016). In that context, a
relevant, personalized, and ubiquitous customer service can be achieved solely by undertaking
actions such as identification, knowing, evaluating, developing, customizing, retaining, and
anticipating (Casariego Sarasquete, 2017, pp. 5758).
The part '360-degree' denotes 'complete' or 'all-around', whilst the part 'view' refers to
the ability to see something from a particular place or angle. Therefore, the term '360-degree
view of a customer' suggests the ability to use the best available and most relevant
information about each customer to enhance sales, marketing, and servicing decisions.
Recognizing the need for gathering all the e-Customers' interactions with a particular
virtual store during their online shopping sessions, as well as the necessity to understand and
learn from their online behavior, thus being able to anticipate their next moves, needs or
expectations, the paper focuses on developing a framework for building a 360-degree view of
e-Customers, based on the utilization of Customer Behavior Model Graphs (CBMGs) and
NoSQL graph databases.
The rest of the paper is organized as follows. Section 2 provides an overview of some
of the most prominent research, including white papers, made on the concept of a '360-degree
view of customers' during recent years. The corresponding graph database model, based on
the Customer Behavior Model Graph (CBMG) and the graph database model of an arbitrary
e-Commerce website, is being proposed in Section 3, whilst Section 4 provides illustrative
examples of how the graph database, implemented in Neo4j, can be effectively used for
answering several CRM-related questions using Cypher Querying Language (CQL). The last
section concludes.

43
 International May Conference on Strategic Management – IMCSM20
September 25 - 27, 2020, Bor, Serbia

2. RELATED RESEARCH

Due to the increasing significance, the research on the 360-degree view of customers
is becoming a focusing point of a growing number of studies worldwide. What follows is just
a glimpse of some of the most prominent research made lately on this hot topic.
Barker (2011) points out that the need to collect and manage data in a bid to recognize
consumers at any touchpoint is greater than ever. By comparing internal and external
customer data integration management, the article looks at the cutting-edge techniques for
gathering and maintaining data both on- and off-line, while taking into account single
customer databases and 360-degree views of customers. The author concludes that better data
quality management can maximize value from customers' database records, and clearer and
more accurate decisions can be made, which can ultimately improve profitability and
revenues.
Recognizing the fact that there is a significant gap regarding the usage of Customer
Knowledge Management (CKM) in practice, Vasireddy (2016) attempts to address the
complex research problem of how to manage customer knowledge to achieve a 360-degree
view of customers in organizations, by proposing a comprehensive CKM practice framework,
based on the findings from the conducted empirical studies as well as from knowledge gained
from the existent CKM literature overview.
Shahina et al. (2016) focus on the implementation of a customer-centric approach vis-
à-vis e-Commerce-based friends' recommendation systems using the Neo4j graph database,
thus enhancing their performances in terms of time and complexity.
Satish & Yusof (2017, p. 278) refer to the 360-degree view of customers from the
perspective of Big Data analytics for enhanced customer experiences with crowdsourcing, as
a means for showcase customer's interaction history with the corresponding outcomes. This
can help to know the level of customer satisfaction and dissatisfaction with various products.
From the crowdsourcing perspective, it should point out both positive and negative outcomes
of product promotions, customer loyalty, and retention. The visualizations of the 360-degree
view of customers should be able to give recommendations on how to promote products and
new customer acquisitions to enhance customer support and co-learning of different aspects,
as well as to troubleshoot errors and do anomaly detections, to conduct omnichannel pricing
and promotions, and payment flow analysis.
According to Casariego Sarasquete (2017), "Being able to anticipate your customer's
behavior is the holy grail for every business leader." The observed customer's behavior is
made up of repetitive transactions and recurring purchases, as well as navigation and
interactions through digital properties, channels, devices, applications, and social networks
that can now be tracked and stored. The research establishes the foundation for a common
data representation model of tracked customer behavior in a form of a 360-degree holistic
view of the customer and its behavior that can be used effectively by organizations to
represent any given customer-centric business model, thus allowing the analysis of the most
popular marketing problems like segmentation, cross-selling, retention, etc.
Finsterwalder (2018) investigates the 360-degree view of actor engagement in service
co-creation, pointing out the need for interpreting it as an encompassing concept with
'receptive' (i.e. psychological) as well as 'transmissive' (i.e. behavioral) properties.

44
 International May Conference on Strategic Management – IMCSM20
September 25 - 27, 2020, Bor, Serbia

3. BUILDING A MODEL OF A 360-DEGREE VIEW OF AN E-CUSTOMER

The majority of e-Customers that visit a particular e-Commerce website can be

classified into two, three, or more classes according to their similar behavior patterns they
exhibit during online shopping sessions. To describe the online shopping behavior of specific
classes of e-Customers, special types of graphs, known as Customer Behavior Model Graphs
(CBMGs) have been introduced (Menascé & Almeida, 2000; Menascé & Almeida, 2000).
CBMGs are graph-based models that characterize web shopping sessions of e-Customers
belonging to a specific class, while they are visiting a particular virtual store. Put differently,
they capture the navigational patterns of e-Customers through a particular e-Commerce
website, as viewed from the web server perspective (Figure 1).
Each particular e-Customer exhibits various behaviors during various online shopping
sessions, i.e. no two online shopping sessions, even of the same e-Customer, are identical.
The specific online shopping behavior of an e-Customer determines both the dynamics (i.e.
the frequency and the randomness) and the structure (i.e. the order) of the invoked e-
Commerce functions, such as LOGIN, REGISTER, LOGOUT, SEARCH, BROWSE,
BUY_NOW, etc. It should be notified that, due to simplicity reasons, the simplified CBMG
shown in Figure 1 does not depict many well-known e-Commerce functions that can be found
with contemporary online stores, such as SAVE_FOR_LATER, MAKE_OFFER,
ADD_TO_WATCHLIST, etc.
For instance, having minded a CBMG of a hypothetical generic e-Commerce website
(Figure 1), a specific e-Customer who knows what to buy and is highly determined to make
an online purchase, may well invoke a specific set of e-Commerce functions in the following
order:
ENTER_PAGE  LOGIN  SEARCH  VIEW_ITEM  BUY_NOW 
LOGOUT  EXIT_PAGE
On the other hand, previously, the same e-Customer, without knowing what to buy,
and, therefore, without being determined to make an online purchase, could invoke e-
Commerce functions in the following sequence:
ENTER_PAGE  BROWSE  VIEW_ITEM  …  BROWSE  VIEW_ITEM
 SEARCH  VIEW_ITEM  BROWSE  VIEW_ITEM  …  BROWSE 
VIEW_ITEM  EXIT_PAGE
Finally, the same e-Customer, who knows what to buy, but is rather reluctant, because
he/she is not determined to make an online purchase, could invoke a specific set of e-
Commerce functions in the following order:
ENTER_PAGE  SEARCH  VIEW_ITEM  LOGIN  ADD_TO_CART 
VIEW_CART  REMOVE_FROM_CART  EXIT_PAGE
Each particular CBMG, specific to an e-Commerce website, is comprised of N states,
where state #1 is always the ENTRY state (designated by letter 'E' in Figure 1), and state #N
is always the EXIT / LOGOUT state (designated by letter 'X' in Figure 1), whilst the states 2,
3, …, N1 correspond to the states HOME (1), BROWSE (2), …, REMOVE_FROM_CART
(11), respectively. Besides the characteristic set of states, a CBMG is being also described by
a set of possible transitions between two particular states i and j, designated by directed arcs
from state i to state j. The set of states and the set of possible transitions refer to the static
aspect of a CBMG since they reflect the structure of the e-Commerce website and does not
depend on the way e-Customers access and use it. All e-Customers of a particular e-
Commerce website share the same static aspect of the CBMG.

45
 International May Conference on Strategic Management – IMCSM20
September 25 - 27, 2020, Bor, Serbia

The N  N transitional probability matrix P  p i, j   pi j , whose elements are the
probabilities of transiting from state i to state j in one step, represents the dynamic aspect of a
CBMG. In Figure 1, such probabilities, which denote, in fact, relative frequencies of invoking
specific e-Commerce functions, are being designated in a form of labels pi j , assigned to
each directed arc between certain pairs of states in the CBMG.
The proposed graph database model for gaining a 360-degree e-Customer view of a
particular online retailer website is based solely on the static aspect of its specific CBMG. The
e-Customer-centric graph database model, resembling the interaction between a particular
e-Customer and a hypothetical generic B2C e-Commerce website, is depicted in Figure 2. All
N states of a CBMG shown in Figure 1 are being transformed into relationships in the graph
database model, starting from a single central node, labeled as INTERNET_USER (Figure 2),
and ending into several other nodes' types, representing corresponding entities with specific
attributes, which are constitutive parts of the generic e-Commerce relational data model
proposed by Williams (2009). This way, the invocations of various e-Commerce functions,
which reflect the interaction between a specific e-Customer and the online retail store during
online shopping sessions, are naturally represented as relationships in the graph database
model. The semantic power and the semantic context of the proposed graph database model
rely entirely on the relationships' names, directions, and attributes since graph databases are
purpose-built to store and navigate relationships. Relationships are 'first-class citizens' in
graph databases, and most of the value of graph databases is derived from these relationships.
In this particular case, the relationships connecting pairs of nodes are more valuable than the
data itself because they represent the invocations of e-Commerce functions, i.e. they reflect
the e-Customer's interaction with the e-Commerce website.
The resulting graph database could address "one of the great macroscopic business
trends of today: leveraging complex and dynamic relationships in highly connected data to
generate insight and competitive advantage", because "the ability to understand and analyze
vast graphs of highly connected data will be key in determining which companies outperform
their competitors over the coming decade" (Robinson et al., 2015).
All the relationships represented in Figure 2, besides their specific attributes, include a
set of three custom attributes: date_time (a timestamp), session_GUID (a globally unique
identifier of each particular online shopping session), and IP_address (the IP address of the
Internet user/e-Customer interacting with the e-Commerce website). These attributes are the
key ones in the process of traversing the graph during the execution of various queries that
contribute to the completion of 360-degree views of e-Customers. The date_time attribute
stores the date and time information in a format 'yyyy-mm-dd hh:mm:ss.sss' according to the
ISO 8601 international standard. The session_GUID attribute stores a 128-bit integer number,
which is both big and distinctive enough to identify each online session uniquely. Each online
shopping session starts at the moment when the Internet user accesses the e-Commerce
webpage for the first time and lasts till the moment when he/she exits it or closes the window
rendering the HTML code of the e-Commerce webpage. The IP_address attribute stores the
IPv4 address of the Internet user who accesses and interacts with a particular e-Commerce
store in dotted decimal notation.
As an example, Figure 3 displays an excerpt from the test graph database,
implemented in Neo4j, a highly scalable and schema-free (NoSQL) graph database
management system, offering ACID-compliant transactions and native graph storage and
processing. The test graph database itself is built entirely using the graph database model
presented in Figure 2 as a blueprint. It shows only the interactions of a particular e-Customer
with a specific e-Commerce website during accomplishing three of his/her online shopping

46
 International May Conference on Strategic Management – IMCSM20
September 25 - 27, 2020, Bor, Serbia

sessions, each comprised of a set of e-Commerce functions invoked in a specific order, being
mentioned previously in this Section. For demonstration purposes, the e-Customer accesses
the e-Commerce webpage from two distinct IP addresses, whilst all the values of nodes’ and
relationships’ attributes have been chosen arbitrarily.

Figure 1. CBMG of a hypothetical generic B2C e-Commerce website (rotated)

47
 International May Conference on Strategic Management – IMCSM20
September 25 - 27, 2020, Bor, Serbia

Figure 2. E-Customer-centric graph database model, resembling the interaction between

e-Customers and a hypothetical generic B2C e-Commerce website

48
 International May Conference on Strategic Management – IMCSM20
September 25 - 27, 2020, Bor, Serbia

Figure 3. Excerpt from the Neo4j-based graph database, depicting the interactions between a
specific e-Customer and a hypothetical B2C e-Commerce website during three online
shopping sessions from two distinctive IP addresses

4. ANSWERING CRM-RELATED QUESTIONS USING A GRAPH DATABASE

Table 1 contains some of the most important CRM-related questions (first column), as
well as the Cypher Query Language (CQL) programming code, for defining corresponding
queries (second column) to address the posed questions vis-à-vis the implementation of the
previously proposed graph database model in Neo4j.

Table 1. Examples of CRM-related questions and the CQL code needed for their addressing
CRM-related question Cypher Query Language (CQL) code
How many times has e-Customer 'MK2014- match (a:Internet_User)-[r:LOGIN]-
>(b:e_Customer_Profile) where
hris' logged in to the e-Commerce webpage r.e_cust_username="mk2014-hris" and
during August 2020? left(r.date_time, 4)="2020" and
substring(r.date_time, 5, 2)="08" return count(*) as
number_of_logins

What distinct IP addresses have e-Customer match (a)-[r:LOGIN]->(b) where

b.e_cust_username="mk2014-hris" return distinct
'MK2014-hris' used to log in to the a.IP_address
e-Commerce webpage so far?
Knowing that e-Customer 'MK2014-hris' match (a)-[:BROWSE]->(b) where
a.IP_address="213.135.176.85" or
has logged in to the e-Commerce website a.IP_address="213.135.177.49" return distinct
from two IP addresses, 213.135.176.85 and b.product_cat_description
213.135.177.49, what unique product
categories has he/she browsed so far?
Knowing that e-Customer 'MK2014-hris' match (a:Internet_User)-[r]->(b) where
(a.IP_address="213.135.176.85" or
has accessed the e-Commerce website from a.IP_address="213.135.177.49") and
the IP addresses 213.135.176.85 and (left(r.date_time, 4)="2020" and
213.135.177.49, how many online shopping substring(r.date_time, 5, 2)="08") return
sessions has he/she accomplished during count(distinct r.session_GUID) as shopping_sessions
August 2020?

49
 International May Conference on Strategic Management – IMCSM20
September 25 - 27, 2020, Bor, Serbia

Knowing that e-Customer 'MK2014-hris' match (a:Internet_User)-[r:VIEW_ITEM]-

>(b:Product_Item) where
has accessed the e-Commerce website from (a.IP_address="213.135.176.85" or
the IP addresses 213.135.176.85 and a.IP_address="213.135.177.49") and
213.135.177.49, how many different (left(r.date_time, 4)="2020" and
products has he/she seen during August substring(r.date_time, 5, 2)="08") return
count(distinct b.productID) as viewed_product_items
2020?
How many items did e-Customer 'MK2014- match (a:Internet_User)-[r:BUY_NOW]-
>(b:Product_Item) where
hris' buy instantly so far during all online a.IP_address="213.135.177.49" return
shopping sessions made from the IP address count(b.productID) as instantly_bought_product_items
213.135.177.49?
Knowing that e-Customer 'MK2014-hris' match (a:Internet_User)-[r:BUY_NOW]-
>(b:Product_Item) where
has accessed the e-Commerce website from a.IP_address="213.135.176.85" or
the IP addresses 213.135.176.85 and a.IP_address="213.135.177.49" return
213.135.177.49, what is the total amount sum(b.product_price) as total_amount_spent_instantly
he/she spent on buying products instantly so
far?
Regarding all online shopping sessions of match (a:Internet_User)-[r:SEARCH]-
>(b:Product_Catalog) where
e-Customer 'MK2014-hris' made from the a.IP_address="213.135.176.85" and r.keywords<>""
IP address 213.135.176.85, what keywords return r.keywords
he/she used while searching for products?
What was the online shopping session match (a:Internet_User)-[t:ENTER_PAGE]-
>(c:Home_Page) match (a:Internet_User)-
duration (in seconds) of e-Customer [x:EXIT_PAGE]->(c:Home_Page) where
'MK2014-hris', with a session_GUID = a.IP_address="213.135.177.49" and
"4cc7de0930bf4e27b40bbc968f1e2fb7", t.session_GUID="4cc7de0930bf4e27b40bbc968f1e2fb7"
made from the IP address 213.135.177.49? and
x.session_GUID="4cc7de0930bf4e27b40bbc968f1e2fb7"
return
duration.inSeconds(localtime(right(t.date_time,12)),
localtime(right(x.date_time,12)))

Numerous other CRM-related questions can be answered by a slight modification of the

relatively simple CQL queries presented in Table 1. Moreover, answers to more complex
questions (e.g. finding the exact order of invoked e-Commerce functions during a specific
online shopping session) can be found using graph traversing algorithms like Breadth-First
Search (BFS) or Depth-First Search (DFS), which are often a required first step for many
other types of analyses.

5. CONCLUSION

The proposed framework for attaining a 360-degree view of an e-Customer is based on

the transformation of states found within a CBMG of a specific e-Commerce website into
relationships of a graph database model. In a NoSQL graph database implementation, each
particular invocation of an e-Commerce function generates a relationship with its specific
attributes that contain the complete information of what, when, and how. This way, the graph
database can keep a digital trail of all e-Customers' actions during online shopping sessions,
which is in line with the definition of the 360-degree view, according to which it is a complete
full-circle view of who they are, and, importantly, it sheds light on every angle of their
interaction with the virtual store. It is the biggest benefit of this approach. On the other hand,
depending on the e-Customer's behavior, even a moderate online shopping session in terms of
the duration and intensity of invoking e-Commerce functions can generate a huge number of
relationships in a graph database. Repetitive online sessions accomplished by a single
e-Customer can contribute to a dramatic increase in the amount of data and complexity of
relationships stored within the graph database, which is the biggest disadvantage. However,

50
 International May Conference on Strategic Management – IMCSM20
September 25 - 27, 2020, Bor, Serbia

this issue paves the way for engaging other contemporary technologies, such as Big Data
analytics, having minded the fact that NoSQL databases, including graph databases, are
constituent, yet fundamental parts of such technologies. The proposed approach offers huge
potentials for answering practically all relevant questions regarding e-Customers and
therefore it can be utilized as a solid basis for building powerful CRM information retrieval
systems.

REFERENCES

Barker, D. (2011). Customer data integration: Reaching more consumers with certainty.
Journal of Database Marketing & Customer Strategy Management, 18(3), 214–219.
Casariego Sarasquete, N. M. (2017). A Common data representation model for customer
behavior tracking. Icono 14, 15(2), 5591.
Fiedler, L., Großmaß, T., Roth, M., & Vetvik, O. J. (2016). Why customer analytics matter.
Retrieved from http://www.mckinsey.com/business-functions/marketing-and-
sales/our-insights/why-customer-analytics-matter (accessed August 16, 2020)
Finsterwalder, J. (2018). A 360-degree view of actor engagement in service co-creation.
Journal of Retailing and Consumer Services, 40, 276278.
Forrester.com (2011). Age of the Customer. Retrieved from https://go.forrester.com/age-of-
the-customer/ (accessed August 16, 2020)
Menascé, D. A., & Almeida, V. A. F. (2000). Scaling for E-Business: Technologies, Models,
Performance, and Capacity Planning, Upper Saddle River, NJ: Prentice-Hall PTR.
Menascé, D. A., & Almeida, V. A. F. (2002). Capacity Planning for Web Services: Metrics,
Models, and Methods, Upper Saddle River, NJ: Prentice-Hall PTR.
Robinson, I., Webber, J., & Eifrem, E. (2015). Graph Databases: New Opportunities for
Connected Data, Second Edition. USA: O'Reilly.
Satish, L., & Yusof, N. (2017). A Review: Big Data Analytics for enhanced Customer
Experiences with Crowd Sourcing. Procedia Computer Science, 116 (2017), 274–283.
Shahina, C. P., Bindu, P. S., & Mariam Varghese, S. (2016). Enhancing the Performance of
e-Commerce Solutions by Friends Recommendation System and Neo4j Database.
International Journal on Cybernetics & Informatics, 5(2), 165171.
Vasireddy, P. (2016). Customer Knowledge Management in organisations: Developing a
practice framework to achieve360o view of customer. Ph.D. Thesis. University of
Bolton, Bolton, United Kingdom.
Williams, B. (2009). A Data Model for e-Commerce. Retrieved from
http://www.databaseanswers.org/data_models/e_commerce/ index.htm (accessed
August 10, 2020)

51