A Methodology for Constructing Patterns for the Management of Data

Science Projects

Christian Haertel

, Sarah Schramm, Matthias Pohl

, Sascha Bosse

, Daniel Staegemann

Christian Daase

and Klaus Turowski

Institute of Technical and Business Information Systems, Otto-von-Guericke University, Magdeburg, Germany

{christian.haertel, sarah.schramm, matthias.pohl, sascha.bosse, daniel.staegemann, christian.daase,

Keywords:

Data Science, Project Management, Pattern, Design Science Research.

Abstract:

In the era of Big Data, the successful completion of Data Science (DS) projects is crucial. However, DS

project management is quite challenging due to its interdisciplinary nature. Existing DS process models, such

as CRISP-DM, have limitations, resulting in low success rates for these undertakings. To address this issue,

a novel methodology for the construction of patterns in DS project management has been proposed, using

the Design Science Research methodology. The design draws inspiration from existing pattern concepts to

address common problems in DS project execution. The methodology is demonstrated through the creation of

patterns for best practices in DS project management, synthesized from scientiﬁc literature. The goal of this

approach is to provide a platform for exchanging and standardizing best practices in DS project management.

While initial demonstrations show the general applicability of the methodology, further evaluations and case

studies are necessary to assess its effectiveness and areas for improvement. The study identiﬁes potential

ambiguities in certain activities within the process, suggesting opportunities for reﬁnement. Overall, this

research contributes to the ﬁeld of DS project management by offering a structured method to encapsulate and

disseminate effective practices, supporting the successful execution of data projects in organizations.

1 INTRODUCTION

In a world where the amount of generated data is

steadily increasing, businesses of various domains

aim to derive potential advantages and enhance their

competitive positioning (de Medeiros et al., 2020).

Accordingly, Data Science (DS) as a discipline to

extract knowledge and insights from data using

various methods and techniques (Chang and Grady,

2019), has gained increasing signiﬁcance (Cao,

2017). With its growing importance, the adequate

management of DS projects becomes crucial. How-

ever, organizations often encounter challenges in

implementing data-driven projects (Martinez et al.,

2021a). DS is considered an interdisciplinary ﬁeld

that combines various areas such as statistics, com-

https://orcid.org/0009-0001-4904-5643

https://orcid.org/0000-0002-6241-7675

https://orcid.org/0000-0002-2490-363X

https://orcid.org/0000-0001-9957-1003

https://orcid.org/0000-0003-4662-7055

https://orcid.org/0000-0002-4388-8914

puter science, machine learning, and domain-speciﬁc

knowledge, which poses unique challenges to project

management (Martinez et al., 2021a). Various DS

process models have been developed to support the

implementation and management of DS projects

(e.g., CRISP-DM) (Saltz, 2015). However, research

indicated several weaknesses in these methodologies

(Martinez et al., 2021a). Hence, there is no widely ac-

cepted and applied approach, and custom methods are

derived instead (Saltz, 2015; Saltz et al., 2018). These

issues are reﬂected in the low DS project success rate

(VentureBeat, 2019), demanding improvements for

DS project management (Saltz and Krasteva, 2022).

As the literature identiﬁes common problems in the

execution of DS projects (Martinez et al., 2021a),

the adaptation of the pattern concept to DS appears

promising. Patterns capture solutions to recurring

problems in a domain in a simple and straightfor-

ward form (Fehling et al., 2014). An overview and

structured presentation within patterns could ensure

alleviated and methodology-independent access to

common problems and solutions and, thus, contribute

to the improvement of DS project management

354

Haertel, C., Schramm, S., Pohl, M., Bosse, S., Staegemann, D., Daase, C. and Turowski, K.

A Methodology for Constructing Patterns for the Management of Data Science Projects.

DOI: 10.5220/0012705300003690

Paper published under CC license (CC BY-NC-ND 4.0)

In Proceedings of the 26th International Conference on Enterprise Information Systems (ICEIS 2024) - Volume 1, pages 354-365

ISBN: 978-989-758-692-7; ISSN: 2184-4992

activities. To the best of our knowledge, the pattern

concept has not been applied before to DS (project

management). Therefore, this research proposes a

methodology for pattern creation for the ﬁeld of DS

project management, using the pattern identiﬁcation,

authoring, and application process of (Fehling et al.,

2014) as a basis. Therefore, the following research

question (RQ) will be examined:

RQ: How can a methodology for the construction

of patterns for DS project management be designed

and applied?

For both, researchers and practitioners, this arti-

fact could form a platform for the exchange and stan-

dardization of best practices in the DS domain. Pat-

terns can be created, expanded, modiﬁed, and linked

to related areas. Overall, this study is intended to

contribute to supporting and conducting data projects

in organizations and, in turn, work toward improving

the success rates of these projects. The Design Sci-

ence Research (DSR) Methodology of (Peffers et al.,

2007) will be leveraged to develop the pattern con-

struction methodology. Therefore, in the next section,

the application of the DSR paradigm in the context of

this paper is described. Afterward, the relevant the-

oretical concepts for this work are discussed, includ-

ing DS and related terms, project management, and

fundamentals concerning the pattern concept. Con-

sequently, the pattern development process for DS

project management, an adaption of the process pro-

posed by (Fehling et al., 2014), is outlined. After

demonstrating the application of the artifact, this con-

tribution is concluded with a summary and an outlook

on future research endeavors.

2 METHODOLOGY

Scientiﬁc research seeks new insights and relation-

ships in a speciﬁc ﬁeld by applying scientiﬁc methods

(Eisend and Kuß, 2023). In the context of information

systems, DSR has emerged as a design-oriented ap-

proach to support the creation of artifacts to address

practically relevant problems (Hevner et al., 2004).

A pattern language construction methodology for DS

can also be categorized as an artifact (method), and

DSR can be considered suitable for its development.

The Design Science Research Methodology (DSRM),

according to (Peffers et al., 2007), is widely adopted

in the DSR context. According to the DSRM, an

artifact is designed, developed, and evaluated in six

phases. The individual steps for the work at hand are

explained in the following.

Problem Identiﬁcation and Motivation: DS

projects suffer from high failure rates (VentureBeat,

2019). Hence, the development of new or revised

approaches for DS project management is necessary

(Saltz, 2022). Pattern languages are utilized to doc-

ument solutions to recurring problems in a given do-

main (Fehling et al., 2014) and have not yet been ap-

plied to DS. Accordingly, adapting this concept to this

ﬁeld by proposing a speciﬁc methodology for the de-

velopment of patterns for DS can support addressing

common issues in the execution of these undertak-

ings.

Objectives of a Solution: The next step of the DSRM

involves deriving goals for a solution. Conducting a

DS project often involves encountering challenges re-

lated to team, project, and data and information man-

agement (Martinez et al., 2021a). The artifact of this

research, a methodology for the construction of DS

project management patterns, enables the capture of

solutions for these often-faced problems in DS from

the respective body of knowledge. Therefore, these

patterns can be utilized to overcome these obstacles

and, ultimately, lead to improved success rates in DS

undertakings.

Design and Development: Based on the deﬁnitions

by (Hevner et al., 2004), the developed artifact is

characterized as a method in the DSR context, as

the proposed pattern construction methodology pro-

vides a process on how to synthesize best practices

in the shape of patterns in the context of DS. For this

purpose, the general pattern identiﬁcation, authoring,

and application procedure of (Fehling et al., 2014) is

adapted to DS project management. Therefore, foun-

dations and methodologies from the DS knowledge

base are used.

Demonstration: The application of the artifact to de-

velop patterns for best practices in the management of

DS projects is demonstrated. Because of page restric-

tions, only snippets from the created patterns can be

shown in this paper.

Evaluation: This step aims to observe and measure

the ability of the artifact with regards to providing a

solution to the initially stated problem (Peffers et al.,

2007). Thus, the Build-Evaluate pattern of (Sonnen-

berg and vom Brocke, 2012) is applied for this DSR

project, consisting of the four steps Eval 1 (justiﬁed

research gap), Eval 2 (validated design speciﬁcation),

Eval 3 (proof of applicability), and Eval 4 (proof of

usefulness).

Communication: The ﬁnal step within the DSRM is

achieved through writing, submitting, and presenting

this paper to the scientiﬁc community and interested

practitioners of the ﬁeld.

A Methodology for Constructing Patterns for the Management of Data Science Projects

355

3 THEORETICAL BACKGROUND

The development of a methodology for the construc-

tion of patterns for DS project management initially

requires establishing a sufﬁcient knowledge base of

the underlying concepts. First, this implies discussing

the terminology around DS and its similarities as well

as delimitations to related disciplines. Additionally,

key aspects and existing research concerning (DS)

project management need to be covered. Finally, the

fundamentals of the pattern concept are introduced.

3.1 Data Science and Related Concepts

A widely accepted interpretation describes DS as the

methodology for synthesizing useful knowledge from

data through a process of discovery or formulation

and testing of hypotheses (Chang and Grady, 2019). It

is also characterized as an interdisciplinary ﬁeld (Cao,

2018). Therefore, knowledge and methods from vari-

ous disciplines are brought together to utilize data ef-

fectively (Schulz et al., 2020). DS has emerged as

a unique discipline where a deep understanding of

the application-speciﬁc domain, mathematical knowl-

edge, and a solid technological background are es-

sential (Schulz et al., 2020). DS has been used since

the mid-2000s, focusing on gaining insights from data

(Chang and Grady, 2019). However, similar goals

were pursued earlier under different names. These re-

lated concepts are closely intertwined and cannot be

easily separated from one another (Chang and Grady,

2019). Examples are Data Mining and Big Data.

Data Mining is the exploration of patterns and

relationships in data using specialized algorithms

(Chang and Grady, 2019). Skills in statistics, math-

ematics, machine learning, algorithms, and domain

knowledge are applied in this process (Chang and

Grady, 2019). Data Mining is known as a part of

Knowledge Discovery in Databases (KDD), which

constitutes a comprehensive process for extracting

valuable knowledge from data (Fayyad et al., 1996;

Chang and Grady, 2019). Data Mining is a step in

this process (Fayyad et al., 1996). The proximity to

DS is evident, explaining the sometimes synonymous

use of the terms (Schulz et al., 2020). Generally, Data

Mining can be understood as a subﬁeld of DS (Chang

and Grady, 2019).

Big Data can be deﬁned by the four dimensions

(Vs) Volume (enormous size of datasets), Velocity

(speed of data generation and capturing), Variety (in

data sources and formats), and Variability (changes

in data ﬂow, format, or volume) (Jeble et al., 2018;

Chang and Grady, 2019). Additionally, while other

characteristics such as Value (value of results), Ve-

racity (data quality), or Complexity (data complexity)

can be deﬁned (Jeble et al., 2018), these dimensions

can be considered drivers for new scalable architec-

tures for data-intensive applications. Big Data posed

a challenge for data processing as well as analysis and

gave rise to the term DS to explore new techniques

for this matter (Chang and Grady, 2019). Thus, DS

encompasses Big Data and its analysis (Chang and

Grady, 2019). Based on the outlined overlaps be-

tween these disciplines, approaches for these domains

tend to be overarching.

3.2 Data Science Project Management

Project management can be deﬁned as the ”applica-

tion of knowledge, skills, tools, and techniques to

project activities to meet the project requirements”

(PMI, 2017). Within project management, various

tasks such as project planning, risk management, and

many more are fulﬁlled (Wack, 2007). During project

planning, the scope, schedules, budget, and resource

allocation to achieve the project goals are deﬁned

(Aichele and Sch

onberger, 2014). In IT projects,

the principles, procedures, methods, techniques, and

tools necessary for planning, controlling, and moni-

toring are summarized within IT project management

(Heinrich, 1997; Wieczorrek and Mertens, 2007). In

comparison, DS undertakings differ because of the

data focus and the resulting explorative character (Das

et al., 2015). Accordingly, both typical IT project

management methods and more speciﬁc process mod-

els are used in DS (Saltz and Hotz, 2020). Several

DS process models can be identiﬁed in the literature,

including KDD, CRISP-DM, TDSP, and SEMMA.

Typically, according to the DS lifecycle of (Haertel

et al., 2022), a DS project can be structured in the

broad phases Business Understanding, Data Collec-

tion, Exploration and Preparation, Analysis, Evalua-

tion, Deployment, and Utilization. However, research

suggests that the current DS methodologies are par-

tially unsuitable for tackling common challenges in

DS projects related to team, project, and data and

information management (Martinez et al., 2021a).

Hence, revised and new approaches are needed (Saltz

and Krasteva, 2022).

3.3 Pattern

For experts working on a problem, it is unusual to

develop a new solution that is entirely different from

existing ones (Buschmann, 1996). Often, there is a

tendency to rely on a similar and already solved prob-

lem, using the essential elements of the solution to

address the new problem (Buschmann, 1996). An

ICEIS 2024 - 26th International Conference on Enterprise Information Systems

356

established concept for the structured documentation

of proven solutions is patterns (Fehling et al., 2014).

This principle originates from the architect Christo-

pher Alexander, who laid its foundation back in 1977

(Alexander, 1979; Coplien and Harrison, 2005). Pat-

terns describe a frequently occurring problem and the

core of its solution so that it can be used repeatedly

in different ways (Alexander et al., 1977). Patterns

are recorded in text form following a speciﬁc struc-

ture and generally consist of a particular context, a

problem, and a solution (Alexander, 1979). They are

hierarchically organized, documented, and intercon-

nected. Connections can be drawn within and to re-

lated subject areas, thus creating a complex and ab-

stract solution system that can be individually ac-

cessed (Fehling et al., 2014; Coplien, 2000). The

connections highlight relationships and capture hid-

den structures (Coplien, 2000). Accordingly, naviga-

tion is facilitated, and references to the application of

patterns are provided, resulting in the creation of a

pattern language (Fehling et al., 2014). In essence,

patterns can be understood as a concept for the struc-

tured documentation of proven solutions to recurring

problems in a speciﬁc domain (Fehling et al., 2014).

Although initially developed for architecture, the

concept was successfully transferred to other do-

mains such as the software ﬁeld (Coplien and Har-

rison, 2005). Other authors have adopted, extended,

modiﬁed, and inﬂuenced the initial context-problem-

solution structure shaped by (Alexander et al., 1977).

Ultimately, the compilation, content, and sequence of

pattern sections are fundamentally left to the authors.

There are no predeﬁned rules for the identiﬁcation

and documentation of patterns, but there are guide-

lines for orientation like in (Wellhausen and Fiesser,

2011) and (Harrison, 2003). The creation of pat-

terns is often performed through expertise and expe-

rience or in collaboration with experts (Iba and Isaku,

2012). Another possibility is the extraction from lit-

erature (Fehling et al., 2014; G

unther and Knote,

2017). For instance, (Fehling et al., 2014) describe

a pattern-creation process that is used in this work.

Application of the pattern concept to (DS) project

management also appears sensible since several com-

mon challenges need to be addressed in the course

of a project. Accordingly, the synthesis of knowledge

from experts and the literature for summarizing corre-

sponding solutions in the form of patterns can provide

added value.

Figure 1: Pattern construction process, adopted from

(Fehling et al., 2014).

4 PATTERN CONSTRUCTION

PROCESS FOR DATA SCIENCE

PROJECT MANAGEMENT

In the literature, there are different approaches to the

development of patterns. For the application in this

context, the process according to (Fehling et al., 2014)

was selected and adapted to the ﬁeld of DS project

management since its applicability to various domains

has already been demonstrated. Despite the weak-

nesses outlined in DS methodologies, several publi-

cations discuss success factors and best practices for

DS project management. Therefore, the consolida-

tion of this knowledge is beneﬁcial. (Fehling et al.,

2014) describe a detailed approach to the identiﬁca-

tion and authoring of patterns and, thus, allows the

creation of patterns from the knowledge and exper-

tise conveyed through the literature. The following

subsections delve into the individual steps of this pro-

cess, based on which DS patterns shall be developed.

The method consists of three phases: pattern identi-

ﬁcation, pattern authoring, and pattern application, as

illustrated in Figure 1. Each stage involves several it-

eratively traversed activities to continuously improve

and adapt the developed results (Fehling et al., 2014).

In particular, the ﬁrst two phases are repeated multi-

ple times to discover and form patterns. Finally, the

third phase involves reﬁning the patterns for speciﬁc

use cases or application environments.

4.1 Pattern Identiﬁcation

The ﬁrst phase of the process, according to (Fehling

et al., 2014), is carried out through ﬁve iteratively tra-

versed activities as depicted in Figure 2. The exam-

A Methodology for Constructing Patterns for the Management of Data Science Projects

357

Figure 2: Pattern Identiﬁcation phase, adopted from

(Fehling et al., 2014).

ined domain is initially structured in this phase, and

relevant information is gathered. In the ﬁrst activ-

ity (Domain Deﬁnition), signiﬁcant fundamentals of

the investigated ﬁeld are elaborated and shared with

the group working on the patterns to build common

knowledge and understanding of the domain as es-

tablishing a foundation for the ﬁeld is considered im-

portant (Fehling et al., 2014). Initially, this requires

creating a joint understanding of key terminology and

concepts of DS (e.g., process models, ML algorithms)

as seen in Section 3. Furthermore, since DS is a com-

plex and interdisciplinary ﬁeld, the focus is purely on

DS project management.

The next activity, Coverage Consideration, fo-

cuses on assessing and narrowing down the scope of

the chosen domain (Fehling et al., 2014). The domain

can be extensive, making it impossible to consider the

entire scope. Accordingly, the scope is adjusted to the

size of the research group. Relevant topics are iden-

tiﬁed and aligned with characteristic problems of the

domain. Since DS project management is extensive

itself, it might be sensible to further limit the scope

to the issues for the group (e.g., best practices, certain

DS project stages) that mainly impact its DS project

success. Therefore, considered information sources

can be limited to a subset.

The subsequent task, Information Format Design,

aims explicitly at team collaboration and establish-

ing a uniﬁed structure for information capture and

processing. As patterns will be identiﬁed based on

information extracted from literature, uniform tools

and templates for the collection, ﬁltering, and analy-

sis should be deﬁned to achieve an efﬁcient and trans-

parent process.

The following activity, Information Collection, in-

volves gathering information and coordinating its pro-

cessing within the research group (Fehling et al.,

2014). Since not all required information will be part

of human memory, other information sources are re-

quired. Therefore, we propose the use of a litera-

ture review to acquire relevant material in a structured

manner. The search process can involve both aca-

demic and non-academic databases. Depending on

the chosen focus within the Coverage Consideration,

the inclusion of grey/white literature might be help-

ful. It is recommended to use an established litera-

ture reviewing methodology (e.g., (vom Brocke et al.,

2009)) to ensure rigor in this process. The search

terms and inclusion/exclusion criteria should be de-

termined within the group based on the previously

selected DS scope. To identify patterns in the next

phase, common themes and solutions from the litera-

ture have to be synthesized based on the deﬁned struc-

tures from the Information Format Design.

In the last activity of this phase, Information Re-

view, the information sources and the solutions to be

considered therein are reviewed concerning their fur-

ther processing by the research group (Fehling et al.,

2014). The obtained volume of literature may be too

large for the group to handle, reﬁning the domain

structure to create smaller and achievable sets of so-

lutions. For example, limiting the scope to certain DS

project activities might be useful. The criteria for lit-

erature review, both in terms of content and the pos-

sibilities for pattern development, can also be sharp-

ened accordingly. The result set is further narrowed

down in different ﬁltering stages until a feasible set re-

lated to the speciﬁc problems is identiﬁed. If needed,

forward and backward searches can be appended to

further strengthen the pool of information sources.

4.2 Pattern Authoring

In the second phase, ﬁve activities are similarly under-

taken (see Figure 3), and the patterns are formulated

based on the similarities of existing solutions from

the information basis (Fehling et al., 2014). There-

fore, in the Pattern Language Design, the structure

and sections of the patterns are deﬁned (Fehling et al.,

2014). As there is no universally valid format, these

are individually determined by the pattern authors and

adapted to their speciﬁc needs. Here, we recommend

using a format using the sections name, problem, con-

text, challenges, solution, results, and links, mainly

built on the work of (Wellhausen and Fiesser, 2011).

The problem is documented to support users in identi-

fying and evaluating its suitability for their speciﬁc is-

sues. Context deﬁnes the setting in which the pattern

occurs. The difﬁculties encountered in addressing a

ICEIS 2024 - 26th International Conference on Enterprise Information Systems

358

Figure 3: Pattern Authoring phase, adopted from (Fehling

et al., 2014).

problem are also captured under challenges (Well-

hausen and Fiesser, 2011). The application of the so-

lution leads to results. To facilitate the application of

patterns and illustrate their interconnections, relation-

ships to other patterns are captured in the links (Well-

hausen and Fiesser, 2011). Additionally, patterns are

given a name by which they are identiﬁed. Each pat-

tern in this context should be uniformly built upon the

mentioned elements.

With the next activity, Primitive Deﬁnition, the

deﬁnitions from Information Format Design can be

further elaborated if needed (Fehling et al., 2014),

depending on the insights gained from the literature.

This aims to ensure consistent usage and homogeniza-

tion of primitives within the research group.

In Composition Language Design, guidelines for

sketches and formal speciﬁcations of the composi-

tion language are determined. For DS project man-

agement, this could involve the use of a DS process

model (e.g., CRISP-DM) to clarify further the appli-

cability of a pattern solution within a DS project.

The next activity, Pattern Writing, involves the ac-

tual documentation of the patterns based on the previ-

ously deﬁned structure. The identiﬁed patterns from

the literature are abstracted to a degree so that suf-

ﬁcient information is provided while remaining ab-

stract enough to apply to various cases (Fehling et al.,

2014). The process is iteratively repeated, with the

patterns and individual sections revised and harmo-

nized repeatedly. Discussions with other pattern au-

thors or users are essential. Documentation is per-

formed based on the approach presented in (Well-

hausen and Fiesser, 2011) and begins with the so-

lution of the pattern. Notes for the identiﬁed pat-

terns and the information retrieved from the DS lit-

erature review are used for this purpose. Next, the

problem is formulated concerning the described so-

lution. The problem should not be trivial and is elu-

cidated by questioning the relevance of the solution

and the actual problem being addressed (Wellhausen

and Fiesser, 2011). The result is then formulated to

assess the impact of applying the solution. Subse-

quently, the challenges related to the results and the

solution are identiﬁed. This involves examining why

the described problem is more challenging to solve

than it might initially appear (Wellhausen and Fiesser,

2011). Only afterward is the context described, deter-

mining the circumstances under which the problem

arises. There is no optimal time to specify the name of

the pattern, and it can emerge during the development

process. In this case, the focus is on what supports the

recall of the solution (Wellhausen and Fiesser, 2011).

The links section is ﬁlled at the end once the interre-

lationships with other patterns have been clariﬁed.

In the last activity of this phase, Pattern Lan-

guage Revision, the created pattern language is evalu-

ated and revised (Fehling et al., 2014). This can also

involve the incorporation of DS practitioners to as-

sess the usefulness of the individual patterns. Addi-

tionally, the structure and links between the patterns

should be investigated. Patterns written at the begin-

ning may have fewer connections and require revi-

sion.

4.3 Pattern Application

Figure 4: Pattern Application phase, adopted from (Fehling

et al., 2014).

The third phase, according to (Fehling et al., 2014),

consisting of four subactivities (see Figure 4), per-

tains to the application of the developed patterns and

can be considered and worked on independently of

A Methodology for Constructing Patterns for the Management of Data Science Projects

359

the other two phases. First, Pattern Search and Rec-

ommendation aims to facilitate users’ navigation and

identify suitable patterns for the speciﬁc use case at

hand (Fehling et al., 2014). Accordingly, the derived

DS patterns shall be accompanied by a summary sec-

tion containing a brief description of the problem and

solution of each of the patterns of the pattern language

within a sentence (e.g., in tabular form) (Meszaros

and Doble, 1997; Manns and Rising, 2012). There-

fore, relevant patterns can be selected and further ex-

amined. For improved navigation, the patterns con-

tain a links section, which indicates the relationship

to other patterns. Moreover, it is recommended that

this connection is graphically represented, too.

In Pattern-based Solution Design, support is pro-

vided for translating the abstract solution of a pattern

into a speciﬁc use case (Fehling et al., 2014). The

original literature, from which the solutions were ab-

stracted, can be revisited to include concrete, exist-

ing solutions in the patterns to facilitate their usability

(Fehling et al., 2014). Accordingly, the pattern nota-

tion is expanded by the section ”Examples”, which

provides a detailed reference solution of the DS prob-

lem described within the given pattern.

In the next activity, Reﬁnement of the Solution De-

sign, patterns are constrained and adapted to a speciﬁc

environment where they should be applied (Fehling

et al., 2014). In the DS context, this could involve a

limitation to certain types of DS projects where the

patterns are used to combat the possible differences

in the manual pattern implementations.

Because of the determined pattern reference im-

plementations and limitation to viable use case

type(s), in the last activity, Instantiation of the Solu-

tion Design, the means to manage, conﬁgure, and de-

ploy the patterns are determined (Fehling et al., 2014).

Afterward, a speciﬁc reﬁnement of the DS patterns

based on the previously selected focus is enabled.

5 DEMONSTRATION

In alignment with the adopted DSR methodology, the

application of the artifact to develop patterns for the

management of DS projects is demonstrated in the

following. This section describes how the authors ap-

plied the individual steps of the introduced method

for identifying and notating patterns in the DS project

management domain. Due to the page and time re-

strictions for this research, not all activities can be

outlined to the fullest extent in this paper.

5.1 Pattern Identiﬁcation

Domain Deﬁnition: As the goal of this research is

centered around the development of patterns for DS

project management, a joint understanding of theo-

retical fundamentals in this domain had to be estab-

lished. Since some research group members already

acquired some experience in the ﬁeld, reference lit-

erature on conceptualization (e.g., NIST deﬁnitions)

and common approaches (e.g., process models) were

exchanged and discussed. The common understand-

ing of the team was recorded. Section 3 constitutes

the result of this activity.

Coverage Consideration: Due to the extensive na-

ture of DS project management, it was necessary to

limit the scope. Therefore, the focus was set on gen-

eral successful approaches and the associated best

practices regarding DS project management to ad-

dress common problems and work toward mitigating

the high failure rates encountered in the execution of

data science projects (VentureBeat, 2019). The pat-

terns should guide common challenges that arise dur-

ing DS project phases and contribute to the develop-

ment of effective data science project management.

Information Format Design: To adequately pre-

pare for the Information Collection step, the research

group jointly agreed to the joint use of certain tools

and templates to facilitate collaboration for the fol-

lowing activities. For example, Citavi was used as

a reference management tool for the literature. Here,

the individual ﬁltering stages were also depicted using

the category feature. Accordingly, grouping related

publications and mapping the inclusion/exclusion cri-

teria was possible. Furthermore, to further organize

the analysis of the obtained material, it was decided

to use the concept matrix of (Webster and Watson,

2002). A corresponding template was created in Mi-

crosoft Excel, which the group jointly used.

Information Collection: The Information Collection

for creating a DS project management pattern lan-

guage was performed through a structured literature

review according to the guidelines of (vom Brocke

et al., 2009), consisting of the ﬁve phases deﬁnition

of review scope, conceptualization of topic, litera-

ture search, literature analysis and synthesis, and re-

search agenda. After completing the ﬁrst two steps of

this framework during Domain Deﬁnition and Cov-

erage Consideration, Scopus and SpringerLink were

queried with the search terms shown in Table 1, cor-

responding to the previously deﬁned scope. The

search yielded 286 results for Scopus and 205 pub-

lications for SpringerLink, where merely articles and

conference papers were considered. In alignment

with the outcome of Coverage Consideration, the re-

ICEIS 2024 - 26th International Conference on Enterprise Information Systems

360

Table 1: Applied search terms for Information Collection.

Database Search string

Scopus TITLE((”Data Science” OR ”Big Data” OR ”Data Mining” OR analytics) AND (project OR manage-

ment OR method* OR framework OR process OR model OR cycle)) AND TITLE-ABS-KEY ((”best

practice” OR success OR pattern) AND project)

SpringerLink 1) Title: ”data science”, All the words: project, Exact phrase: best practice

2) Title: ”data science”, All the words: project, One word: success

3) Title: ”data science”, Exact phrase: project management

search group agreed to include articles that describe

approaches or best practices for the execution and

management of DS projects. Publications with an un-

ﬁtting thematic focus, such as particular DS applica-

tion scenarios and predominant coverage of techni-

cal questions (e.g., algorithms), were excluded. Each

group member was assigned a literature subset for re-

view.

Information Review: The obtained material base

was ﬁltered through multiple stages based on the in-

clusion and exclusion criteria. After the title assess-

ment and removal of duplicates, 96 papers remained.

Following the abstract evaluation, 38 articles were left

in the pool. The full examination led to the removal of

an additional 19 papers, resulting in an intermediate

result set of 19 contributions directly relevant to suc-

cessful approaches and best practices in DS project

management. To expand the literature base for pattern

writing, a backward and forward search (Webster and

Watson, 2002) was also performed. This step added

13 more papers (32 in total). Next, a concept matrix

was utilized to capture and analyze the common top-

ics covered in the works to facilitate the extraction of

the relevant information for the patterns.

5.2 Pattern Authoring

Pattern Language Design: Based on the introduced

methodology for pattern construction, the research

group employed the prescribed pattern structure, con-

sisting of the sections name, problem, context, chal-

lenges, solution, results, and links (see Table 3). In

the Pattern Writing step, these sections are completed

based on the literature base.

Primitive Deﬁnition: The design decisions made

during Information Format Design were reviewed and

largely conﬁrmed. Next to the concept matrix used to

group the obtained information from the literature, the

authors further agreed to use the joint Citavi reposi-

tory for making more detailed annotations of impor-

tant information within the included material since

this can alleviate the pattern writing process.

Composition Language Deﬁnition: In this phase, it

was determined to position patterns in the context of

DS lifecycle stages proposed by (Haertel et al., 2022)

to better clarify the applicability of the created pat-

terns. Because of the focus on DS project manage-

ment best practices, it was expected that most patterns

would address the phase of Business Understanding.

Pattern Writing: Following the described approach,

the sections of the patterns are completed in a spe-

ciﬁc order and as brieﬂy as possible. Because of the

page limitations, this step is described based on one

example pattern (”Alignment of Expectations”) that

was created during this process. The full pattern is

depicted below in Table 3 and is written based on

the inputs derived from the articles of (Cato et al.,

2015; G

okay et al., 2023; Martinez et al., 2021b;

Saltz and Shamshurin, 2016; Soukaina et al., 2019;

Sun et al., 2018; Varela and Domingues, 2021; Yeoh

and Koronios, 2010; Yeoh and Popovi

c, 2016; Schulz

et al., 2020). Documentation begins with the solu-

tion of the pattern. An alignment regarding the poten-

tial of the to-be-developed DS application is required

to raise awareness for realistic DS project expecta-

tions. This involves the project team, management,

domain users, and other stakeholders. A situation as-

sessment is needed to evaluate the feasibility of the

set objectives. Based on similar problems, relevant

resources (e.g., data, budget, skills) and their avail-

ability are discussed. Afterward, the problem section

is elaborated to outline the relevance of the solution.

Because of the data focus, DS projects have an ex-

plorative nature, which increases the difﬁculty of es-

tablishing goals and timelines. Additionally, manage-

ment and users tend to have high expectations in DS

applications. The results are written next to clarify

the impact of the solution. This leads to a joint un-

derstanding of suitable expectations and the roughly

required resources in the project. Based on the sit-

uation assessment, conﬁdence is established regard-

ing the feasibility of the project and its added value

for the organization. Challenges in the context of this

pattern mainly relate to the resources, especially data.

A signiﬁcant challenge in DS undertakings is data ac-

cess. Additionally, the data exploration might reveal

the unsuitability of the available data for achieving the

initially set goals. Hence, a modiﬁcation of the expec-

tations could be necessary. This also applies to other

resources like computing infrastructure or personnel.

A Methodology for Constructing Patterns for the Management of Data Science Projects

361

Table 2: Abstracts for the created patterns.

Pattern name Summary

Alignment of Expectations Coordination of all stakeholders to sensitize regarding spe-

ciﬁc requirements, relevant resources, and challenges of the DS

project.

Involvement of Senior Management Upper management support, encouragement, and guidance are

essential for the successful execution of data science projects,

considering their distinctive demands and uncertainties.

Strategic Alignment of the Project By aligning with the organizational strategy, the project enables

the generation of valuable outcomes for the organization.

Scope The project scope is carefully derived and maintained to facili-

tate the implementation of data science projects.

Process Organization Processes are deﬁned and established to facilitate controlled and

targeted project execution, meeting project challenges.

Implementation of Change Management DS projects requires a corresponding willingness and accep-

tance of changes that need to be incorporated development pro-

cess.

Team Composition Forming a team with members from different areas is crucial for

managing DS projects, requiring comprehensive competencies

and skills.

Project Team Competencies Various technical and social competencies are required to man-

age aspects and requirements of DS projects.

Team Management Promoting high productivity and cooperation among team

members through effective leadership and coordination for suc-

cessful project completion.

Ensuring Data Security To ensure data security in data processing, security measures

are integrated into project infrastructure and processes.

IT Infrastructure To enhance productivity in the project, the IT infrastructure has

to be set up under consideration of the unique requirements of

the DS project.

Ensuring Data Quality To be able to achieve the business objectives, suitable data must

be integrated, prepared, and monitored to ensure high data qual-

ity.

Creation and Maintenance of Documentation Documentation provides access to project procedures and (DS)

results.

Project Performance Monitoring Continuous monitoring of applications and processes enables

efﬁcient project implementation and infrastructure, facilitating

project success.

Finally, the context of the pattern is described, which

constitutes a speciﬁcation of the problem to detail the

circumstances under which it arises. Finally, the sec-

tion on the links to related patterns is ﬁlled out and

brieﬂy elaborated (see Table 3). The Pattern Writ-

ing step is repeated multiple times to repeatedly re-

vise and harmonize the sections and patterns. Using

this procedure with the obtained material base, a total

of 14 patterns for DS project management best prac-

tices were created. These are summarized in Table

2. The repeated occurrence and observation of identi-

ﬁed solutions in the literature were signiﬁcant for the

formation of the patterns.

Pattern Language Revision: The components of the

derived patterns were subject to multiple joint revi-

sions within the research group. A further evaluation,

including external experts, is planned to obtain further

insights regarding the usefulness and completeness of

the DS patterns for actual application.

5.3 Pattern Application

As the application of the developed patterns in prac-

tice in a DS project is still pending, this phase has only

been partially completed so far, and thus, intermediate

results are reported.

Pattern Search and Recommendation: This ﬁrst

activity focuses on improved search and navigation

through the patterns. Therefore, each pattern was

summarized in a sentence. The summary for the pat-

tern ”Alignment of Expectations” is shown in Table

2. Furthermore, the links between the patterns were

visually highlighted with the means of a cross matrix.

Pattern-based Solution Design: The pattern lan-

guage notation was expanded with an ”Example” sec-

tion to facilitate usability, using inputs from the ma-

terial base. The result for the example pattern can be

traced in the full notation in Table 3.

Reﬁnement of the Solution Design: As the scope

ICEIS 2024 - 26th International Conference on Enterprise Information Systems

362

Table 3: Pattern Alignment of Expectations.

Name Alignment of Expectations

Problem The explorative nature of DS projects increases the difﬁculty of establishing goals and timelines that

conﬁrm with expectations of management and domain users.

Context DS project expectations are frequently not realized. Oftentimes, possibilities and results strongly de-

pend on available resources, data access, and quality.

Challenges The availability of resources impact the project outcome. A signiﬁcant challenge in DS undertakings

is data access. Additionally, the data exploration might reveal the unsuitability of the available data for

achieving the business objectives. Hence, because of the inherent risks and uncertainties in DS projects,

ﬂexibility regarding the modiﬁcation of the expectations might be necessary. This also applies to other

resources like computing infrastructure or personnel.

Solution The project team, management, domain users, and other stakeholders perform an alignment regarding

the potential and limitations of the envisioned DS application. A situation assessment evaluates the

feasibility of the set objectives and their added value for the organization. Based on detected similar

problems and the corresponding solutions, the relevant resources (e.g., data, budget, competencies) and

their availability are discussed.

Result Development of a joint understanding of appropriate expectations and the approximately required re-

sources and timelines. Based on the situation assessment, conﬁdence is established regarding the feasi-

bility of the DS project and its added value for the organization.

Links Project expectations result from the Strategic Alignment of the Project and Involvement of Senior Man-

agement. Objectives have to be aligned with requirements to the project execution, including the IT

Infrastructure and Team Composition to determine a realistic Scope. Moreover, expectations are de-

ﬁned regarding Project Team Competencies to complete the project tasks. During project execution,

based on Project Performance Monitoring new or revised requirements and goals can arise.

Example This pattern can be assigned to Business Understanding, which is a common phase in various DS

process models. Here, the project circumstances are communicated with involved stakeholder groups

to elaborate opportunities, requirements, and functionalities of the DS application (Schulz et al., 2020).

A feasibility study can be used to evaluate the likelihood of fulﬁlling project requirements and objectives

(Schulz et al., 2020).

of the Coverage Consideration and resulting Informa-

tion Collection was set on general best practices and

successful approaches for DS project management, no

further limitations for application were deﬁned in this

step. However, based on the feedback of the revision

with DS practitioners, this is subject to change.

Instantiation of the Solution Design: Due to the po-

tential for various changes to the patterns in the future,

the patterns are stored in a repository with shared ac-

cess for each member of the research group. Version

control is enabled to track changes that might be nec-

essary based on the evaluation results and possible ex-

tensions to the material base.

6 CONCLUSION

DS projects suffer from high failure rates (Ven-

tureBeat, 2019), which indicates the need for new

approaches for DS project management (Saltz and

Krasteva, 2022). Therefore, in this work, we applied

the pattern concept to DS since it allows for the struc-

tured summarization of solutions for common prob-

lems in a domain. Using the DSR methodology of

(Peffers et al., 2007), the pattern creation process of

(Fehling et al., 2014), consisting of Pattern Identiﬁca-

tion, Authoring, and Application, was adapted to DS

project management. The functionality of the pro-

posed method was demonstrated by the creation of

patterns for DS project management best practices

from the synthesis of scientiﬁc literature. Accord-

ingly, researchers and practitioners can apply the in-

troduced pattern construction method to synthesize

solutions to frequently occurring issues in the exe-

cution and management of DS undertakings. Never-

theless, the study at hand is subject to certain limi-

tations. Because of page restrictions, the evaluation

within the DSR methodology following the guide-

lines of (Sonnenberg and vom Brocke, 2012) was not

covered in this paper. While the ex-ante evaluations

(Eval 1 and 2) were brieﬂy touched upon in the back-

ground, further depth should be provided through lit-

erature reviews and expert interviews. The demon-

stration in Section 5 showed initial tendencies toward

the artifact’s general applicability, but a detailed as-

sessment is still pending. Additionally, further case

studies are needed to conclusively evaluate the use-

fulness of the proposed method and determine areas

for improvement. For instance, the delimitation be-

tween certain activities (e.g., Primitive Deﬁnition and

Composition Language Design) was not always clear,

indicating the potential for consolidating some of the

(sub)stages.

A Methodology for Constructing Patterns for the Management of Data Science Projects

363

REFERENCES

Aichele, C. and Sch

onberger, M. (2014). IT-

Projektmanagement: Efﬁziente Einf

uhrung in

das Management von Projekten. SpringerLink

ucher. Springer Vieweg, Berlin.

Alexander, C. (1979). The timeless way of building. Oxford

Univ. Pr, New York.

Alexander, C., Ishikawa, S., and Silverstein, M. (1977). A

Pattern Language: Towns, Buildings, Construction.

Buschmann, F. (1996). Pattern-oriented Software Architec-

ture: A System of Patterns. Wiley, Chichester.

Cao, L. (2017). Data science: Challenges and directions.

Communications of the ACM, 60(8):59–68.

Cao, L. (2018). Data science: A comprehensive overview.

ACM Computing Surveys, 50(3):1–42.

Cato, P., Golzer, P., and Demmelhuber, W. (2015). An in-

vestigation into the implementation factors affecting

the success of big data systems. In 2015 11th Interna-

tional Conference on Innovations in Information Tech-

nology (IIT), pages 134–139. IEEE.

Chang, W. and Grady, N. (2019). Nist big data interoper-

ability framework: Volume 1, deﬁnitions.

Coplien, J. O. (2000). Software Patterns. SIGS Books &

multimedia, New York.

Coplien, J. O. and Harrison, N. B. (2005). Organiza-

tional Patterns of Agile Software Development. Pear-

son Prentice Hall, Upper Saddle River.

Das, M., Cui, R., Campbell, D. R., Agrawal, G., and Ram-

nath, R. (2015). Towards methods for systematic re-

search on big data. 2015 IEEE International Confer-

ence on Big Data, pages 2072–2081.

de Medeiros, M. M., Hoppen, N., and Mac¸ada, A. C. G.

(2020). Data science for business: beneﬁts, challenges

and opportunities. The Bottom Line.

Eisend, M. and Kuß, A. (2023). Grundlagen em-

pirischer Forschung: Zur Methodologie in der Be-

triebswirtschaftslehre. Springer Fachmedien Wies-

baden and Imprint Springer Gabler, Wiesbaden, 3.,

uberarbeitete aufage edition.

Fayyad, U., Piatetsky-Shapiro, G., and Smyth, P. (1996).

The kdd process for extracting useful knowledge from

volumes of data. Communications of the ACM,

39(11):27–34.

Fehling, C., Barzen, J., Breitenb

ucher, U., and Leymann, F.

(2014). A process for pattern identiﬁcation, authoring,

and application. In Eloranta, V.-P. and van Heesch, U.,

editors, Proceedings of the 19th European Conference

on Pattern Languages of Programs, pages 1–9, New

York, NY, USA. ACM.

okay, G. T., Nazlıel, K., S¸ener, U., G

okalp, E., G

okalp,

M. O., Genc¸al, N., Da

gdas¸, G., and Eren, P. E.

(2023). What drives success in data science projects:

A taxonomy of antecedents. In Computational Intel-

ligence, Data Analytics and Applications, pages 448–

462, Cham. Springer International Publishing.

unther, A. and Knote, R. (2017). How to design patterns in

is research – a state-of-the-art analysis. Proceedings

der 13. Internationalen Tagung Wirtschaftsinformatik

(WI 2017), pages 1393–1404.

Haertel, C., Pohl, M., Nahhas, A., Staegemann, D., and Tur-

owski, K. (2022). Toward a lifecycle for data science:

A literature review of data science process models.

PACIS 2022 Proceedings.

Harrison, N. B. (2003). Advanced pattern writing: Patterns

for experienced pattern authors. EuroPLoP, pages

809–828.

Heinrich, L. J. (1997). Management von Informatik-

Projekten. R. Oldenbourg Verlag M

unchen Wien.

Hevner, A. R., March, S. T., and Park, J. (2004). Design

science in information systems research. MIS Quar-

terly.

Iba, T. and Isaku, T. (2012). Holistic pattern-mining pat-

terns: A pattern language for pattern mining on a

holistic approach. 19th Pattern Languages of Pro-

grams conference.

Jeble, S., Kumari, S., and Patil, Y. (2018). Role of big data

in decision making. Operations and Supply Chain

Management, Vol. 11(No. 1):36–44.

Manns, M. L. and Rising, L. (2012). Fearless Change: Pat-

terns for Introducing New Ideas. Addison-Wesley.

Martinez, I., Viles, E., and Olaizola, I. G. (2021a). Data

science methodologies: Current challenges and future

approaches. Big Data Research 24.

Martinez, I., Viles, E., and Olaizola, I. G. (2021b). A survey

study of success factors in data science projects. In

2021 IEEE International Conference on Big Data (Big

Data), pages 2313–2318.

Meszaros, G. and Doble, J. (1997). A pattern language for

pattern writing. Pattern languages of program design,

pages 529–574.

Peffers, K., Tuunanen, T., Rothenberger, M. A., and Chat-

terjee, S. (2007). A design science research method-

ology for information systems research. Journal of

Management Information Systems, 24(3):45–77.

PMI (2017). A guide to the project management body of

knowledge (PMBOK guide). Sixth edition edition.

Saltz, J. (2022). Nine questions to evaluate a data science

team’s process: Exploring a big data science team

process evaluation framework via a delphi study. In

2022 IEEE International Conference on Big Data (Big

Data), pages 2667–2672. IEEE.

Saltz, J., Hotz, N., Wild, D., and Stirling, K. (2018). Explor-

ing project management methodologies used within

data science teams. AMCIS 2018.

Saltz, J. S. (2015). The need for new processes, method-

ologies and tools to support big data teams and im-

prove big data project effectiveness. IEEE Interna-

tional Conference on Big Data 2015.

Saltz, J. S. and Hotz, N. (2020). Identifying the most com-

mon frameworks data science teams use to structure

and coordinate their projects. In 2020 IEEE Inter-

national Conference on Big Data (Big Data), pages

2038–2042. IEEE.

Saltz, J. S. and Krasteva, I. (2022). Current approaches for

executing big data science projects - a systematic lit-

erature review. PeerJ Computer Science, 8(e862).

Saltz, J. S. and Shamshurin, I. (2016). Big data team pro-

cess methodologies: A literature review and the iden-

tiﬁcation of key factors for a project’s success. In

ICEIS 2024 - 26th International Conference on Enterprise Information Systems

364

2016 IEEE International Conference on Big Data (Big

Data), pages 2872–2879.

Schulz, M., Neuhaus, U., Kaufmann, J., Badura, D.,

Kuehnel, S., Badewitz, W., Dann, D., Kloker, S.,

Alekozai, E. M., and Lanquillon, C. (2020). Intro-

ducing dasc-pm: A data science process model. ACIS

2020.

Sonnenberg, C. and vom Brocke, J. (2012). Evaluations in

the science of the artiﬁcial – reconsidering the build-

evaluate pattern in design science research. In Pef-

fers, K., Rothenberger, M., and Kuechler, B., ed-

itors, Design science research in information sys-

tems, SpringerLink B

ucher, pages 381–397, Berlin.

Springer.

Soukaina, M., Anoun, H., Ridouani, M., and Hassouni, L.,

editors (2019). A study of the factors and methodolo-

gies to drive successfully a big data project.

Sun, S., Cegielski, C. G., Jia, L., and Hall, D. J. (2018).

Understanding the factors affecting the organizational

adoption of big data. Journal of Computer Informa-

tion Systems, 58(3):193–203.

Varela, C. and Domingues, L. (2021). Risks of data science

projects - a delphi study. pages 982–989.

VentureBeat (2019). Why do 87% of data science projects

never make it into production?

vom Brocke, J., Simons, A., Niehaves, B., Reimer, K., Plat-

tfaut, R., and Cleven, A. (2009). Reconstructing the

giant: On the importance of rigour in documenting the

literature search process. ECIS 2009.

Wack, J. (2007). Risikomanagement f

ur IT-Projekte:

Zugl.: Hamburg, Univ., Diss., 2006, volume 54

of Betriebswirtschaftliche Forschung zur Un-

ternehmensf

uhrung. Dt. Univ.-Verl., Wiesbaden, 1.

auﬂ. edition.

Webster, J. and Watson, R. T. (2002). Analyzing the past to

prepare for the future: Writing a literature review. MIS

Quarterly, Vol. 26, No. 2 (Jun. 2002), pages 13–23.

Wellhausen, T. and Fiesser, A. (2011). How to write a pat-

tern? a rough guide for ﬁrst-time pattern authors. Pro-

ceedings of the 16th European Conference on Pattern

Languages of Programs.

Wieczorrek, H. W. and Mertens, P. (2007). Manage-

ment von IT-Projekten: Von der Planung zur Real-

isierung. Xpert.press. Springer-Verlag Berlin Heidel-

berg, Berlin, Heidelberg, 2.,

uberarbeitete und erweit-

erte auﬂage edition.

Yeoh, W. and Koronios, A. (2010). Critical success factors

for business intelligence systems. Journal of Com-

puter Information Systems, Vol. 50(No. 3):23–32.

Yeoh, W. and Popovi

c, A. (2016). Extending the under-

standing of critical success factors for implementing

business intelligence systems. Journal of the As-

sociation for Information Science and Technology,

67(1):134–147.

A Methodology for Constructing Patterns for the Management of Data Science Projects

365