Toward big data and analytics governance: redefining structural governance mechanisms

Abstract

Big Data and Analytics (BDA) enable innovative business models and, simultaneously, increase existing business processes' efficiency and effectiveness. Although BDA's potential is widely recognized, companies face a variety of challenges when adopting BDA and endeavoring to generate business value. Researchers and practitioners emphasize the need for effective governance to delineate data and analytics' roles and responsibilities. Existing studies focus either on data or on analytics governance, even though both approaches are closely interlinked and depend on each other. Our study aims to integrate these two distinct research perspectives into a unified view on structural mechanisms for BDA. Using design science research, we iteratively develop data and analytics roles, clarify their responsibilities and provide guidelines for their organizational assignment. Our study contributes to advancing research on data and analytics governance and supports practitioners managing BDA.

Full text

Toward big data and analytics governance:
redefining structural governance mechanisms
Martin Fadler
Faculty of Business and Economics
(HEC), University of Lausanne
martin.fadler@unil.ch
Christine Legner
Faculty of Business and Economics
(HEC), University of Lausanne
christine.legner@unil.ch
Abstract
Big Data and Analytics (BDA) enable innovative
business models and, simultaneously, increase existing
business processes’ efficiency and effectiveness.
Although BDA’s potential is widely recognized,
companies face a variety of challenges when adopting
BDA and endeavoring to generate business value.
Researchers and practitioners emphasize the need for
effective governance to delineate data and analytics’
roles and responsibilities. Existing studies focus either
on data or on analytics governance, even though both
approaches are closely interlinked and depend on each
other. Our study aims to integrate these two distinct
research perspectives into a unified view on structural
mechanisms for BDA. Using design science research,
we iteratively develop data and analytics roles, clarify
their responsibilities and provide guidelines for their
organizational assignment. Our study contributes to
advancing research on data and analytics governance
and supports practitioners managing BDA.
1. Introduction
Companies with a clear strategy to monetize their
data can strengthen their competitive advantage
through new business models, data-driven insights, and
improved business processes [1]. Although the
potentials of leveraging Big Data and Analytics (BDA)
are well known, enterprises face various challenges
with BDA adoption and value generation. Among
these challenges are processing large data volumes,
ensuring the data quality, protecting privacy-related
data, as well as governing data [2]. Researchers point
out that “without appropriate organizational structures
and governance frameworks in place, it is impossible
to collect and analyze data across an enterprise and
deliver insights to where they are most needed” [3,
p.417]. Despite its practical relevance, data governance
is a challenging topic that has received much less
attention than IT governance [4]. Existing data
governance concepts focus on master data and data
quality [5]–[7], but do not consider new requirements
that have emerged with BDA [8]. In the past, analytics
governance was associated with business intelligence
and data warehousing [9], [10], with research only
recently opening up to advanced analytics [11].
However, this research stream has not as yet integrated
findings from data governance studies. In the context
of BDA, an overarching understanding is required of
data and analytics governance. This comprises a better
understanding of structural governance mechanisms
such as roles and responsibilities [13] which are
fundamental to any governance design and considered
a catalyst of BDA value creation [6]. The following is
therefore our research question: How do enterprises
(re-)define their roles and responsibilities in the Big
Data and analytics (BDA) context?
Our overarching research objective is to develop a
reference model for enterprise-wide data and analytics
governance. In this paper, we emphasize the structural
governance mechanisms, which we will complement
with procedural and relational mechanisms in further
research. Following the guidelines for design science
research [12], we developed a role model through
multiple iterations involving data and analytics experts
from large corporations. Learning from enterprises
with extensive experience in data governance and
adopting BDA technologies, we derived unique
insights into the inner workings of designing data and
analytics governance in large enterprises. We also
identified common patterns regarding the allocation of
roles and responsibilities.
The remainder of this paper is structured as
follows. First, we review literature on IT, data, and
analytics governance to clarify the commonalities and
differences, as well as the research gap that our study
examines. We then motivate and describe our design
science research approach and outline the research
process. Thereafter we present our framework for data
and analytics’ roles and responsibilities in detail.
Finally, we discuss our results, and provide an outlook
on future research.
Proceedings of the 54th Hawaii International Conference on System Sciences | 2021
Page 5696
URI: https://hdl.handle.net/10125/71311
978-0-9981331-4-0
(CC BY-NC-ND 4.0)

2. Background
Although often considered a practitioner topic,
governance has gained increasing interest from IS
researchers in recent years. Generally speaking, IT
governance is associated with “[…] the distribution of
IT decision-making rights and responsibilities among
different stakeholders in the enterprise, and defines the
procedures and mechanisms for making and
monitoring strategic IT decisions” [13, p.7]. An
increasing number of studies investigate the
phenomenon from different perspectives (see Table 1):
(1) the governance of IT artifacts (IT governance), (2)
the content of IT artifacts (data/information
governance), and (3), more recently, the analysis of IT
artifacts’ content (analytics governance).
In the following, we describe each form in detail
and discuss their commonalities and differences.
Table 1 Prior research on IT, Data, and Analytics Governance
IT
Data
Analytics
Governance forms
Differentiation of
structural, procedural,
and relational
mechanisms [10], [12]
IT decision domains and
organizational
archetypes [14]
Governance in specific
contexts: Eco-system/
platform governance
[15], [16], application
governance [17], and IT
consumerization [18]
Structural, procedural,
and relational
information governance
mechanisms [19]
Roles, decision areas,
and main activities of
data quality
management [5]
Data management’s
decision domains and
locus of accountability
[20]
Structural, procedural,
and relational analytics
governance mechanisms
[11]
Typical roles/process of
analytics competence
centers [21]
Governance in specific
contexts: accountability in
algorithmic decision-
making [22], AI
governance [23], and
BDA ownership types [8]
Contingency
Multiple contingencies
and their interaction
(reinforcing, conflicting,
and dominating) in IT
governance design [24]
Multiple contingencies
in data governance
design and data quality
management success [5]
Uncertainty and
business unit similarity
regarding a decision to
centralize/ decentralize
data management [25]
No studies found
2.1. IT governance (IT artifact)
Most of IS studies on governance have been
conducted with a focus on governing IT artifacts on the
firm level [4]. Previous research can be broadly
divided into two distinct streams with one focusing on
IT governance forms and the other on IT governance
contingency influences [26].
Studies on IT governance forms investigate the
organizational assignment of the decision-making
authority and the organizational structuring of IT
activities in order to increase the return on investment
[26, p.700]. Typical discussions include the advantages
and disadvantages of assigning IT decisions in a
central, decentral, or federated organization. While a
centralized organization emphasizes control over IT
standards and increases the opportunity for an
economy of scale, a decentralized allocation allows for
customizing IT solutions to meet the business needs
and enables flexibility. The federated IT governance
form allows companies to gain advantages from the
two extremes. Here, a central IT organization seeks
economies of scale and provides core IT services to the
entire enterprise, while business units retain their
flexibility and build their own solutions if the offered
services were to not address the business needs
sufficiently. In addition, researchers also examined the
IT governance architecture [13]. They suggested
differentiating between the structural (e.g. roles and
responsibilities), procedural (e.g. processes), and
relational (e.g. alignment) governance capabilities or
mechanisms [13], [27].
Studies on IT governance contingency influences
aim to understand which form is most suitable for
which company type. They analyze factors that affect
individual IT governance framework success [26,
p.703], such as the organizational structure, business
strategy, industry, and firm size. Researchers
investigate either single or multiple contingency
factors. The seminal study by Sambamurthy and Zmud
applies the theory of multiple contingencies to examine
the factors that influence decisions on the IT
governance form [24]. It reveals that the contingency
factors interact with one another along three scenarios:
reinforcing, conflicting, and dominating.
Although IT governance has been extensively
researched, innovative technologies and changes in
technology lead to new requirements arising. The
governance of platforms and eco-systems, for instance,
requires stakeholders' involvement beyond the firm’s
boundaries [15], [16]. Furthermore, application
governance becomes more important when companies
increasingly rely on Internet-based delivery models,
such as software-as-a-service (SaaS), which challenge
traditional governance assumptions [17]. Other
researchers emphasize that IT consumerization could
transform the fundamentals of IT governance [18].
2.2. Data governance (Content)
Compared to IT artifacts’ governance, the content
perspective has received much less research attention,
but BDA and the processing of massive amounts of
data have made it more important [4]. While IT
governance aims to manage IT assets in the sense of
hardware and software components that help support
the automation of well-defined tasks, data governance
aims to manage data (or information) assets as facts
having value or potential value that are documented
Page 5697

[20, p.148]. Consequently, the need for data
governance evolved with data’s importance for
enterprises. [28] describe how data management
evolved in three phases: data administration (since the
1980s), quality-oriented data management (since the
1990s) and extensions of strategic data management
(since the 2010s). Data governance issues were only
raised in the second phase when data quality became a
significant success factor in system integration projects
(e.g. data warehouse implementations). Similar to IT
governance, existing research can be broadly divided
into understanding data governance forms and
analyzing contingency influences. The first data
governance frameworks emerged when companies
started managing master data (a company’s core data
objects, e.g. its business partners or products), which
requires data quality management and coordination
across business units [29], [30]. [8] use the IT
governance framework’s general structure suggested
by [14] to derive five key decision domains (data
principles, data quality, metadata, data access, and data
lifecycle) for data governance and map them to the
locus of accountability (central to decentral). These
authors use an example to illustrate that data principles
should, for instance, be centrally defined, while data
quality should be managed in a decentralized way.
Other researchers focus on data governance
mechanisms. [19] conclude that many of the factors
leading organizations to adopt information governance
practices are equally relevant when governing physical
IT artifacts (p. 170). These authors therefore use the
same structure of [13], [27] to define structural,
procedural, and relational information governance
mechanisms. [5] define typical roles (executive
sponsor, data quality board, chief steward, business
data steward, and technical data steward), decision
areas, and the main activities of data quality
management. The business data steward is, for
example, a key role in data quality management and
responsible for detailing, from a business perspective,
corporate-wide data quality standards and policies for
his/her area of responsibility [5, p.11]. The authors
further analyze the influence of seven contingency
factors (e.g. performance strategy or organization
structure) on data governance design and data quality
management success. While this study has a narrow
focus on data quality management, a more recent study
examines the influences of similarities between
business units and uncertainty on a decision to
centralize or decentralize their data management [25].
One of their findings is that when business units are
very similar and uncertainty is low, their data
management should be centralized, which will, for
example, reduce the coordination effort.
Only a few studies have investigated data
governance in the BDA context. [31] argues that data
governance practices need to balance value creation
and risk exposure to gain a competitive advantage and
maximize the business value through BDA. This
balance requires companies to continuously evaluate
the business value of their data assets, while ensuring
that the risks are monitored and assessed. [32] argues
that information governance is required to facilitate
BDA capabilities. [8] endeavor to explicate the
fundamentals of a BDA governance model by defining
three data ownership types: data owner, data platform
owner, and data product owner. The latter study
stresses that new roles and responsibilities are required
to manage BDA on a firm level. While data owners
ensure controlled access and use of data at the data
source level, data product owners are accountable to
ensure that data products generate business value over
their lifecycle. Since data are a strategic asset, [33]
introduce the role of a chief data officer (CDO).
2.3. Analytics governance (Analysis of content)
More sophisticated forms of analytics involve
artificial intelligence and automated decision-making,
requiring new roles and responsibilities, but also
leading to new risks. Governance should therefore not
be limited to the content, but should also encompass its
analysis.
Researchers emphasize that, in addition to IT and
data governance, analytics governance mechanisms are
required to overcome challenges, such as the alignment
between business users and analytics practitioners [11].
Another well-known challenge is that data scientists
still spend 80% of their time “[...] finding, cleaning,
and organizing data” [34, p.4]. To improve this
situation, the alignment with business users and with
data management experts is of particular importance.
Consequently, researchers have started investigating
analytics governance forms. Based on a literature
review, [9] suggest a framework of the structural
(organization structure, coordination and alignment,
and roles and responsibilities), procedural (process
model, monitoring and evaluation, and development),
and relational mechanisms (shared perceptions,
collaboration, and transfer of knowledge). These
authors apply their framework to analyze three case
studies, with only one case appearing to have reached
high analytics governance maturity. Other studies
focus on specific areas of analytics governance or
particular organizational setups. [21], for instance,
examine analytics competence centers to understand
how analytics capabilities can be cultivated in
enterprises. While the core of these researchers’
investigation is strategic by nature and on capability
Page 5698

development, they also identify analytics competence
centers’ (ACC) typical roles and responsibilities and a
general process for ACCs. Their findings can therefore
be considered structural and procedural governance
mechanisms, although they are not phrased as such.
With the increasing application of artificial
intelligence and machine learning, new risks are
encountered, for example, discriminatory effects and
privacy infringements [35]. These risks have led to
increasing public awareness and questions about how
these technologies should be governed in the future.
[22] investigates accountability in algorithmic
decision-making, proposing an algorithmic
transparency standard. [23] formulates a research
agenda for artificial intelligence (AI) governance to
overcome risks such as “[…] labor displacement,
inequality, an oligopolistic global market structure,
reinforced totalitarianism, shifts and volatility in
national power, strategic instability, and an AI race that
sacrifices safety and other values” [23, p.1]. While
these AI governance considerations could have
implications for enterprises, their scope is more
political.
2.4. Research gap
BDA has led to new challenges emerging and
enterprises struggling to gain a return on their BDA
investments. In these situations, governance -
specifically defined responsibilities and data
accountabilities - is considered a catalyst for BDA
value creation [3, p.417]. Existing studies focus either
on data or on analytics governance, even though both
approaches are closely interlinked and depend on each
other [8]. We therefore argue that an integrated view is
needed to overcome the challenges and ensure that
BDA investments generate value. Nevertheless, most
studies have investigated data and analytics
governance by uncovering general governance
mechanisms. The identified structural, procedural, and
relational mechanisms mirror the IT governance
literature. While existing studies mainly focus on
identifying the generic mechanisms, we lack further
details of structural mechanisms in the context of BDA
and insights into their design and implementation in
enterprises. Research on decision rights is, for instance,
part of the structural governance mechanisms stream,
but has not investigated the roles and responsibilities in
greater detail. While governance models for master
data and data quality have elaborated on the roles and
responsibilities, the first attempts have been made to
define common analytics roles, albeit for educational
purposes [36]. Consequently, the definition and
assignment of roles and responsibilities for data and
analytics are a promising direction, which [4] also is
emphasizes.
3. Methodology
In this paper, our objective is to develop structural
mechanisms for enterprise-wide data and analytics
governance with a particular focus on roles and
responsibilities. Using design science research (DSR)
[12], we worked closely with five enterprises (see
Table 2) for a period of 17 months to understand their
current challenges and approaches in term of governing
BDA and developed a role model in multiple iterations.
All the enterprises are suitable with regard to
supporting our research goals, because they (1) have a
high maturity regarding managing data (2) have gained
first experiences with adopting BDA (e.g. all the
companies have an established data lake), and (3) are
in the process of (re-)designing their data and analytics
governance models.
Table 2 Companies involved in the research process
Company
Industry
Revenue/
# Employ.
Current situation
A
Consumer
goods
50–100 B$ /
~80 000
Central data and analytics
management unit to operate a
central big data platform and
distributed BI infrastructures
B
Public
transport-
tation
1–50 B$ / ~35
000
Central data management
organization, decentralized
analytics teams working with
multiple analytics infrastructures
C
Industry
products
50–100 B$ /
~110 000
Central data management org. and
advanced analytics group operating
multiple data lakes and data
warehouses
D
Consumer
goods
1–50 B$ / ~30
000
Central data and analytics
management organization with a
high business intelligence maturity
managing a central enterprise data
warehouse and one data lake
E
Manu-
facturing
1–50B$ /
~90 000
Central data management
organization and central platform
team that enable digital innovations
and industrialize analytics products
The chosen DSR process model encompasses six
steps [12]. Before starting this process, an adequate
research entry point had to be selected. All the
companies already had an established role model, but
with a focus on data management. The major
drawbacks were, on the one hand, that the existing
roles and responsibilities originated from master data
management and could only address the new BDA
requirements partly. On the other hand, analytics-
related roles were somewhat defined, but not integrated
into the overall governance framework. We therefore
initiated the research process with an objective-
centered solution, asking the question: what would a
Page 5699

better role model accomplish? [12]. In the first step
(identify problem & motivate), we conducted one
expert interview with each company to understand
their initial situation. We also held a one-day focus
group workshop in February 2019, where we presented
the findings and discussed the challenges of managing
data lakes compared to traditional business intelligence
environments. The group agreed that existing data
governance models needed to be extended to cope with
BDA requirements and achieve a unified enterprise
view. In the second step, we defined the objectives of
the solution. In March 2019, we allowed the companies
to first present their current roles and responsibilities
and thereafter discussed the missing roles and
responsibilities in the BDA context in a second focus
group meeting that was held virtually. During this
session, the group reached consensus that (1) the
general structuring of data and analytics organizations
needed to be reviewed in the context of BDA and (2)
that the outcome should be a framework depicting best
practices that could be used as reference to design roles
and responsibilities for data and analytics governance.
To fulfill these requirements, we developed the
model in three design & development, demonstration,
and evaluation iterations (see Figure 1).
In the first iteration (Mar–Aug 2019), we focused
on the roles and responsibilities of the emerging
analytics platforms. In order to do so, we analyzed
each company’s role models for their data lakes, and
conducted a literature review to attain a list of relevant
data and analytics roles. We then defined an area of
responsibility for each role, based on an information
supply chain depicting the steps needed to provide
analytics products. In a focus group, we demonstrated
how this first version of the role model could be used
to assign roles and allocate their responsibilities,
thereafter asking the participants to assess the model’s
applicability and usefulness. While the companies
appreciated the ease with which they understood the
framework, they remarked that the focus was too
narrowly on analytics platforms and failed to determine
the required responsibilities for managing data and
analytics enterprise-wide. In addition, we evaluated the
list of data and analytics roles by means of a survey,
asking companies to assess the completeness of the
roles and assign the area of responsibility in the context
of their company.
In the second iteration (Sept 2019–Dec 2019), we
revised the model, integrating the previous iteration’s
feedback. We structured the framework by
categorizing the roles and responsibilities according to
their organizational level with regard to the strategy,
governance, and operations. Furthermore, we extended
the information supply chain with an additional step to
emphasize the creation and maintenance of data on the
source level. We demonstrated the new version by
applying it to two analytics products (advanced
analytics model and dashboard) during an executive
course with a group of 15 professionals (Sept 2019). In
addition, we ran a focus group with the involved
companies. While the categorization regarding the
strategy, governance and operations ensured that all the
relevant roles and responsibilities were assigned, it
became apparent that the accountable roles had to be
allocated first. This led to discussions about the locus
of accountability and the decision to identify core
data/analytics roles that have to interact with business
and IT roles.
In the third iteration (Jan 2019–May 2020), we
designed the final version (see Section 4). We
demonstrated its general applicability by mapping the
companies’ role models to the framework and
presented the results in a focus group. The group
reached consensus that the framework was useful and
could be used to allocate roles and responsibilities in
the context of BDA, thus fulfilling the solution’s
defined objectives.
Figure 1 Design science research iterations according to [8]
4. Roles and responsibilities for BDA
In the following, we present the suggested role
model for enterprise-wide data and analytics
governance. First, we discuss the general principles
which guide the design of roles and responsibilities for
data and analytics governance in enterprises. Second,
we present the roles and responsibilities and illustrate
the framework in terms of a typical organization.
4.1. Design principles
At the beginning of our research process, three
important design decisions were taken that guided the
definition and assignment of roles and responsibilities
for data and analytics in enterprises.
Iteration 1
03-08/19
Iteration 2
09-12/19
Iteration 3
01-05/20
Design &
Development Demonstration Evaluation
V 0.3
V 0.7
V 1.0
Demonstration of the
role model’s general
mechanisms to a
focus group.
List of roles and
assignment of
responsibilities on
the information
supply chain.
Focus group
discussion on the
applicability and
usefulness. Survey
evaluation.
Instantiation of
the role model for
two analytics
products during an
executive course.
Categorization of roles
according to the organi-
zational level. Extension
of the information supply
chain.
Focus group
discussion on its
applicability and
usefulness.
Demonstration by
mapping the com-
panies’ role concepts
to the new model in
the focus group.
Design of final
model that allocates
roles and
responsibilities on
two axes.
General consensus
that the role model
fulfills the
solution’s defined
objectives.
Page 5700

Federated model for enterprise-wide data and
analytics governance - Any data management and
analytics activities require alignment and collaboration
with business and IT departments. The complete
centralization of data and analytics in an enterprise
would potentially increase the economies of scale, but
diminish business units’ flexibility (no customization)
and value generation through data and analytics (no
business self-responsibility, because the ownership
remains with the data and analytics organization).
Conversely, complete decentralization makes business
units flexible, but leads to data silos and hinders data
sharing and integration across functions [25].
Consequently, a federated approach is needed for
enterprise-wide data and analytics governance. This
implies that the roles and responsibilities can be
assigned to employees who work in different parts of
the enterprise:
• A federated data and analytics organization relies
on central teams with core data and analytics
roles that coordinate data management and
analytics delivery activities at the enterprise level.
They ensure that business requirements are
correctly transformed into data and analytics
products.
• Since the ownership of data and analytics
products lies with business [8], [37], business
roles play an important role. They define business
requirements for data and analytics products and
own their data and analytics products.
• IT roles support data management and analytics
delivery by means of infrastructure and IT
services. This includes the operation of analytics
products and the development of analytics
platforms.
Governance at the intersection between
strategy and operations - Generally speaking,
governance implements a strategy by means of
oversight and control mechanisms [38] and
complements strategic as well as operational tasks:
Strategy is doing the right things, operations are doing
things right, and governance is ensuring that the right
things are done right.
The suggested model therefore considers
governance roles as complementing the roles on the
strategic and operational levels.
• The objective and long-term direction for data
and analytics are defined at the strategic level.
This includes sponsorship, strategic direction,
funding, and the coordination of data
management and analytics activities at an
enterprise-wide level.
• The governance level implements the strategy
through oversight and control mechanisms. While
enterprise-wide data and analytics governance is
cross-functional, defines the overarching
governance framework and controls its
implementation, it needs to be detailed for the
different business units or departments by
defining the standards and the policies of the
areas of responsibility.
• The operations level executes the strategy through
day-to-day activities, operates the data and
analytics product lifecycle based on the defined
standards, and takes responsibility for the
correctness of the data content and the use of
analytics products.
Figure 2 Data and analytics governance at the intersection of
strategy and operations
Data and analytics roles facilitate the
information supply chain - While the data roles
emphasize the provision of data for different business
purposes, the analytics roles endeavor to deliver
analytics products throughout the enterprise and
integrate data across the business units. Both roles
clearly depend on each other and facilitate information
supply chains. Their responsibilities can be
summarized as follows:
• Data roles aim to make data fit for use for
business processes, data-driven insights, and
digital products and services.
• Analytics roles aim to deliver different types of
analytics products, for example, reports, ad-hoc
analysis, data science experiments, and
production.
According to these design considerations, we
allocate roles and responsibilities along two axes (see
Figure 2). The horizontal axis defines whether a role is
part of the core data and analytics organization, or
primarily a business and/or IT role. The vertical axis
allocates roles’ general responsibility according to the
strategy, governance, and operations. Using this
framework, roles can be allocated according to the
described dimensions. In this regard, it is important to
note that a role can be allocated to bordering areas if
Data and analytics
StrategyOperations Governance
Business IT
Defines the strategy:
sponsorship, strategic direction, funding,
and coordination of data management and analytics at an enterprise-wide level.
Implements the strategy by means of
corporation-wide oversight and control
mechanisms to ensure that the business
requirements are correctly transformed into
data and analytics products.
Executes the strategy through day-to-day activities:
operation of the data and analytics product lifecycle based on defined standards and
responsibility for the correct data content and the use of analytics products.
Define business
requirements and
own data and
analytics
products.
Define
infrastructure and
IT services to
deliver data and
analytics producs.
Page 5701

they only belong partially to the data and analytics
organization. This applies to data owners, for instance,
who are part of a business function but are assigned by
the data organization.
In the following, we present the governance-related
roles and responsibilities in detail. In contrast to
existing research, which mostly focuses on the
allocation of decision control rights (accountable roles)
[39], we define the accountable and responsible roles.
This is motivated by our observation that companies
are increasingly required to define their roles and
responsibilities at this level of detail to enforce
governance.
4.2. Roles and responsibilities in detail
Before we delve into the data and analytics
governance roles and responsibilities, we briefly define
the role of the Chief Data Officer, who plays as an
integrating role between the strategy and governance
level.
Chief Data Officer as integrating role - The role
of the Chief Data Officer (CDO) - also called head of
data and analytics or chief data and analytics officer –
is becoming of major importance in enterprises,
because data are increasingly recognized as a strategic
asset [33]. A recent study established that companies
with a CDO are twice as likely to have a clear digital
strategy [40]. Of the surveyed companies, 67,9% had a
CDO assigned in 2018 [40]. A CDO is the head of the
central data and analytics organization, is responsible
for the overall data management and analytics strategy,
and accountable for its implementation. This range of
activities requires continuous exchanges with the data
and analytics organization’s executive sponsor on the
business side, as well as with the chief information
officer (CIO) on the IT side. In the role model
suggested by [5], a CDO fulfills the chief data steward
role and extends his or her accountability to the
analytics organization.
In addition, companies increasingly establish a
dedicated data and analytics board comprised of C-
level executives to align the stakeholders on the
enterprise level. This board is accountable for defining
the data and analytics strategy, controlling its
implementation (including compliance requirements),
and setting priorities.
Data roles and responsibilities - An effective data
governance design (see Table 3) requires data
ownership to remain with the business functions. It
also requires a rather central organization of data
stewards and data architects, who, for instance, set and
enforce enterprise-wide standards for data
documentation, or facilitate data unification activities
to enable experimentation with and exploration of data
lakes.
Two types of ownership need to be distinguished
in terms of data ownership: the data definition owner
and the data content owner. Both roles are usually
assigned to senior executives responsible for a defined
business domain (e.g. business process) and who have
strategic responsibility (e.g. head of sales). In respect
of the data in his/her domain, the data definition owner
is accountable for data definitions of business and
quality rules, data access policies, data lifecycle, and
the conceptual data model. The data definition ensures
that data are created and used in controlled ways. The
data definition owner collects business requirements
for the defined area of responsibility (e.g. a particular
data domain like a business partner or product) from
other business process owners and from the
compliance officer.
Table 3 Data roles and responsibilities
Role name
Decision right and area
Allocation
Data
definition
owner
Accountable for the data definition in specific areas
of responsibility (e.g. a specific data domain).
Business
Data
steward
Responsible for the data definition in specific areas
of responsibility (e.g. the data field attributes of a
data object in a specific data domain).
Data & analytics
organization/
Business
Data
architect
Responsible for designing, creating, deploying, and
managing conceptual and logical data models as
well as for the mapping to physical data models.
Accountable for the implementation and
maintenance of data pipelines.
Data & analytics
organization/ IT
Data
content
owner
Accountable for data creation and maintenance
(data lifecycle) according to a specific area of
responsibility’s data definition.
Business/Shared
service center
Data
editor
Responsible for data creation and maintenance
(data lifecycle) according to a specific area of
responsibility’s data definition.
Business/Shared
service center
Data
expert
Responsible for communicating data the definition
and for training data editors.
Business/Shared
service center
While the data definition owner is accountable,
she or he is often at a high hierarchical level, for
instance, the head of material or supplier data
purchasing. The data steward is responsible for specific
areas of responsibility’s data definition and is often
part of the central data and analytics team. Here, the
data steward takes care of certain data fields of a data
object in a specific data domain. This includes defining
data while enforcing data quality measures and
ensuring that data is fit for use [41]. The data architect
supports the data steward by designing, creating,
deploying, and managing conceptual and logical data
models, as well as with mapping to physical data
models. In the role model defined by [5], the data
architect role corresponds to the technical data steward
role and complements the business steward. With the
Page 5702

emergence of new data types (e.g. IoT/sensor data) and
analytics use cases, the data definition needs to be
continuously adapted and serves as a central element to
ensure facilitated access and use across the enterprise.
The data steward therefore needs to handle data
requests from different business functions.
The data content owner’s role is usually assigned
to executives with operational responsibilities (e.g. the
head of sales of a specific country), who are
accountable for creating data according to the relevant
data definition. This role manages a team of data
editors, who are responsible for data creation. The data
expert is another typical role on the operations level.
This expert has no other major responsibility besides
communicating the data definitions to the data editors
and training them.
Analytics roles and responsibilities - An effective
analytics governance design (see Table 4) requires the
requestors and users of analytics products to
collaborate with the data and analytics organization
and IT.
On the business side, executives in business
domains who sponsor and request analytics products
usually represent the analytics product requirement
owner’s role. In this sense, she or he is accountable for
the business value and specification of an analytics
product’s business requirements. Accordingly, the
person assuming this role has to stimulate the
identification and use of analytics products in her or his
area of responsibility in order to increase data-driven
decision-making and communicate with important
business stakeholders. A business analyst, in the
analytics product requirement owner’s area of
responsibility, is responsible for the specification of the
analytics product on the operations level. While the
analytics product requirement owner specifies the
business requirements, the analytics product lifecycle
owner is accountable for implementing these
requirements in a specific analytics product, doing so
by coordinating its development, deployment, and
maintenance. In addition, this analytics product
lifecycle owner is responsible for defining analytics
product standards and guidelines, assuring quality, and
for managing the lifecycle as part of her or his
governance responsibility.
On an operations level, the coordinates the data
analysts, data scientists, and data engineers responsible
for analytics products’ development and deployment.
In order to do so, she or he involves the business
stakeholders to ensure that the business requirements
are met. The analytics product lifecycle owner is
typically a person with project management experience
with technical know-how of analytics product
development. The Analytics product architect’s role is
meant to ensure applications’ reusability and
scalability across the enterprise. This architect is
responsible for analytics products and analytics
product architecture’s design, which requires close
collaboration with the IT organization. Consequently,
this role is allocated to the bordering area of
analytics/IT.
Two data governance roles are of particular
importance for the analytics organization. The data
architect is accountable for data pipelines’
implementation and maintenance by providing the data
models that data engineers use. The data steward, a key
role for data governance, is responsible for managing
analytics projects’ data requests and for supporting the
data onboarding process. This support is of particular
importance to increase the analytics practitioners’
efficiency and reduce the time spent on finding and
preparing data.
Table 4 Analytics roles and responsibilities
Role name
Decision right and area
Allocation
Analytics
product
requirement
owner
Accountable for the business value and the
specification of the business requirements of
an analytics product.
Business
Analytics
product
architect
Responsible for the design of analytics
products and analytics product architecture.
Data & analytics
organization/ IT
Analytics
product
lifecycle
owner
Accountable for the implementation
(development and deployment) and
maintenance of an analytics product.
Responsible for analytics product standards
and guidelines, quality assurance, and the
lifecycle management.
Data & analytics
organization
Business
analyst
Responsible for the business value and
specification of an analytics product’s
business requirements.
Business
Data analyst
Responsible for the implementation
(development and deployment) and
maintenance of reports and ad-hoc analyses.
Data & analytics
organization
Data scientist
Responsible for the implementation
(development and deployment) and
maintenance of advanced analytics models.
Data & analytics
organization
Data engineer
Responsible for data pipelines’
implementation and maintenance.
Data & analytics
organization/ IT
Analytics
expert
Responsible for the training of analytics
product users.
Business/ Data &
analytics
organization
A central data and analytics organization ensures
that requests for new analytics products (e.g. data
science use case) are prioritized and specified within
an enterprise-wide demand management process.
Although all companies still distinguish between the
delivery of BI (e.g. reporting) and advanced analytics
products (e.g. predictive modelling), they seek an
integrated, unified view on analytics products’ demand
and delivery in the long term, in order to bundle
resources and facilitate BDA capabilities. Business
roles’ involvement guarantees that the business
Page 5703

requirements are met, and the domain knowledge is
transferred to analytics products.
5. Conclusions and outlook
Our study aims to integrate two distinct research
perspectives on data and analytics governance into a
unified view. Based on a close collaboration with five
large corporations with extensive BDA experience, we
provide a role model that explicates how structural
governance should be pursued in the context of BDA
on an enterprise level. More specifically, it defines
roles and responsibilities for data and analytics and
provides guidelines for their organizational
assignment. The enterprise-wide governance of data
and analytics requires a federated organization,
because business units need to assume ownership of
data content and analytics products. On the other hand,
a central data and analytics organization ensures that
data are "fit for use" and analytics products are
delivered efficiently across business functions and in
line with the strategic direction. Furthermore, roles and
responsibilities for data and analytics need to be
assigned on the strategy, governance, and operations
levels. This means that data and analytics governance
roles have to work hand-in-hand with strategic roles
and with those involved in day-to-day operational
tasks. The chief data officer (or chief data and analytics
officer) plays an important role in linking the strategic
and governance levels. Although data and analytics
have largely been viewed separately, their
interdependence in managing today’s information
supply chains cannot be questioned. The data steward
and data architect each play an integrating role
between the data and analytics organizations.
The role model and derived design considerations
contribute to IS governance literature in general and to
structural governance mechanisms in particular.
Practitioner can use the framework to design their own
data and analytics organizations.
Our study does have limitations. Since we only
collaborated with large, multi-national corporations to
build the framework, our results might not be
applicable to smaller size organizations. Furthermore,
we could not evaluate the actual implementation of the
defined roles and responsibilities since most companies
have not yet implemented all roles. However, all the
companies use the framework to design their own
governance models.
As already mentioned, this study contributes to a
larger research program in which we develop a
reference model for data and analytics governance. The
latter not only comprises structural mechanisms, but
also procedural and relational ones. The roles
presented in this paper help to clarify responsibilities in
processes and to define the required interactions. While
we have a clear plan for our research activities, we also
see interesting avenues for future research: The
presented roles can be used as a reference to
investigate how companies implement data and
analytics governance and compare different
organizational setups. Besides comparing centralized
to decentralized setups, a more in-depth investigation
of the coordination between data and analytics roles
with regard to overarching governance goals (for
instance, data quality) might be an interesting research
opportunity to foster the unified view that we suggest
in our study.
6. References
[1] B. Wixom and J. Ross, “How to Monetize Your Data,”
MIT Sloan Management Review, vol. 58, no. 3, 2017.
[2] U. Sivarajah, M. M. Kamal, Z. Irani, and V.
Weerakkody, “Critical analysis of Big Data challenges
and analytical methods,” Journal of Business Research,
vol. 70, pp. 263–286, Jan. 2017.
[3] V. Grover, R. H. L. Chiang, T.-P. Liang, and D. Zhang,
“Creating Strategic Business Value from Big Data
Analytics: A Research Framework,” Journal of
Management Information Systems, vol. 35, no. 2, pp.
388–423, Apr. 2018.
[4] A. Tiwana, B. Konsynski, and N. Venkatraman,
“Special Issue: Information Technology and
Organizational Governance: The IT Governance Cube,”
Journal of Management Information Systems, vol. 30,
no. 3, pp. 7–12, Dec. 2013.
[5] K. Weber, B. Otto, and H. Österle, “One Size Does Not
Fit All---A Contingency Approach to Data
Governance,” Journal of Data and Information Quality,
vol. 1, no. 1, pp. 1–27, Jun. 2009.
[6] J. J. Korhonen, I. Melleri, K. Hiekkanen, and M.
Helenius, “Designing Data Governance Structure: An
Organizational Perspective,” Journal on Computing,
vol. 2, no. 4, 2013.
[7] B. Otto, “Data Governance,” Business & Information
Systems Engineering, vol. 3, no. 4, pp. 241–244, Aug.
2011.
[8] M. Fadler and C. Legner, “Who owns data in the
enterprise? Rethinking data ownership in times of big
data and analytics,” Marrakesh, Morocco, 2020.
[9] H. J. Watson, C. Fuller, and T. Ariyachandra, “Data
warehouse governance: best practices at Blue Cross and
Blue Shield of North Carolina,” Decision Support
Systems, vol. 38, no. 3, pp. 435–450, Dec. 2004.
[10] R. Winter and M. Meyer, “Organization of data
warehousing in large service companies - A matrix
approach based on data ownership and competence
centers,” Journal of Data Warehousing, vol. 6, no. 4,
pp. 23–29, 2001.
[11] J. Baijens, R. W. Helms, and T. Velstra, “Towards a
Framework for Data Analytics Governance
Mechanisms,” 2020.
Page 5704

[12] K. Peffers, T. Tuunanen, M. A. Rothenberger, and S.
Chatterjee, “A Design Science Research Methodology
for Information Systems Research,” Journal of
Management Information Systems, vol. 24, no. 3, pp.
45–77, Dec. 2007, doi: 10.2753/MIS0742-1222240302.
[13] R. Peterson, “Crafting Information Technology
Governance,” Information Systems Management, 2004.
[14] P. Weill and J. Ross, “A Matrixed Approach to
Designing IT Governance,” MIT Sloan Management
Review, vol. 46, no. 2, pp. 26–34, 2005.
[15] A. Tiwana, B. Konsynski, and A. A. Bush, “Research
commentary-Platform evolution: Coevolution of
platform architecture, governance, and environmental
dynamics,” Information Systems Research, vol. 21, no.
4, pp. 675–687, 2010.
[16] J. Wareham, P. B. Fox, and J. L. Cano Giner,
“Technology Ecosystem Governance,” Organization
Science, vol. 25, no. 4, pp. 1195–1215, Mar. 2014.
[17] T. J. Winkler and C. V. Brown, “Horizontal Allocation
of Decision Rights for On-Premise Applications and
Software-as-a-Service,” Journal of Management
Information Systems, vol. 30, no. 3, pp. 13–48, Winter
2013.
[18] R. W. Gregory, E. Kaganer, O. Henfridsson, and T. J.
Ruch, “IT Consumerization And The Transformation
Of It Governance,” MIS Quarterly, vol. 42, no. 4, pp. 1–
9, 2018.
[19] P. P. Tallon, R. V. Ramirez, and J. E. Short, “The
Information Artifact in IT Governance: Toward a
Theory of Information Governance,” Journal of
Management Information Systems, vol. 30, no. 3, pp.
141–178, Dec. 2013.
[20] V. Khatri and C. V. Brown, “Designing Data
Governance,” Communication of the ACM, vol. 53, no.
1, pp. 148–152, 2010.
[21] R. Schüritz, E. Brand, G. Satzger, and J.
Bischhoffshausen, “How to cultivate analytics
capabilities within an organization? - Design and types
of analytics competency centers,” in Proceedings of the
25th European Conference on Information Systems
(ECIS), Guimarães, Portugal, Jun. 2017, pp. 389–404.
[22] N. Diakopoulos, “Accountability in algorithmic
decision making,” Communications of the ACM, vol.
59, no. 2, pp. 56–62, Jan. 2016.
[23] A. Dafoe, “AI Governance: A Research Agenda,”
Centre for the Governance of AI, Future of Humanity
Institute, University of Oxford, 2018.
[24] V. Sambamurthy and R. W. Zmud, “Arrangement for
Information Technology Governance: A Theory of
Multiple Contingencies,” MIS Quarterly, vol. 23, no. 2,
pp. 261–290, Jun. 1999.
[25] C. K. Velu, S. E. Madnick, and M. W. Van Alstyne,
“Centralizing Data Management with Considerations of
Uncertainty and Information-Based Flexibility,”
Journal of Management Information Systems, vol. 30,
no. 3, pp. 179–212, Dec. 2013.
[26] A. E. Brown and G. G. Grant, “Framing the
Frameworks: A Review of IT Governance Research,”
Communications of the Association for Information
Systems, vol. 15, 2005, Accessed: Apr. 15, 2020.
[Online]. Available:
https://aisel.aisnet.org/cais/vol15/iss1/38.
[27] S. De Haes and W. Van Grembergen, “IT Governance
and its Mechanisms,” Information Systems Control
Journal, vol. 1, pp. 27–33, 2004.
[28] C. Legner, T. Pentek, and B. Otto, “Accumulating
Design Knowledge with Reference Models: Insights
from 12 Years of Research on Data Management,”
Journal of the Association for Information Systems, vol.
21, no. 3, 2020.
[29] D. Loshin, Ed., “Master Data Management,” in Master
Data Management, Boston: Morgan Kaufmann, 2009,
pp. i–ii.
[30] H. A. Smith and J. D. McKeen, “Developments in
Practice XXX: Master Data Management: Salvation Or
Snake Oil?,” CAIS, vol. 23, no. 4, 2008.
[31] P. P. Tallon, “Corporate Governance of Big Data:
Perspectives on Value, Risk, and Cost,” Computer, vol.
46, no. 6, pp. 32–38, Jun. 2013.
[32] P. Mikalef, J. Krogstie, R. van de Wetering, I. O.
Pappas, and M. N. Giannakos, “Information
Governance in the Big Data Era: Aligning
Organizational Capabilities,” 2018.
[33] Y. Lee, S. Madnick, R. Wang, F. Wang, and H. Zhang,
“A Cubic Framework for the Chief Data Officer
(CDO): Succeeding in a World of Big Data,” MIS
Quarterly Executive, vol. 13, no. 1, 2014.
[34] H. Bowne-Anderson, “What Data Scientists Really Do,
According to 35 Data Scientists,” Harvard Business
Review Digital Articles, pp. 2–5, Aug. 2018.
[35] B. Custers, “Data Dilemmas in the Information Society:
Introduction and Overview,” in Discrimination and
Privacy in the Information Society: Data Mining and
Profiling in Large Databases, B. Custers, T. Calders, B.
Schermer, and T. Zarsky, Eds. Berlin, Heidelberg:
Springer Berlin Heidelberg, 2013, pp. 3–26.
[36] J. Saltz, F. Armour, and R. Sharda, “Data Science Roles
and the Types of Data Science Programs,”
Communications of the Association for Information
Systems, vol. 43, pp. 615–624, 2018.
[37] M. Van Alstyne, E. Brynjolfsson, and S. Madnick,
“Why not one big database? Principles for data
ownership,” Decision Support Systems, vol. 15, no. 4,
pp. 267–284, Dec. 1995.
[38] A. Tiwana and S. K. Kim, “Discriminating IT
Governance,” Information Systems Research, vol. 26,
no. 4, pp. 656–674, Oct. 2015.
[39] T. J. Winkler and M. Wessel, “A Primer on Decision
Rights in Information Systems: Review and
Recommendations,” in Proceedings of ICIS 2018, San
Francisco, CA, 2018, p. 17.
[40] Forbes, “Forbes Insights: Rethinking The Role of Chief
Data Officer,” Forbes.
https://www.forbes.com/sites/insights-
intelai/2019/05/22/rethinking-the-role-of-chief-data-
officer/ (accessed Jul. 14, 2020).
[41] R. Y. Wang and D. M. Strong, “Beyond accuracy: What
data quality means to data consumers,” Journal of
Management Information Systems, vol. 12, no. 4, pp. 5–
33, 1996.
Page 5705