Paper title (Paper Title style)


Australasian Journal of Information Systems Balakrishnan, Das & Chattopadhyay 
2020, Vol 24, Research Article Implementing Data Strategy 

 1 

Implementing Data Strategy: Design Considerations and 
Reference Architecture for Data-Enabled Value Creation 

Radhakrishnan Balakrishnan 

Indian Institute of Management 

Raipur, India 

radhakrishnan.efpm2016@iimraipur.ac.in 

Satyasiba Das 

Indian Institute of Management 

Raipur, India 

Manojit Chattopadhyay 

Indian Institute of Management 

Raipur, India 

Abstract  

With the arrival of Big Data, organizations have started building data-enabled customer value 

propositions to increase monetizing and cost-saving opportunities. Organizations have to 

implement a set of guidelines, procedures, and processes to manage, process and transform 

data that could be leveraged for value creation. This study has approached the journey of an 

organization towards data-enabled value creation through four levels of data processing, such 

as data extraction, data transformation, value creation, and value delivery. This study has 

critical inferences on using data management solutions such as RDBMS, NoSQL, NewSQL, 

Big Data and real-time reporting tools to support transactional data in internal systems, and 

other types of data in external systems such as Social Media. The outcome of this study is a 

methodological technology independent data management framework an organization could 

use when building a strategy around data. This study provides guidelines for defining an 

enterprise-wide data management solution, helping both the academicians and practitioners. 

Keywords: Data Management; Data Management Reference Architecture; Data Strategy; 

Design considerations for Big Data; Governance for Big Data; Data-enabled value creation; 

1 Introduction 

Organizations across industries consider data as an essential factor of production similar to 

other critical assets such as labour, capital, and land. In the past, organizations were primarily 

focusing on enterprise-specific structured data, and of late, organizations have started 

collecting data regardless of the size, structure, and the speed at which it is created (Manyika 

et al., 2011). Big Data is no longer a buzz word used by a select number of technology 

companies and consultants. Today, most organizations have to deal with large volumes of 

data as their interactions with stakeholders are recorded with increasing level of granularity. 

The initial definition of Big Data confining only to Volume has expanded to include Velocity 

at which the data gets created, Variety such as structured, semi-structured and unstructured 

formats, Veracity or accuracy of data, and the Value that gets generated through data insights 

(T. H. Davenport, 2014). There is value hidden within Big Data collected by organizations, and 

the effective extraction and application of that value have become a key differentiator in the 

battle for supremacy in the marketplace.  



Australasian Journal of Information Systems Balakrishnan, Das & Chattopadhyay 
2020, Vol 24, Research Article Implementing Data Strategy 

 2 

Traditionally, organizations have been collecting and analysing data from various 

transactional systems to identify patterns and arrive at inferences with which they have 

enhanced operational efficiency and reduced cost. Now there is a plethora of additional data, 

such as extensive customer and supplier information, detailed product catalogues, more 

information surrounding online transactions, and an increase in the volume of communication 

with third parties, including email and voice recording systems. This is further supplemented 

with data available on social media (SM) that exist outside of the organization. With this 

barrage of data, organizations have to identify the right means to store, retrieve and process 

data to arrive at actionable inferences.  

The study on data management has received extensive attention in the recent past, and some 

of the topics covered in data management are summarized in Table 1. In the literature related 

to data management, ‘How do organizations build a robust data architecture to support 

different types of data management systems, and what are the critical elements of the 

proposed architecture that enable data-driven value generation?’ has been identified as a less 

researched topic. To this end, the proposed study attempts to arrive at a reference architecture, 

including different types of data management solutions, and how these systems can combine 

to drive value-creation for customers. There are two objectives of this study: (i) to identify how 

organizations support a variety of data management solutions from traditional RDBMS to 

NewSQL, and the recommended data architecture for supporting data-enabled value creation, 

and (ii) to understand the design considerations in implementing the proposed architecture. 

This study is organized as follows: In Section-2, the study presents a literature review on the 

need for data management solutions to create data-enabled value creation. In Section-3, the 

study covers the research methodology, research setup, and data analysis associated with the 

investigation. The findings are presented in Section-4, followed by a discussion on the findings 

in Section-5, and the conclusion in Section-6. The study has identified a methodological 

technology independent framework an organization could use when building a strategy 

around data-enabled value creation and delivery.  

 
Aspects of Data Management Notable Contributions 

Fundamentals on Data Analytics (T. H. Davenport, Harris, Long, & Jacobson, 2001) 

Data Analysis on Web 3.0 (Hendler, 2009; Spivack, 2011) 

Data management – Value Creation Framework (Lim et al., 2018) 

Data Management – Big Data based value creation (Mazzei & Noble, 2017) 

Data Management – Building blocks for a data-driven 

culture 
(Anderson & Li, 2017) 

Data Strategy – Guidelines for development (Bowen & Smith, 2014) 

Data Management – usage of relational database 

management systems (RDBMS) and NoSQL 
(Link & Prade, 2019; Pokorny, 2013) 

Data Management – Migration from RDBMS to NoSQL (Rocha, Vale, Cirilo, Barbosa, & Mourao, 2015) 

Data Management – Features of NoSQL systems (Mohan, 2013; Pavlo & Aslett, 2016) 

Data Management – Use of NewSQL systems (Kaur & Sachdeva, 2017; Simsek, 2019) 

Table 1: Aspects of Data Management and Notable Research Contributions 



Australasian Journal of Information Systems Balakrishnan, Das & Chattopadhyay 
2020, Vol 24, Research Article Implementing Data Strategy 

 3 

2 Literature Review 

The first section of the literature review discusses how data strategy is critical for an 

organization, and the design considerations for data-enabled value creation. The second 

section discusses the use of traditional relational database management systems (RDBMS) to 

store structured data and provides a brief on the issues related to RDBMS. The third section 

provides an overview of how NoSQL systems could be used to overcome some of the 

problems related to RDBMS. The fourth section discusses NewSQL systems and how they 

could be leveraged to build data management solutions. The last section discusses the need 

for data governance and critical elements of data governance. 

2.1 Data Strategy – A critical component of Web 3.0 

John Markoff of the New York Times had coined the term Web 3.0 to refer to the intelligent 

web such as the “applications using semantic web, natural language search, data mining, machine 

learning (ML), and artificial intelligence (AI) technologies that emphasize machine-facilitated 

understanding of information to provide a more productive and intuitive user experience” (Spivack, 

2011). Web 3.0 technologies enable firms to respond to external changes quickly by integrating 

data and applications and providing the ability to “infer relationships between data available in 

different applications or different parts of the same application” (Hendler, 2009). Web 3.0 comprises 

of technologies such as ubiquitous connectivity, network computing, open technologies, and 

the intelligent internet (Spivack, 2011). The intelligent web consists of a variety of technology 

solutions associated with data stores and data processing using natural language processing 

(NLP), ML and AI.  

In the existing literature on data management, the terms data and information have often been 

used interchangeably. This study differentiates data and information based on the widely 

popular data-information-knowledge hierarchy (Braganza, 2004; T. H. Davenport et al., 2001). 

Data are facts and observations, and in a given context, processed data becomes information 

(Zack, 1999). The information that leads to results becomes part of the knowledge. Data 

analytics is defined as the transformation of quasi-finished data into useful insights and 

knowledge (T. H. Davenport et al., 2001). The quasi-finished data goes through a process of 

cleansing, amending, and augmenting before going through further processing by statistical 

techniques, computer programs, and manual analysis.  

Organizations consider knowledge as a strategic asset, and organizations have to create, 

capture, and share knowledge to remain competitive (Zack, 1999). Organizations implement 

knowledge management mechanisms by building repositories of explicit knowledge, defining 

roles to execute and manage the knowledge accumulation process, and developing solutions 

using information technologies to support knowledge repositories and processes (Zack, 1999). 

An organization achieves data-enabled value creation by focusing on four areas, such as data 

collection, information creation, value creation, and distribution through provider networks 

(Lim et al., 2018). 

Organizations use data in a variety of ways, such as for exploring new opportunities, for 

arriving at insightful analysis on various internal and external factors, and for improving the 

organization’s data eco-system by compiling all possible data from end customers (Mazzei & 

Noble, 2017). Organizations could build innovative business models by building data-enabled 

assets and resources. Organizations, by developing analytical capabilities, could improve 

product development, customer service, and other value-chain activities (Mazzei & Noble, 



Australasian Journal of Information Systems Balakrishnan, Das & Chattopadhyay 
2020, Vol 24, Research Article Implementing Data Strategy 

 4 

2017). Organizations look at data and analytics as tools for enhancing innovation and 

operational efficiency, thus leading to additional revenue streams. Organizations could reap 

huge rewards by implementing effective means of data management solutions. Data-driven 

organizations have higher output and productivity than their less data-driven counterparts. 

Organizations could facilitate data-driven culture by implementing various measures such as 

building a repository that provides a single version of truth, building a repository containing 

data dictionary to understand what the data means, giving access to information for all 

relevant roles in the organization, and enabling managers to support data-driven decision 

making (Anderson & Li, 2017). Technology is the critical enabler for big-data aided value 

generation. However, the quality and integrity of data used for arriving at inferences depend 

on business processes and rules rather than on technology (Bowen & Smith, 2014).  

Firms generate value by analysing data available at their disposal, and the value derived from 

analysis helps on both internal operations front, and external market services front. The data-

enabled value generation options range from managerial actions such as implementing data-

driven process management to managing operations through data-driven workforce 

management and creating new market offerings based on data analytics, ML and AI 

techniques. For an organization to build data-enabled value creation, it has to deploy the right 

set of data management processes, tools, and methods for facilitating the building blocks 

necessary to create a data-driven culture. This study is an attempt in that direction to provide 

insights on a reference data architecture for handling all types of data in a corporate 

environment such as data in traditional transactional systems, data warehouses, Big Data 

repositories, schema-less data stores, and a newer variant of the scalable transactional system 

such as NewSql. The next few sections deal with various types of data, an orientation on the 

usage, shortcomings, and alternative options.  

2.2 Relational Database Management Systems (RDBMS) 

RDBMS store and represent data in two-dimensional tables. This is a simple and effective way 

of automated storage and retrieval of structured data. While using RDBMS, programmers 

employ query-based scripting languages to perform data management tasks. This continues 

to be useful when the data is in a structured format, and the scale of the data is manageable. 

RDBMS are suitable for online transaction processing (OLTP) applications and characterized 

by atomicity (A), consistency (C), isolation (I) and durability (D). The ACID properties of 

RDBMS ensure that the related transactions get completed as one unit – all or nothing, 

transactions happen independently, transactions that are written to the storage result in 

consistent data stores, and data stays durable irrespective of system failures (Pokorny, 2013).  

RDBMS have been used for decades and are mature systems, yet, in the face of ever-changing 

types of data and the complicated relationships among them, they become difficult to use. In 

RDBMS, especially in OLTP systems, data gets stored in normalized forms following Codd’s 

rules, eliminating redundancy and making sure that all changes in the data are consistent 

across the models (Link & Prade, 2019). The more complex the collection of data is, the more 

levels of hierarchy and cross relationships, the less possible it is to represent the collected data 

within the simple tabular structures of an OLTP relational database system.  

The issues associated with normalized data resulting in inefficient retrievals and lack of 

support for complex structures such as hierarchies are eliminated in data-warehouses. Data-

warehouses are relational in nature and store aggregate information that can readily be used 

for analysis and reporting. RDBMS supporting data warehouses feature non-normalized or 



Australasian Journal of Information Systems Balakrishnan, Das & Chattopadhyay 
2020, Vol 24, Research Article Implementing Data Strategy 

 5 

de-normalized data stores to facilitate easy retrieval for analysis and reporting. Data 

warehouses are built with star-schemas or snowflake schemas. Data warehouses comprise of 

dimensional stores through which hierarchies can be supported. Also, data warehouses enable 

storage of business metrics associated with various dimensional stores.  

When it comes to performance, RDBMS could scale up very well, but always with an 

associated cost. The initial investment needed for RDBMS is high as one has to pay for both 

software systems such as Oracle, and high-end servers with excellent processing capabilities. 

If one uses RDBMS for storing some reasonably common forms of data like storing the content 

from channels such as SM, articles, blogs, forums, to identify prospects in various domains, it 

will result in processing heavy-duty data. While using RDBMS for this purpose, one would 

run into issues on cost, performance, scalability, and even on building complex programming 

logic for text processing. The applications using a massive amount of unstructured data 

streaming on a real-time basis would need a scalable system with extensive storage and high 

availability (Rocha et al., 2015). RDBMS support scaling-up vertically by moving the database 

to a machine with better hardware, resulting in additional cost. The sources of data such as 

real-time feeds and SM warrant a new type of system called NoSQL that facilitates scaling-up 

horizontally (or scaling-out) by adding more low-cost machines or commodity servers (Pavlo 

& Aslett, 2016).  

2.3 ‘Not Only SQL or NoSQL’ – A Schema-less database solution 

With the arrival of Web 2.0 and Web 3.0, different types of data formats and various data 

channels emerged, and this new paradigm warranted for a different approach for data storage 

and retrieval. The modified approach to storing data in NoSQL is called CAP. CAP deals with 

data consistency - data being correct all the time, data availability - reading and writing data 

all the time, and partition tolerance - no complete failure of the system, and the system 

becomes consistent when it comes online. The type of systems based on CAP is not ACID and 

called BASE systems (Pokorny, 2013). BASE reflects the features of NoSQL systems that 

include BAsic-availability, Soft-state, and Eventual-consistency. The critical feature of NoSQL 

system is that it sacrifices robust transactional guarantees and data models of RDBMS in 

favour of eventual consistency and more flexible data models such as key-value pairs, 

document-stores, columnar stores, graph databases and likes (Pavlo & Aslett, 2016). 

Performance and Flexibility are the two critical reasons for firms to move to NoSQL systems 

(Lourenço, Cabral, Carreiro, Vieira, & Bernardino, 2015). Performance is focused on sharing 

and managing distributed data, and flexibility is the ability to handle different types of data 

such as unstructured and semi-structured data that may arise on the web. 

In the BASE model, the main emphasis is on data availability despite multiple system failures. 

NoSQL databases spread data across many storage systems with a high degree of replication. 

NoSQL solutions operate similarly as other distributed systems, providing critical features 

such as scalability, concurrency, data fail-over and recovery, and data security. The storage 

failures do not necessarily result in a complete outage of the system. In RDBMS, data is 

organized in the form of database tables, whereas NoSQL databases use different approaches 

for storing data (Sareen & Kumar, 2015). For example, a key-value NoSQL database such as 

MongoDB stores values for each key and distributes those values across the database to allow 

for easy and efficient retrieval. The NoSQL database does not have a standard schema or 

physical model that stores the meta-data for the data elements. In other words, there is no 

schema available for NoSQL databases, and they could be categorized as schema-less databases. 



Australasian Journal of Information Systems Balakrishnan, Das & Chattopadhyay 
2020, Vol 24, Research Article Implementing Data Strategy 

 6 

NoSQL systems are highly optimized for storing high volumes of data and efficient data 

retrieval. NoSQL system could be a better fit for storing blogs, articles, SM content and any 

unstructured data that does not fit into a relational database table. For NoSQL systems, 

developers do not create logical data models, and hence, data is unconstrained (Stantic & 

Pokorny, 2014). There are query tools such as XQuery and JSONiq developed to support data 

retrieval from NoSQL systems. However, they are not widely used. In cases where NoSQL 

query tools are not used, the NoSQL query options are built into the application layer. NoSQL 

systems are inherently huge in nature because of the type of data that gets saved. The usage 

of unconstrained data in NoSQL systems is generally supported by metadata such as data 

catalogues for facilitating efficient and faster data retrievals. 

NoSQL systems complement RDBMS, and hence these systems co-exist. The issues associated 

with RDBMS, such as data model change management, and data integrity enforcement, are 

eliminated in NoSQL systems. NoSQL systems are useful when the data’s nature does not 

require a relational data model (Sareen & Kumar, 2015). NoSQL systems present a different 

set of issues such as lack of designer tools to manage NoSQL systems, and lack of routines or 

options to migrate from one NoSQL system to another NoSQL system (Mohan, 2013). NoSQL 

systems do not have any standardized way of storing and retrieving data like RDBMS and 

SQL based systems, and hence, facilitating change management on NoSQL based systems 

could be frightening (Mohan, 2013). NoSQL systems are not recommended for applications 

requiring functionality, such as data integrity and security. However, they could be used for 

analysing high volume and real-time data such as web-log data, and click-stream data (Stantic 

& Pokorny, 2014). NoSQL cannot be used as the standard data store for all types of data, and 

organizations have to use NoSQL systems for the right use cases to facilitate efficient data 

storage, retrieval, and change management. This study attempts to understand use cases 

relevant for NoSQL systems and how these systems could work with existing traditional 

RDBMS supporting online transaction processing (OLTP) solutions.  

2.4 NewSQL systems 

NewSQL systems represent a newer version of the already existing RDBMS technologies. 

NewSQL systems are the ones that combine the critical features of both RDBMS and NoSQL 

systems (Pavlo & Aslett, 2016). NewSQL systems are based on distributed architecture 

operating on shared-nothing resources and contain components to support multi-node 

concurrency control, fault tolerance through replication, and distributed query processing 

(Pavlo & Aslett, 2016). This includes cloud computing providers that offer NewSQL database-

as-a-service (DBaaS) products. NewSQL DBaaS products are hosted on cloud infrastructure, 

and the customers are provided with a universal-resource-locator (URL) to connect to the 

system. DBaaS enables the pay-per-use of computing resources.  

NewSQL systems support horizontal scaling by adding new servers, and support partitioning, 

sharding, and clustering. NewSQL systems, by combining features from both NoSQL and 

RDBMS, could support both ACID transactions and enable high-availability systems. 

NewSQL systems meet many of the requirements for data management in cloud 

environments, combined with the benefits offered by traditional RDBMS (Grolinger, 

Higashino, Tiwari, & Capretz, 2013). NewSQL could be considered as an alternative to NoSQL 

systems for new OLTP applications (Kaur & Sachdeva, 2017). Some of the NewSQL systems, 

such as VoltDB, support in-memory processing of complex ACID transactions in a distributed 

manner. NewSQL systems are developed from scratch using shared-nothing architecture, or 



Australasian Journal of Information Systems Balakrishnan, Das & Chattopadhyay 
2020, Vol 24, Research Article Implementing Data Strategy 

 7 

built by augmenting the existing RDBMS, enabling them to scale-out. Through augmentation, 

the features of NewSQL could be extended on top of existing RDBMS solutions, and this 

feature would be useful for organizations that are not willing to migrate to new database 

solutions. The traditional one-size-fits-all approach of RDBMS does not hold good anymore, 

and NewSQL empowers organizations to move towards specialized database designs catering 

to their specific data storage and processing needs (Simsek, 2019). 

2.5 Data Governance  

Information is a critical resource for organizations, and it forms the basis for business 

intelligence and data analytics (Newman & Logan, 2009). Information takes on new 

importance in IT-related strategies such as service-oriented architecture. Data governance is 

defined as an organizational approach to data and information management through a set of 

policies and procedures, to manage the life cycle of data from acquisition to use to disposal 

(Korhonen, Melleri, Hiekkanen, & Helenius, 2013). With data governance, organizations could 

define guidelines and standards for data quality management and assure compliance with 

laws governing data (Weber, Otto, & Osterle, 2009). Data or information governance is a 

mechanism through which data that is considered a liability, get converted into a trusted 

strategic asset (Kooper, Maes, & Lindgreen, 2011) 

The organization’s use of Information Technology (IT) is primarily focused on operational 

efficiency; however, the evolving complex IT landscape has resulted in redundant and 

inconsistent data in enterprise environments (Korhonen et al., 2013). Incorrect and erroneous 

data add additional costs and risks and lead to poor managerial decisions. Ever since the 

inception of Big Data and streaming data, organizations have started considering data they 

collect from various internal systems and external sources, as assets. Information assets are 

facts that have value or potential value that are documented (Khatri & Brown, 2010).  

The critical business benefit of information governance is the realization of data as a valuable 

and manageable organizational asset, and other business benefits include improved business 

decisions due to accurate data, and increased user trust within the organizations’ data sources 

(Marco, 2006). Data governance strategy results in improved reliability, traceability, and 

authenticity of data, thus providing the backbone communication platform across all 

stakeholders.  

Some of the critical elements of data governance include data quality, meta-data management, 

and data life-cycle management. Having high-quality information is a pre-requisite to 

building data-driven customer value propositions (Weber et al., 2009). Hence, it is critical for 

organizations to focus on data quality so that they could make effective managerial and 

strategic decisions. Data quality management helps an organization to plan, provision, 

organize, use, and retire high-quality data (Weber et al., 2009). For a firm to implement the 

right governance program, it has to manage various aspects of data management such as 

information security, risk management, data privacy, data security, data loss, and business 

continuity. Organizations in the digital era have to adopt a holistic approach to manage the 

three critical challenges associated with data management such as data security, data privacy 

and compliance obligations (Salido, 2010). These problem areas of data management have to 

be adeptly supported by three capabilities of firms such as people, process and technology. 

Organizations design three kinds of governance mechanisms such as decision-making 

structures, alignment processes, and formal communications (Weill & Ross, 2005).  



Australasian Journal of Information Systems Balakrishnan, Das & Chattopadhyay 
2020, Vol 24, Research Article Implementing Data Strategy 

 8 

Successful information governance warrants accountability, and accountability is executed 

through a formalized process called stewardship (Newman & Logan, 2009). The people aspect 

of data management deals with identifying data-stewards - people collectively responsible for 

data quality in the organization by defining policies and procedures for the classification, 

protection, use, and management of data (Salido, 2010). The data stewards engage in 

information governance by implementing managerial activities such as setting standards, 

enforcing policies around data, resolving intra-team issues and implementing day-to-day 

operational procedures (Newman & Logan, 2009).  

The process aspect of data management deals with defining a set of processes to manage 

elements such as information security needs and compliance needs of the organization. The 

process aspect of information governance deals with monitoring and reconciling data at 

critical points along data’s path through the organization’s IT systems, and this includes initial 

data entry points and data aggregation points (Griffin, 2005). At the data-entry points, the IT 

systems could check for valid values or required values. At data aggregation points, the 

process steps could include eliminating duplicate data entries and checking for valid cross-

references to strengthen the authenticity of data. Also, the process options include features for 

manual corrections to exceptions and errors that cannot be fixed automatically (Griffin, 2005).  

The technology aspects of information governance and data management deal with activities 

such as providing secure infrastructure, implementing identity and access control, 

strengthening information protection features, and implementing robust auditing and 

reporting functionalities (Salido, 2010). The software solution considered for big data-related 

data management solutions should scale to accommodate large datasets, leverage the 

underlying hardware platforms efficiently, and bridge the increasing gap between growing 

data and computing power (Kambatla, Kollias, Kumar, & Grama, 2014).  

The terms data governance and data quality are often used interchangeably, but both are different 

in theory and practice. Data quality happens at both the front-end through data validations 

and back-end through databases. The process of data quality is defined, structured, and 

implemented through a well-designed data governance framework. In the data governance 

framework, rules and policies related to data ownership, data processes, and data technologies 

are clearly defined. In other words, data governance provides a framework for managing data 

quality (Ghosh, 2019).  

The existing literature on data management covers a variety of topics, as mentioned in Table 

1. However, in the current research of data management, queries such as ‘How do 

organizations support variety of data management solutions ranging from traditional RDBMS 

to NewSQL, and how do organizations integrate them for facilitating data collection, 

aggregation, and analysis?, ‘What is the recommended data architecture for supporting data-

enabled value creation?’, and ‘What are the design considerations in implementing the 

proposed architecture, in terms of data governance, data security, disaster management, and 

business continuity?’, have been identified as less explored, and recommended for research. 

This study attempts to delve into various aspects of data management, to arrive at inferences 

on reference architecture and design considerations to enable data strategy for customer value 

creation.  



Australasian Journal of Information Systems Balakrishnan, Das & Chattopadhyay 
2020, Vol 24, Research Article Implementing Data Strategy 

 9 

3 Research Methodology 

3.1 Research setting 

This study is based on an explorative approach to understand data management solutions 

used by organizations, and the inferences are generic in nature. The study is executed through 

interviews with three data architects focusing on building solutions using the latest data 

management tools and technologies. All three data architects have significant industry 

experience, worked with multi-national corporations, and architected complex data 

management solutions for global clients. One of the architects – A1, has been working on 

building solutions for clients in North America. Another interviewee – A2, has been working 

with clients in Western Europe and the Americas, and the third one – A3 has experience in 

building solutions for e-commerce, banking and financial services clients in North America.  

This research is conceptual in nature, and there are no control variables, moderating, and 

mediating factors considered for this study. This study includes interviews that have enabled 

the researchers to understand the data architecture used for creating a robust data 

management solution covering all design aspects, end-user aspects, and governance aspects. 

The interviewees have provided documents on processes used while building data 

warehouses and presented architecture diagrams that accommodate various data 

management solutions.  

The study involves interviewing three candidates, and for insightful qualitative research, the 

number of interviewees may not be considered sufficient, as interviewing three people does 

not satisfy the condition of having a representative sample. However, the three senior people 

found for this study A1, A2 and A3 are playing leadership roles in their respective teams. A1, 

as a team lead, manages fifteen data architects, A2, manages twenty-five data architects, and 

A3 manages fifteen data architects. All interviewees have exposure to variety of business 

challenges in data management domain, and they have provided solutions to complex 

business problems, with help from team members. They have a collective knowledge base on 

various data management techniques and worked on all tools and technologies related to data 

management. This criterion qualifies them as subject matter experts (SME) who could provide 

insights on all aspects of data management. The SME has profound knowledge in the subject, 

and their experience was evident as they could comment and voice their opinion on all 

prominent and latest methodologies and technologies used for data management. All the 

chosen interviewees are from different organizations, and hence their comments and opinions 

are independent, and not influenced by other members. In addition to these discussions, the 

researchers have met with the team members to have informal discussions on finer aspects of 

data management. During those meetings, the team members had shared artifacts covering 

the data management solutioning and implementation details for various clients. The 

researchers have used the transcripts, documentation, and artifacts to arrive at the inferences 

and contributions for this paper. 

3.2 Research Design and Data Collection 

The scope of the research is limited to four dimensions of data-enabled value creation, as 

suggested by Lim et al. (2018), and the dimensions include data collection, information 

creation, value creation, and distribution or information delivery. The scope is technical in 

nature, and covers governance aspects of data management such as information security, data 

privacy, and data quality. The queries covering these four dimensions are created in 



Australasian Journal of Information Systems Balakrishnan, Das & Chattopadhyay 
2020, Vol 24, Research Article Implementing Data Strategy 

 10 

consultation with faculties teaching courses on data management for under-graduate 

engineering students. The study does not focus on covering the aspects related to hardware or 

infrastructure components associated with data management.  

The study comprises of interviews with subject matter experts, and the research is exploratory 

in nature. The researchers have avoided any leading queries to interviewees and included 

probing queries to offer more clarity on the base queries when needed. The questions are open-

ended, and some of the queries used for this study are, “How are various data management 

systems integrated? From your area of expertise, could you provide insights on why the integration is 

needed, and how it was implemented?”, “How would you handle exceptions while loading different 

types of data?” “What is the new customer value proposition that comes with the data exploration 

tools?” “How does your new value proposition get monetized? Could you give specific scenarios on how 

you helped your client monetize on data offerings?”, and “What kind of data governance mechanisms 

you have implemented? What are the design considerations for business continuity?”. The study is 

conducted without any underlying assumptions, and the researchers have personally met all 

the three data architects, conducted interviews and transcribed the audio files. The transcripts 

are reviewed with the interviewees for approval, thus enhancing the accuracy of collected data 

(Huber & Power, 1985). The study comprises of 150 minutes of audio recordings, and 16000 

words in transcripts. Also, the authors have referred to various process documentation, and 

discussions with critical team members through informal question-and-answer sessions to 

answer queries, for a period of another 140 minutes;  

3.3 Data Analysis 

The researchers have transcribed the recorded conversations and validated the transcribed 

content through the maker-checker process, a dual-approval process typically used in the 

financial services industry (Commercial Bank, 2018). In this dual-approval process, a maker 

enters the transactions into a system, and a checker reviews these transactions and approves 

the same. For this study, an author analysing the transcript files, gets his analysis reviewed 

and approved by the other author. The researchers have analysed the data, verified each 

other’s work, and frequently referred to both the data and existing literature when developing 

the research report. This study has identified findings across four aspects of data-enabled 

value creation, and these inferences are summarized in the next section.  

4 Research Findings 

The study proposes a methodological technology independent framework an organization 

could implement to support data-enabled value creation. All the critical elements of the 

structure supporting data-enabled value creation are portrayed in Fig.1. These components 

are discussed at length in subsequent sections. The first section provides details on different 

types of data analytics solutions implemented by organizations. The second section contains 

the findings on data integration methods adopted by firms, explicitly covering all types of data 

such as structured data, semi-structured data, and unstructured data. The third section 

summarizes the findings on data transformation methods used by firms to create information 

that can be part of the knowledge. The fourth section lists the findings on the value creation 

resulting from such an architecture, and the last section summarizes the findings on other 

business critical considerations such as data quality, data security, and business continuity.  



Australasian Journal of Information Systems Balakrishnan, Das & Chattopadhyay 
2020, Vol 24, Research Article Implementing Data Strategy 

 11 

 

Fig.1 Reference Architecture for creating Big Data enabled value creation 

4.1 Use Cases for various data management technologies 

One of the data architects has spelled out the use cases of these data management technologies, 

and opined, 

“anything that is transactional in nature – where you have structured schema and transactions 

…things like ERP and stuff – RDBMS is the preferred choice; … when you have aspects related 

to data integrity and things, that is where RDBMS comes into play; …. the challenge comes as 

(when) the data volume increases; the only way you can handle volumes is to grow vertically; 

like you have to increase the processing capability of the server. ….we had influence of social 

media data coming; influence of semi-structured data coming; semi-structured in the form of 

XML, JSON, and other files have come; these things resulting in change in data structures as 

well; that is how we got NoSQL; …lots of vital data that needs to be stored and does not need 

to be in an integrated manner (data integrity need not be there) then, we have to go for NoSQL 

systems … a new generation data management solution – you could handle big volumes and 

scaling like NoSQL – that is where NewSQL comes into (the) picture; low latency data 

retrievals – huge storage, scaling up - horizontally, with very low latency;” (A2, Data and 

Solution Architect) 

Data management solutions are implemented for descriptive, predictive and prescriptive 

analytics solutions. The following table lists the suggestions as indicated by the interviewees, 

on the type of data management solutions possible with different types of data. 

 

 

 

 

 

 

 

 

 

 

-- 

 

 

 

 

 

 

 

Internal sources 

CRM 

Retail sales 

ERP 

Ecommerce 

Inventory 

External sources 

Twitter 

Facebook 

E 

T 

L 

Data Lake 

EDW 

HR 

Order  

Inventory 

Customer 

NoSQL or 

Big Data 

Environment 

S
e
n
ti
m

e
n
t 

m
e
tr

ic
s 

P
ro

d
u
c
t 

A
ff

in
ity

 m
e
tr

ic
s 

S
e
rv

ic
e
 Q

u
a
lit

y
 m

e
tr

ic
s

 

Enterprise 

Functional         

metrics 

Social 

Media 

metrics 

NoSQL or 

Big Data 

Environment 
EDW 

Fail-over/Secondary Instance 

Streaming 

Analysis 

Business analytics 

  

  

  

Real time trends 
Sentiment metrics 
Other Social Media metrics 

Passive replication 
Active replication 

E 

L 

T 

.... 

Data Marts 



Australasian Journal of Information Systems Balakrishnan, Das & Chattopadhyay 
2020, Vol 24, Research Article Implementing Data Strategy 

 12 

Type of Data 

Management 

Technology 

Descriptive Analytics Solutions 
Predictive and Prescriptive Analytics 

Solutions 

RDBMS  

 

(Supports 

Structured Data 

and ACID 

transactions) 

Various reporting solutions based on 

Customer Relationship Management (CRM) 

and Enterprise Resource Planning (ERP) 

applications; Management Information 

Systems (MIS) generating daily reports;  

Retail: Recommendation solutions 

 

Banking: Fraud monitoring solutions 

 

Health-care: Identifying the right solution 

based on given health conditions 

(Prescriptive solution) 

NoSQL 

 

(Supports 

unstructured and 

semi-structured 

data) 

Statistics information such as tag-clouds, 

sentiments on social-media content, and 

visual reporting on hierarchical information   

Arriving at the sentiment trend based on the 

tone of sentiments collected over a period of 

time; 

 

Predicting the most-preferred features in a 

new product or service based on the 

comments collected from semi-structured and 

unstructured data sources 

NewSQL 

 

(Structured data) 

Similar to RDBMS; However, NewSQL is 

mostly used for real-time analytics and 

predictive analytics 

 

 

Streaming analytics – on customer 

sentiments; Identifying trends in user 

perceptions; Real-time recommendations on 

product purchases; Streaming analytics on 

data from IoT; Market basket analysis in 

Retail industry; 

Solutions 

supporting Big 

Data 

 

(HDFS based 

systems) 

All types of data; 

 

Mainly used as ‘Data staging’ area and data 

archival area; Helpful to deal with ‘Save now 

and analyze later’ kind of data; 

Staging area for Extract-Load-Transform 

(ELT) kind of data extraction;  

 

Storage for archived-data wherever needed; 

 

System can support all type of data to 

support descriptive, predictive and 

prescriptive analytics solutions 

Table 2: Data Management Technologies and Use Cases supported 

4.2 Data Integration 

Data integration involves combining data from disparate and heterogeneous source systems 

to arrive at a unified view of the data. Data integration consolidates applications within one 

firm to provide a unified view of the firm’s data assets (“Data Integration,” 2018). Data 

integration enables businesses to provide users with a real-time view of business performance 

by combining data residing in different sources, and inclusion is the first step towards 

transforming data into meaningful insights such as business metrics and key-performance-

indicators (KPI) (Doyle, 2017).  

Organizations use various means to extract data from source systems, integrate them, arrive 

at different business metrics, and store those KPI in a relational data warehouse that could 

drive further analysis. Organizations could use either ETL (Extract-Transform-Load) or ELT 

(Extract-Load-Transform) to extract data from the source systems. When using ETL, the data is 

extracted from source systems, transformational rules are applied to the data, all necessary 

business metrics are calculated, and the transformed information is loaded into the target 

database. In the case of ELT, the extracted data is loaded into a temporary work area called 

‘staging area’, transformation rules are applied to arrive at business metrics, and the 



Australasian Journal of Information Systems Balakrishnan, Das & Chattopadhyay 
2020, Vol 24, Research Article Implementing Data Strategy 

 13 

transformed information or business metrics is loaded into the target data warehouse. To 

answer a specific query on the right use cases for using either of ETL or ELT, A3 claims, 

“If the systems you extract data from (source systems) are all local (within your 

internet/intranet), then ETL would be good; there won’t be much of latency issues here as the 

data is internal; the data you have on external systems – especially social media, etc – you extract 

data, you load them as-is into some sort of temporary store (data staging area) – and then 

transform and load/persist into data warehouse; (Use ELT) “ (A3, Data Architect) 

ETL is the standard way of data extraction when sourcing from internal transaction processing 

systems. However, data from SM such as Twitter and Facebook get loaded first into a data 

store like NoSQL, necessary metrics such as sentiment scores are calculated, and the results 

are loaded into a relational data warehouse. SM data is usually extracted through scripts, 

software programs, and crawler programs.  

An ETL or ELT process might throw exceptions or errors for data that cannot be processed 

completely. These exceptions are sent back to source systems, and alert messages are raised to 

business users. Business users view error records through user interfaces or applications 

wherein they would have options to fix the data, and resubmit for processing. The data 

extraction and load procedures maintain data integrity when loading data into target systems, 

and reconcile target data with source systems to load all relevant information into target 

systems. The data extraction process, depending on the business rules, could be set as an 

incremental-refresh or full-refresh procedures. In the case of incremental-refresh, new transactions 

since the last incremental-refresh, will get extracted from the source system, and loaded into the 

target system. The incremental transactions are usually identified through a timestamp field 

in transactions. In the case of full-refresh, the entire target data gets overwritten by the source 

data. The following quote summarizes how incremental-refresh and full-refresh data extraction 

procedures work: 

“When you do data extraction, and if you know that the data (source data) is changing, then, 

you will have to set that as a full refresh; if the data doesn’t change much, it has to be incremental 

refresh; if I have to quote example from the credit card expense transactions, the transactions 

happen on a daily basis, and no need to do full refresh (on those daily transactions); likewise, 

your banking transactions – no need to do full refresh as all the transactions like deposits, 

withdrawals, etc – are incremental in nature; and also, once you do a transaction you hardly 

change anything; maybe there is a credit onto your account, but that is a new transaction – 

which can again be incremental; full refreshes happen when you set up the target (data) for the 

first time; also – things like billing for one of your corporate clients: there, may be billing has to 

be done as a full refresh on a periodic basis – say 3 months or 6 months or even a year;” (A3, 

Data Architect) 

Data transformation is a process through which the data extracted from source systems is 

transformed to reflect the precise business metrics used for end-user analysis. The data 

transformation happens either at the source system or at the target system after the load to the 

staging area is completed. For data extraction and transformation, one could choose to 

transform at the source, provided the source systems are internal to the firm or available 

within the firm’s intranet. Data extracted from SM and external sites is brought inside the 

firm’s intranet, loaded into the staging area, transformed to arrive at metrics such as sentiment 

scores and product affinity scores, and loaded into a target data warehouse for further 

analysis. 



Australasian Journal of Information Systems Balakrishnan, Das & Chattopadhyay 
2020, Vol 24, Research Article Implementing Data Strategy 

 14 

Organizations could implement data transformation using in-memory techniques that are part 

of new data management solutions called NewSQL systems. NewSQL technologies offer in-

memory computing features that support real-time analysis of data. In this case, the data that 

comes in real-time such as Internet-of-Things (IoT) sensor data and real-time Twitter feeds, 

could be loaded into the NewSQL database, and analysed on a real-time basis. The new 

generation analytics tools such as Spotfire and Tableau provide in-memory-computing 

features through which the entire data used for reporting could be cached into the reporting 

server, and cached data in memory would expedite information delivery and visualization. 

One of the subject matter experts has mentioned the following while discussing in-memory 

computing: 

“It (In-memory analysis) is something that we typically use it for visualization; when you want 

to write a query and load lots of data; it is supposed to be available for visualization very quickly 

– then you can use In-memory; …In-memory computing will be beneficial for sporting, casino 

kind of businesses where the update should be in seconds and decisions should be taken in 

seconds.” (A1, Senior Data Architect) 

Organizations store business metrics and KPI data in a data warehouse that is usually 

represented by a star-schema wherein the business-relevant KPI are stored in one or more fact 

tables (or a data store), and all reporting entities are stored as dimensions. For an organization, 

there would be an enterprise-wide data warehouse (EDW) that hosts business metrics relevant 

for all functional teams within the organization. In addition to this EDW, there would be many 

subject-area data warehouses one for each functional unit within the organization. These 

subject-area data warehouses are otherwise known as data-marts. The data-marts are built by 

extracting subject-specific data from EDW, or in some cases, an organization could build EDW 

from the underlying subject-specific data-marts. A1 has suggested,  

“The other important thing is for all data warehouses to have subject areas (on their own); it 

would help in semantic analysis; you put customer data into ‘customers’ store, sales into a 

separate store; you can put that into (create) order, inventory; likewise, you can create subject 

areas, and keep the data into different subject areas; anybody who want to do customer analysis 

can easily do it from the subject-area data-mart (data base); it is always better to persist in 

subject-matter data store and do further analysis; “ (A1, Senior Data Architect) 

Data transformation process is preceded with the pre-processing phase during which data 

from disparate systems get cleansed, validated and modelled to support transforming and 

loading the data into target database. The processes of extracting data from source systems, 

transforming data, and loading necessary business-relevant metrics into EDW are portrayed 

in Fig.2.  



Australasian Journal of Information Systems Balakrishnan, Das & Chattopadhyay 
2020, Vol 24, Research Article Implementing Data Strategy 

 15 

 

Fig.2 EDW built using ETL and ELT processes 

The findings from this study are in line with inferences on big data extraction methods 

identified by Paakkonen and Pakkala (2015). In addition, this study provides inferences on 

extraction options for both internal and external systems, and outlines the need for data-

staging to handle data from external systems.  

4.3 Data Lakes and Dataspaces 

Data lakes enable large-scale data storage at a relatively low cost. It is cheap and persistent 

storage that can support huge volumes of data (T. Davenport & Verma, 2018). The data 

stewards have to identify all possible sources from where the data has to be brought into the 

enterprise data lake. Once the causes are identified, the architects have to identify the right 

means to get the incremental and full dump of data from the source, and these mechanisms 

would include techniques such as writing a crawler job, developing a program to call API 

function to get data, configuring an ELT tool to extract data, and using vendor (source) 

provided adapters or scripts. The architects have to define an integrated semantic layer as 

mentioned by A2,  

“to give a specific example – assume you are getting social media comments on your product, 

service quality and product ideation; these are three different things – or you could call it 

‘subject areas or domains’. When you bring it over (the data over to staging), you put it (the 

extracted data) on a staging (temporary) data store; a semantic layer integrating all these three 

domains have to be modelled – to be precise, integrating your products through product-id and 

relating those products to the comments you get on service quality (related to those products), 

and user feedback on new features that could be added to the product through product ideation. 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Internal Systems 

Package applications (SAP, 
Oracle ERP, MS-CRM) 

Legacy (Mainframe, AS400) 

OIA (Other internal “home-
grown” Applications) 

• Business rules driven data 
extraction 

• Restructure data as needed 
• Calculate business metrics 
• Load into target EDW 

HR 

Marketing 

Sales 

Social 

media 

metrics 

• Business rules driven 
extraction 

• Load data into staging area 
 

Loading 

Area 

 

 

 

External Systems 

Online forums 

Social media 

E 

T 

 L 

EDW 

E 

L 

T 

Data Lake 

Enterprise 

Functional         

metrics 

R
e
s
tr

u
c
tu

re
 d

a
ta

 a
s
 n

e
e
d
e
d

 
C

a
lc

u
la

te
 b

u
s
in

e
s
s
 m

e
tr

ic
s
 

L
o
a
d
 i
n
to

 t
a
rg

e
t 
E

D
W

 

 

… 

Data Marts 



Australasian Journal of Information Systems Balakrishnan, Das & Chattopadhyay 
2020, Vol 24, Research Article Implementing Data Strategy 

 16 

Building this integrated model is a complex process as the relationships are not defined in a 

concrete way; you will have to do lot of trial and error things (modelling); one need to create 

ontologies, tag-clouds for often-used words, classifications on the type of comments used, etc – 

to arrive at this integrated (semantic) layer; data analytics and your predictive stuff (analytics) 

can be driven off of this (integrated layer); all these things constitute your data lake” (A2, Data 

and Solution Architect) 

Hadoop based Big Data environments provide flexibility to organizations when dealing with 

high volumes of data from different source systems. Since systems supporting big data are 

hosted on low-cost commodity hardware, organizations could choose to save data from less 

popular social forums, newspapers and online magazines into a big-data environment, 

without worrying about how it could be leveraged for customer value creation. In the words 

of an architect, 

“We always tell them (clients) – ‘forget about what you want to do with data; Have a data-lake 

in Hadoop; any data, you come across – make sure that you store in Hadoop (Big Data 

environment)’; We are not saying that high-volume data only should go to Hadoop; All other 

data – streaming, etc -more like a ‘Staging’ area; Dump everything into one store; we then 

figure out a way to use it, build analysis and store the inferences;” (A2, Data and Solution 

Architect) 

All the subject matter experts handling data management solutions feel that data lakes could 

be used to archive all types of data that may or may not be used later. This would result in a 

data store that is humongous in nature, and hence retrieving data for further processing might 

be time-consuming and cumbersome in nature. Dataspaces would help overcome these issues. 

Dataspaces are not part of a data integration layer, and they provide base functionality over 

all data sources, regardless of how integrated they are (Franklin, Halevy, & Maier, 2005). 

Dataspace comprises of data models with a collection of relationships between data 

repositories. One of the basic dataspace services is cataloguing data elements from all 

participating data sources (Franklin et al., 2005). Dataspace platform should provide means 

for searching and querying the metadata. The metadata forms the basis for all retrieval 

operations, and the metadata in dataspace could be configured at various levels such as time-

bound data, source-specific data, and certainty of data. Dataspace is considered as a hybrid of 

a database management system storing data from variety of sources, a search engine, a data-

sharing system and an information integration system (Mirza, Chen, & Chen, 2010).  

In the words of a team member from A2, 

“Any data that you get into data lake, one has to create a model or a catalogue supporting all 

sources of data, where the data is retrieved from, time-stamp associated with data, all the meta-

data associated with data like which user (if possible to identify) has created it, geo (geography) 

associated with the data, relationship between data elements from various sources – if any, etc. 

we could build more intelligence in the meta data by providing filtering elements such as the 

key tags that are being referred, the impact the data have – in terms of decision making, etc. all 

this could be part of the meta data ; when we retrieve data, usually you retrieve it through an 

app(application) layer; there, you query the meta-data to identify the right source we should 

query from, and initiate our retrieval; by having an effective cataloguing, you would be able to 

isolate which sources need to be queried, and also – how impactful the inference could be. If you 

don’t have catalogue information or namespace kind of setup, querying from a data lake might 



Australasian Journal of Information Systems Balakrishnan, Das & Chattopadhyay 
2020, Vol 24, Research Article Implementing Data Strategy 

 17 

be impossible as no one knows where to get it (your search results) from” (Team-A2, Data and 

Solution Architect) 

The inferences presented in this study that are related to data spaces is similar to the concepts 

suggested by Franklin et al. (2005) and Mirza et al. (2010). Additionally, this study emphasizes 

the need for having subject-area specific data stores along with catalogue and meta-data for 

each subject-area. Business Intelligence solutions built on this premise enable combining data 

across various subject-areas to arrive at valuable inferences that could drive business 

decisions.  

4.4 Customer Value Creation  

The organizations of the new digital era use data as a means by which they could create 

customer value proposition and leverage the same for monetization. A2 has opined, 

“Data monetization is an area where every organization wants to get into; Data is very valuable 

and not that easy to get; now organizations look at – we have data now, and how they can make 

money out of it? Something we are looking into – of late is ‘benchmarking’; benchmarking is 

very relevant for people in health-care industry; benchmarking gives an idea on where an 

organization stands with respect to various parameters that are standardized for the industry – 

and gives an indication of where the company should focus on; there are organizations in health-

care industry, who collect data (from various firms), these people collect non-sensitive data, 

gender data, demographic data – and start creating benchmarks with data; the data can be 

federated and third-party people can bank on the data for benchmarking; that is one way of 

monetizing; this probably would fall into ‘data-as-a-service’ model as well..” (A2, Data and 

Solution Architect) 

One of the architects has opined on the data-enabled value creation saying, 

“(by listening to data in SM and other channels), businesses get more informed in terms of what 

products have to be developed, and how they have to be positioned … if you provided targeted 

discounts, people tend to buy more; ..people tend to buy more as there is a discount which in 

turn increases sales. if you look at a data warehouse, it helps – and gives directions to salespeople 

on what to do (which product to develop), when and where (Positioning in different markets)?” 

(A1, Senior Data Architect) 

Firms have deployed self-service business intelligence tools for doing explorative data 

analysis. The information visualization tools support a variety of analysis such as what-if 

analysis, scenario analysis, correlation analysis, and enable business users to understand business 

KPI in a better way. The new generation data explorative tools are more tuned for business 

use than for IT use, and engage customers more efficiently than the previous generation 

reporting tools. The inferences on value creation and value delivery are aptly summarized by 

one of the interviewees:  

“you provide value-add like self-service reports (to business users), real-time analysis, help 

customers in their product launch, product positioning, enhancing your target market reach - 

those are your monetization opportunities; internally, you use data-driven decisions to 

improvise your operational metrics; that is your cost saving opportunities; any data-driven 

initiatives (value propositions) help you on two fronts: open (enables) new avenues for revenue 

(additional revenue generation), and cost-saving opportunities by process optimization; and by 

doing your data management the right way – you get to benefit from both ends (revenue 

generation and cost cutting);” (A3, Data Architect) 



Australasian Journal of Information Systems Balakrishnan, Das & Chattopadhyay 
2020, Vol 24, Research Article Implementing Data Strategy 

 18 

The findings from this study resonate with the ideas of using data-driven decisions for process 

optimization and value creation that are presented in various other studies. Furthermore, this 

study highlights options for data-enabled new value creation and revenue generation 

opportunities such as building data-as-a-service model.  

4.5 Other Critical Considerations: Data Quality, Data Security and Business 
Continuity 

The organization’s ability to create a data-driven value proposition is severely impacted by 

low-quality data. Organizations strive hard to strengthen the quality of data used for building 

EDW. A slang expression ‘Garbage in, Garbage out (GIGO)’ (Techopedia, 2011) refers to 

inaccurate results due to erroneous input data. Data governance mechanisms make sure that 

the data entered into both transactional systems and reporting systems are valid, valuable, 

safe and could be used effectively for decision making. One of the subject matter experts has 

to say this, when queried on the aspects of data governance: 

“implementing master data management, cleansing rules for data, business rules for data 

validation, implementing maker-checker for enforcing transactional integrity, data backups and 

archival based on business rules - at stipulated time periods – all are part of data governance 

mechanisms; an organization has to look at these things holistically – to set up their governance 

mechanisms for ensuring ‘clean data’;” (A3, Data Architect) 

Organizations have to comply with stringent data security mandates, and one of the rules 

imposed by Sarbanes-Oxley (SOX) considers encryption as a critical security control, and one 

of the best practices, for both data-at-rest, and data-in-transit (McAfee, 2019). This implies that 

all types of data residing in enterprise systems within the organization, have to be encrypted. 

Data encryption happens at the database level or the hardware level. Organizations could 

choose to encrypt entire databases such as MongoDB or Oracle hosting transactional 

applications. The encryption at the database level is preferred over incorporating security at 

the code level because database-level encryption provides uniformity across all functional 

elements of the application. Data centres hosting servers running enterprise applications could 

be protected by the use of firewall software and access control mechanisms. For encrypting 

data-in-transit, firms could use a secured protocol such as https instead of http. For data-

transfers from the source system to the target system, firms could use secured-ftp instead of 

regular FTP.  

Organizations implement data replication to support recovery from data failures and business 

continuity. Data replication happens through two modes such as active-replication and passive-

replication. The databases are replicated on a real-time basis during active-replication, and 

during passive-replication, databases are replicated when they are offline. Organizations 

usually keep primary server hosting data and a secondary, failover server in case of system 

failures. In the case of system failures or data loss, the fail-over server takes over automatically. 

Organizations define business rules for configuring data replication frequencies between 

primary and fail-over sites. For all critical transactional systems, replication happens 

frequently, or on a real-time basis. For all non-critical transactions, replication could occur 

during off-hours. The non-critical data such as facts and dimensions in an EDW are replicated 

less-frequently, usually once a day. In case of data loss, this incremental data in EDW could be 

built from the data in transactional systems. The objective of replication is to keep the data loss 

at a very minimal level.  



Australasian Journal of Information Systems Balakrishnan, Das & Chattopadhyay 
2020, Vol 24, Research Article Implementing Data Strategy 

 19 

5 Discussion  

This study on data architecture related to using diverse data management solutions has 

inferences on using data-access concepts such as data virtualization, implementing alternate 

options for real-time analytics, and defining general guidelines for building a robust data 

architecture for enterprise data lake and EDW. These inferences are discussed at length in 

subsequent sections. The last section discusses the implications for both practitioners and 

academicians and the limitations of this study. 

5.1 Options available for Data Extraction 

While extracting data from source systems, organizations could choose either ETL or ELT as 

an option to transform and load data into target systems. For both ETL and ELT data 

extraction, organizations create necessary scripts or use commercial-off-the-shelf tools to 

complete the data extraction and load jobs. Creating these jobs is time-consuming, and one 

needs to go through rigorous testing to cover all possible data scenarios. One of the options 

available for an organization to extract data from source systems instead of setting up ETL or 

ELT jobs is to build data virtualization on top of underlying databases. Data virtualization 

integrates data from disparate source systems without replicating the data, to create a single 

virtual data layer that delivers unified data to multiple applications and end-users (Patrizio, 

2019). When answering a query on specific scenarios for using data virtualization over ETL or 

ELT, A1 has opined, 

“Data virtualization is a case where you want to do ad-hoc-queries (for reporting); when your 

business wants to do adhoc-querying, and you want to run something which is not already 

existing. They (business users) can look at the data, define some queries on it, and explore the 

data further; so, data virtualization helps with ad-hoc queries which are not existing already; 

and it adds value without adding a long ETL process…. I would rather say – use data 

virtualization for all PoCs (proof of concepts) that you want to run; find the value of it; and for 

better performance (if it needs to be repeated), you make it into ETL or ELT (for further 

extraction, processing, and analysis);” 

The suggestion is to use the data virtualization option to query directly from underlying 

systems when working on proof-of-concept or demo-version of applications. Organizations 

could choose to go with ETL or ELT jobs when they want to implement the proof-of-concept 

version as a full-fledged application. Data virtualization helps a firm to increase productivity 

when building EDW, brings down the cost as the data does not get replicated through ETL, 

and lessens the data governance needs (Patrizio, 2019).  

5.2 Real-time analytics 

Organizations could use streaming analytics or real-time analytics to query continuous data 

streams and detect conditions instantly, within a small period from the time the data is 

received (Perera, 2018). Real-time analysis is the right choice in scenarios where the data 

insights are more valuable only when the data is processed shortly after it is received, and the 

value diminishes fast with time (Perera, 2018). Some of the typical use cases for real-time 

analytics are applications such as providing real-time personalization options, detecting 

anomalies and frauds in real-time, providing real-time emergency health-care services, 

implementing smart device applications, and building algorithmic trading applications 

(Lawton, 2019). Options such as an e-commerce portal recommending a product to buy, or 

offering a discount based on the transactions one has made in the last few hours, are made 



Australasian Journal of Information Systems Balakrishnan, Das & Chattopadhyay 
2020, Vol 24, Research Article Implementing Data Strategy 

 20 

possible through real-time analytics. Organizations could use technologies such as NewSQL to 

drive real-time analytics. One of the architects has commented on the popularity and relevance 

of NewSQL systems, 

“I have my apprehensions about NewSQL; if you look at the way it has come up, it has been 

there for almost 8 years now; 2011 is when it came first; – not much of adoption happened in 

the industry; that is what I have seen; even though it (NewSQL) happened, it happened in a 

small way; In places like ours (e-commerce portals), it is difficult to adopt (a tool or technology) 

which is not previously adopted or proven (by others); … the maturity of NewSQL is still a 

question; (A1, Senior Data Architect) 

The architects feel that NewSQL systems handle real-time analysis at a higher cost compared 

to other streaming analytics tools that are open-source or offered at a significantly lower cost. 

One of the subject matter experts used for this study, feel that tools such as Kafka, provided 

by Apache Foundation, could listen to messages streaming from SM, and synchronously do 

analysis such as trends and sentiments. Information visualization tools could directly connect 

to streaming analytics tools and plot real-time reports on inbound comments. The messages 

coming through Kafka are cached in Kafka for a duration that could be configured. Developers 

could write scripts to move the streaming messages coming through Kafka, and persist them 

in a data-store if needed (Yang, 2016). A1 has opined, 

“you will have a lot more channels Kafka listens at (to); you could push messages to various 

channels – the product could be one, and so on. … If they (firms) want to do that (real-time 

analysis), we will have (a tool like) Tableau listening to Kafka; then, you can do real-time 

analytics; So, Kafka would be your data source for Tableau; It is like messaging queue; the data 

gets persisted for a period of time (in Kafka); then, you could refresh that also; if you do not want 

to keep that data (in Kafka) for a long time, you could set the configuration to ‘delete the 

messages’, or you could move the messages (to your data warehouse or transactional systems) 

also” (A1, Senior Data Architect) 

The big data and data management reference architecture depicted in Fig.1 covers all the 

components of EDW and data lake. The streaming-analytics part in Fig.1 is explained in Fig.3, 

which contains high-level architecture and process flow associated with Kafka enabled 

stream-processing.  

 

Fig.3 Stream Processing using Kafka – High-level Architecture and Process Flow  

 

                    

 

 

 

 

 

 

 

Persist data to 

disk(optional) 

Data request API 
Twitter 

 

 

 

 

 

 

Facebook 

Python 

Tableau 
Batch data 

Streaming data 

Data sources 

Extraction 

Python 

Script 

 

Kafka Storage & 

Messaging 

Storage 

Kafka Connect 

HDFS Hive Store 

Analysis & Report 



Australasian Journal of Information Systems Balakrishnan, Das & Chattopadhyay 
2020, Vol 24, Research Article Implementing Data Strategy 

 21 

5.3 Guidelines for building a robust data architecture 

Organizations have predicaments on the validity of the data management solutions discussed 

in this paper, and the chances of these solutions becoming obsolete. The solutions mentioned 

in this study such as RDBMS, NoSQL, and NewSQL or real-time analysis tools, complement 

each other and will continue to co-exist in an enterprise. The choice of using an appropriate 

data management solution is dependent on the type of data that needs to be managed, the way 

the data has to be integrated and transformed, and the method the data has to be analysed. In 

the interest of time, effort and cost, organizations tend to standardize the usage of data 

management solutions. The standardization of solutions is considered a good initiative, but it 

may not be the right choice. As one could notice, data keeps on evolving. As the data evolves, 

the data management solutions have to evolve as well. With that, the adage of one-size-fits-all 

would not work anymore, when discussing data management solutions. 

It is suggested that organizations should extract all types of data available on external systems, 

load them into a low-cost Big Data setup, regardless of whether it could be leveraged or not. 

When organizations arrive at a need to use the saved data from external sources, they could 

implement processes to transform the staged data into meaningful business metrics and persist 

into an EDW. Organizations have to be mindful of data quality, as the data quality determines 

the effectiveness of data analytics. Organizations have to implement a business-rule driven 

data cleansing, data profiling, and data validation procedure to make sure that the data is 

analysis-ready before it is transformed and loaded into EDW. For transactional systems, to 

strengthen the data quality, organizations could implement maker-checker, a widely 

implemented, successful process implemented in the financial services industry (Commercial 

Bank, 2018). This process mandates a transaction entered by a maker to be verified and 

approved by a checker, to make sure that the data is valid and useable.  

The data extracted from external source systems such as SM, forums, blogs, and newsletters 

get inserted into data lakes, and data governance enables the data to be tagged. In other words, 

data governance mechanisms warrant architects to catalogue the data that gets into data lakes 

by building a meta-data repository. The data has to be governed in terms of ownership, 

accountability, sharing, and usage, to avoid accumulating unwanted data in the form of data 

swamps (Ghosh, 2019).  

The organization’s success in building a robust data management system depends on 

implementing solutions such as master data management (MDM) solution, implementing data 

privacy, and maintaining data portability to other environments when needed (Addagada, 

2019). Master data is defined as an organization’s core data, containing the necessary 

information that is critical to run the business, and the data is relatively stable in nature (Foote, 

2019a). Some of the master data domains include customers, products, and locations. In the 

words of A1, 

“the centralized master data – used by all systems in the enterprise; generally, one (the firm) 

stores only one instance of master data – that gets replicated or federated to other systems; the 

data in master data should reflect ‘the single version of your most important entities in the 

enterprise – like customer and part/product details’. A well-defined master data of customers 

and products can enable the firm to get a unified view of customer-wise product sales across 

various channels – in an effective way in very short span of time; MDM increases data analytics 

efficiency” (A1, Senior Data Architect) 



Australasian Journal of Information Systems Balakrishnan, Das & Chattopadhyay 
2020, Vol 24, Research Article Implementing Data Strategy 

 22 

As the type of data evolves, data modelling has become even more important (Foote, 2019b). 

The benefits associated with modelling NoSQL data include improved data quality, lot more 

identifiable information, and enhanced business intelligence. Architects could start with 

simple models, and make the models agile, keeping them evolve as new data sources and 

types of data get added into data lakes (Foote, 2019b). In the words of A2 and his team, 

“Data models are very important regardless of the data you use; all of us are familiar with OLTP 

and OLAP related data models (normalization and relational databases); now, with all the 

evolving database types, we need to constantly add meta-data repositories and models to define 

and hold that data; this (modelling NoSQL data) will play a critical role when you do analysis; 

to give an example, if you want to monitor twitter feeds for a given trending topic, the data 

could be stored in a relational model that store, topic-id, topic-name, comment or tweet, and 

some meta-data on the person posted the tweet including the date-time stamp; once, you have it 

in this form, assuming that it is saved in BigData Hadoop cluster, querying this data using 

tools like Hive would be much easier; you could query, do your aggregates and analytics on top 

of this; as and when new types and sources evolve, one should keep adding flexible data models 

in the form of your meta-data, and also tables to host your data.. this is very important to have 

a control over what comes into your system” (A2, Data and Solution Architect) 

Good data governance mechanisms, coupled with data management best practices contribute 

to the successful implementation of data-driven value creation. The data governance issues 

related to data security, data privacy, and data quality could be managed effectively through 

data-stewardship, the right set of processes for extraction, transformation and loading, and 

the right tools and technologies. In this section, as mentioned earlier, the aspects to related to 

technology governance are included, and mechanisms related to people and process aspects 

of governance are not dealt with. 

5.4 Implications and Limitations of this study 

This study has implications for both practitioners and academicians. For the practitioners, this 

study provides an overview of building an EDW by encompassing data from disparate sources 

and external systems. This study provides insights to managers on the choices of data 

management solutions available, and the ideal use cases for the same. With this information, 

managers would be able to choose the right data management tool for a specific business 

purpose. This study provides inferences to academicians on the methods used for creating 

value propositions from data available in internal systems and external systems. Academicians 

would be able to understand the components involved in building an EDW, and the use of 

various data management solutions. This study provides extensive coverage of the processes 

needed for creating an EDW. The subject matter experts used for the study have offered 

insights on generic processes used in data analysis. However, the insights offered on data 

architecture are oriented more towards practitioners. This study proposes research to be 

conducted in academic settings, to understand the concepts and theory behind implementing 

data management solutions. This suggested research would be able to bring perspective on 

the nature of transactions, properties of data conducive for certain analysis, extraction and 

transformation methods, and options available for analytics. 

6 Conclusion 

Many of the findings from this study are already available as part of technical white papers, 

research articles, and conference proceedings. However, in this paper, the researchers have 



Australasian Journal of Information Systems Balakrishnan, Das & Chattopadhyay 
2020, Vol 24, Research Article Implementing Data Strategy 

 23 

attempted to collate all those findings to arrive at a reference architecture for all types of data 

management solutions. The critical contribution from this paper is to provide a consolidated 

reference material that would provide input to practitioners on: a) using the right technology 

and methods for data analysis, b) implementing the right use cases for various data 

management technologies, c) recommending a reference architecture for managing various 

data management solutions and design considerations for implementing the solution, d) 

providing a literature review and overview of all types of data management solutions, their 

pros and cons, and how they are part of the data management eco-system, and e) providing a 

consolidated view on various technologies used for data management such as data lakes, 

dataspaces, virtualization, and real-time monitoring tools.  

This study has approached the journey of an organization towards data-enabled value creation 

through four levels of data processing, such as data extraction, data transformation, data 

exploration or information visualization, and value creation. The study has critical 

observations with respect to using RDBMS, NoSQL, and in-memory data management 

solutions. The query on organizations supporting a variety of data management solutions, and 

the recommended data architecture for supporting the same, is aptly answered by one of the 

subject matter experts. The entire process of building an EDW that enables further data 

analysis is succinctly quoted here: 

“The traditional data in various transactional systems, you store in relational databases; you 

have data from social media, and whatever other external sources – you want to bring it over to 

a NoSQL kind of environment which is hosted on a Big Data environment, or, you could use 

any other form of Big Data systems (systems supporting big data - provided by various third 

parties), and then you use some level of processing and transformation to take all metrics that 

are relevant for your business needs, and then bring it (those metrics and KPI) into a structured 

kind of data warehouse, and you drive your information visualization (reporting and analysis) 

on top of that” (A1, Senior Data Architect) 

Organizations could create value by providing real-time analysis on data retrieved from SM, 

and an architect, when discussing the technology used for real-time analytics, has mentioned, 

“Any data that does not have to be persisted, and you need real-time analysis, then go for 

NewSQL; here, what you do is – you stream the data into NewSQL in-memory database, enable 

all your analysis on the same; … NewSQL is not proven, and a bit expensive, and you have 

other options available as well – for doing real-time analysis (Like Kafka)” (A3, Data Architect) 

Organizations manage the data-enabled value creation through governance mechanisms 

implemented in a holistic way to address challenges associated with data security, data 

privacy, data loss and recovery, and business continuity. The design considerations should 

include elements such as encryption at the database and hardware-level for data-at-rest, 

security through firewall at the data-centre level, data transfer over secured protocols for data-

in-transit, access controls at various levels, scheduled back-ups for data, data storage through 

both primary and fail-over instances, and data replication through active and passive modes. 

Organizations have to implement robust information-governance mechanisms across data 

handling processes to achieve higher level of quality. One of the architects has commented on 

the aspect of data governance, 

“’ clean and good data’ enables good metrics, and good metrics enable great inferences – and 

hence, these things (data governance mechanisms and best practices) are very critical for 



Australasian Journal of Information Systems Balakrishnan, Das & Chattopadhyay 
2020, Vol 24, Research Article Implementing Data Strategy 

 24 

implementing quality data management solutions that would drive your business;” (A3, Data 

Architect) 

In summary, data is considered as the gold of the 21st century (Aviza, 2017), and using it the 

right way would yield economic returns at an epic scale. Organizations could harvest the value 

embedded in data by implementing the right data management technologies, coupled with 

appropriate information governance mechanisms.  

Acknowledgements 

We thank our interviewees Rajendran Subramanian, Narayanan Kandanchatha, 

Mukundakumar KG, Brijesh Madhavan, Shiju Jalaludin, Murali Avanamuthu, and their team 

members for sharing their pearls of wisdom during the course of this research. We are 

immensely grateful to Dr. S. Varadhaganapathy for helping us in framing our research queries. 

We extend our sincere thanks to Latha Venkitachalam and Baiju Ambadan for providing 

comments that have greatly improved the manuscript. Please note that any errors in the 

manuscript are our own, and the errors should not tarnish the reputations of these esteemed 

persons.  

References 

Addagada, T. (2019). Customer Data Protection: Deriving Value and Ownership. Retrieved May 

24, 2019, from https://www.dataversity.net/customer-data-protection-deriving-value-

and-ownership/ 

Anderson, C., & Li, M. (2017). Five building blocks of a data-driven culture. Retrieved May 27, 

2019, from https://techcrunch.com/2017/06/23/five-building-blocks-of-a-data-driven-

culture/ 

Aviza, E. (2017). Data is the Gold of the 21st Century. Retrieved May 20, 2019, from 

https://www.cloudbakers.com/blog/data-is-the-gold-of-the-21st-century 

Bowen, R., & Smith, A. R. (2014). Developing an enterprisewide data strategy. Healthcare 

Financial Management, 68(4). 

Braganza, A. (2004). Rethinking the data – information – knowledge hierarchy : towards a case-

based model. International Journal of Information Management, 24, 347–356. 

https://doi.org/10.1016/j.ijinfomgt.2004.04.007 

Commercial Bank, C. G. (2018). Why Dual Approval Matters. Retrieved May 6, 2019, from 

https://businessaccess.citibank.citigroup.com/basprod/citiiwt/images/Why_Dual_Appr

oval_Matters.pdf 

Data Integration. (2018). Retrieved May 20, 2019, from https://www.dataintegration.info/data-

integration 

Davenport, T. H. (2014). Big Data at work: Dispelling the myths and uncovering the opportunities. 

Harvard Business Review Press. 

Davenport, T. H., Harris, J. G., Long, D. W. De, & Jacobson, A. L. (2001). Data to Knowledge 

to Results: Building an Analytic Capability. California Management Review, 43(2). 

Davenport, T., & Verma, A. (2018). It’s time to modernize your big data management techniques. 

Retrieved May 18, 2019, from 



Australasian Journal of Information Systems Balakrishnan, Das & Chattopadhyay 
2020, Vol 24, Research Article Implementing Data Strategy 

 25 

https://www2.deloitte.com/us/en/insights/topics/analytics/data-management-

techniques-approaches-tools.html 

Doyle, M. (2017). The Importance of a “Data Integration First” Strategy. Retrieved May 15, 2019, 

from https://www.dqglobal.com/2017/07/25/importance-data-integration/ 

Foote, K. D. (2019a). A Brief History of Master Data. Retrieved May 12, 2019, from 

https://www.dataversity.net/a-brief-history-of-master-data/ 

Foote, K. D. (2019b). Data Modeling in an Agile World. Retrieved May 15, 2019, from 

https://www.dataversity.net/data-modeling-in-an-agile-world/ 

Franklin, M., Halevy, A., & Maier, D. (2005). From databases to dataspaces: A new abstraction 

for information management. SIGMOD Record, 34(4), 27–33. 

https://doi.org/10.1145/1107499.1107502 

Ghosh, P. (2019). Data Governance and Data Quality Use Cases. Retrieved May 12, 2019, from 

https://www.dataversity.net/data-governance-and-data-quality-use-cases/ 

Griffin, J. (2005). Data governance: a strategy for success. DM Review, 15(8), 15, 70. 

Grolinger, K., Higashino, W. A., Tiwari, A., & Capretz, M. A. M. (2013). Data management in 

cloud environments : NoSQL and NewSQL data stores. Journal of Cloud Computing. 

https://doi.org/https://doi.org/10.1186/2192-113X-2-22 

Hendler, J. (2009). Web 3.0 Emerging. IEEE Computer Society, 42(January), 111–113. 

Huber, G. P. ., & Power, D. J. . (1985). Retrospective Reports of Strategic-Level Managers : 

Guidelines for Increasing Their Accuracy. Strategic Management Journal, 6(2), 171–180. 

Kambatla, K., Kollias, G., Kumar, V., & Grama, A. (2014). Trends in big data analytics. Journal 

of Parallel and Distributed Computing, 74(7), 2561–2573. 

https://doi.org/10.1016/j.jpdc.2014.01.003 

Kaur, K., & Sachdeva, M. (2017). Performance Evaluation of NewSQL Databases. In 

International Conference on Inventive Systems and Control (pp. 1–5). 

https://doi.org/10.1109/ICISC.2017.8068585 

Khatri, V., & Brown, C. V. (2010). Designing Data Governance. Communications of the ACM, 

53(1). https://doi.org/10.1145/1629175.1629210 

Kooper, M., Maes, R., & Lindgreen, R. E. (2011). Information Governance as a Holistic 

Approach to Managing and Leveraging Information Prepared for IBM Corporation. 

International Journal of Information Management, 31. 

Korhonen, J. J., Melleri, I., Hiekkanen, K., & Helenius, M. (2013). Designing Data Governance 

Structure : An Organizational Perspective. GSTF Journal On Computing, 2(4), 11–17. 

https://doi.org/10.5176/2251-3043 

Lawton, G. (2019). 7 enterprise use cases for real-time streaming analytics. Retrieved May 20, 2019, 

from https://searchbusinessanalytics.techtarget.com/feature/7-enterprise-use-cases-for-

real-time-streaming-analytics 

Lim, C., Kim, K., Kim, M., Heo, J., Kim, K., & Maglio, P. P. (2018). From data to value : A nine-

factor framework for data-based value creation in information-intensive services. 



Australasian Journal of Information Systems Balakrishnan, Das & Chattopadhyay 
2020, Vol 24, Research Article Implementing Data Strategy 

 26 

International Journal of Information Management, 39(January 2017), 121–135. 

https://doi.org/10.1016/j.ijinfomgt.2017.12.007 

Link, S., & Prade, H. (2019). Relational database schema design for uncertain data Relational 

Database Schema Design for Uncertain Data $. Information Systems. 

https://doi.org/10.1016/j.is.2019.04.003 

Lourenço, J. R., Cabral, B., Carreiro, P., Vieira, M., & Bernardino, J. (2015). Choosing the right 

NoSQL database for the job: a quality attribute evaluation. Journal of Big Data, 2(1), 1–26. 

https://doi.org/10.1186/s40537-015-0025-0 

Manyika, J., Chui, M., Brown, B., Bughin, J., Dobbs, R., Roxburgh, C., & Byers, A. H. (2011). 

Big data: The next frontier for innovation , competition , and productivity. McKinsey 

Global Institute, (May). 

Marco, D. (2006). Understanding data governance and stewardship, Part 1. DM Review, 16(9), 

28. 

Mazzei, M. J., & Noble, D. (2017). Big data dreams : A framework for corporate strategy. 

Business Horizons, (60), 405–414. 

McAfee. (2019). Overview of Serbanes-Oxley. Retrieved May 20, 2019, from 

https://www.skyhighnetworks.com/cloud-compliance/sarbanes-oxley-encryption-

compliance-requirements/ 

Mirza, H. T., Chen, L., & Chen, G. (2010). Practicability of dataspace systems. International 

Journal of Digital Content Technology and Its Applications, 4(3), 233–243. 

https://doi.org/10.4156/jdcta.vol4.issue3.23 

Mohan, C. (2013). History Repeats Itself : Sensible and NonsenSQL Aspects of the NoSQL 

Hoopla. IBM Alamaden Research Center, 11–16. 

Newman, D., & Logan, D. (2009). Governance Is an Essential Building Block for Enterprise 

Information Management. Gartner Research, (May 2006). 

Pääkkönen, P., & Pakkala, D. (2015). Big Data Research Reference Architecture and 

Classification of Technologies , Products and Services for Big Data Systems. Big Data 

Research, 2(4), 166–186. https://doi.org/10.1016/j.bdr.2015.01.001 

Patrizio, A. (2019). What is Data Virtualization? Retrieved May 20, 2019, from 

https://www.datamation.com/big-data/what-is-data-virtualization.html 

Pavlo, A., & Aslett, M. (2016). What is really new with NewSQL ? SIGMOD Record, 45(2), 45–

55. 

Perera, S. (2018). A Gentle Introduction to Stream Processing. Retrieved May 20, 2019, from 

https://medium.com/stream-processing/what-is-stream-processing-1eadfca11b97 

Pokorny, J. (2013). NoSQL databases: A step to database scalability in web environment. 

International Journal of Web Information Systems, 9(1), 69–82. 

https://doi.org/10.1108/17440081311316398 

Rocha, L., Vale, F., Cirilo, E., Barbosa, D., & Mourao, F. (2015). A Framework for Migrating 

Relational Datasets to NoSQL ∗. Proceedia Computer Science, 51, 2593–2602. 

https://doi.org/10.1016/j.procs.2015.05.367 



Australasian Journal of Information Systems Balakrishnan, Das & Chattopadhyay 
2020, Vol 24, Research Article Implementing Data Strategy 

 27 

Salido, J. (2010). Data Governance for Privacy, Confidentiality and Compliance: A Holistic 

Approach. ISACA Journal, 6, 1–7. 

Sareen, P., & Kumar, P. (2015). NoSQL Database and its comparison with SQL Database. 

International Journal of Computer Science & Communication Networks, 5(5), 293–298. 

Simsek, G. (2019). What is new about NewSQL? Retrieved June 7, 2019, from 

https://softwareengineeringdaily.com/2019/02/24/what-is-new-about-newsql/ 

Spivack, N. (2011). Web 3.0: The Third Generation Web is Coming. Retrieved May 18, 2019, from 

https://lifeboat.com/ex/web.3.0 

Stantic, B., & Pokorny, J. (2014). Opportunities in Big Data Management and Processing. 

Databases and Information Systems VIII. https://doi.org/10.3233/978-1-61499-458-9-15 

Techopedia. (2011). Garbage In, Garbage Out (GIGO). Retrieved May 15, 2019, from 

https://www.techopedia.com/definition/3801/garbage-in-garbage-out-gigo 

Weber, K., Otto, B., & Osterle, H. (2009). One Size Does Not Fit All — A Contingency Approach 

to Data Governance. ACM Journal of Information Quality, 1(1). 

https://doi.org/10.1145/1515693.1515696.http 

Weill, P., & Ross, J. (2005). A matrixed approach to designing IT governance. MIT Sloan 

Management Review, 46(2), 26–34. 

Yang, F. (2016). Building a Streaming Analytics Stack with Apache Kafka and Druid. Retrieved May 

20, 2019, from https://www.confluent.io/blog/building-a-streaming-analytics-stack-

with-apache-kafka-and-druid/ 

Zack, M. H. (1999). Managing Codified Knowledge. Sloan Management Review, 40(4), 45–58. 

 

 

Copyright: © 2020 Balakrishnan, Das & Chattopadhyay. This is an open-access article 

distributed under the terms of the Creative Commons Attribution-NonCommercial 3.0 

Australia License, which permits non-commercial use, distribution, and reproduction in any 

medium, provided the original author and AJIS are credited. 

doi: https://doi.org/10.3127/ajis.v24i0.2541 

 

http://creativecommons.org/licenses/by-nc/3.0/au/
http://creativecommons.org/licenses/by-nc/3.0/au/