The Impact of Resource Allocation on the Machine Learning Lifecycle - Bridging the Gap between Software Engineering and Management

Aug 4, 2026

Uploaded by Maeve

An organization’s ability to develop machine learning (ML) applications depends on its available resource base. Without understanding relevant resources and how they shape the ML lifecycle, organizations risk inefficient allocations and monopolization tendencies. The study presents a framework that interweaves resources with the procedural and technical dependencies across the ML lifecycle. Using design science research, literature review, and interviews, it extends prior work by adding previously under-discussed resources and identifying six direct and indirect resource effects and contextual properties.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages17

Reading Time43 min

AccessFree to download

Introduction

Research Motivation and Knowledge Gaps

Framework Overview and Research Approach

Method

Design Science Research Paradigm

Literature Review and Interview Study

Results

Resource Effects on the ML Lifecycle

Contextual Properties and Practical Value

Bus Inf Syst Eng 66(2):203–219 (2024) [https://doi.org/10.1007/s12599-023-00842-7](https://doi.org/10.1007/s12599-023-00842-7) RESEARCH PAPER The Impact of Resource Allocation on the Machine Learning Lifecycle Bridging the Gap between Software Engineering and Management Sebastian Duda • Peter Hofmann • Nils Urbach • Fabiane V¨olter • Amelie Zwickel Received: 14 May 2022 / Accepted: 11 September 2023 / Published online: 14 November 2023 􀀂 The Author(s) 2023 Abstract An organization’s ability to develop Machine Learning (ML) applications depends on its available resource base. Without awareness and understanding of all relevant resources as well as their impact on the ML lifecycle, we risk inefﬁcient allocations as well as missing monopolization tendencies. To counteract these risks, the study develops a framework that interweaves the relevant resources with the procedural and technical dependencies within the ML lifecycle. To rigorously develop and evaluate this framework the paper follows the Design Science Research paradigm and builds on a literature review and an interview study. In doing so, it bridges the gap between the software engineering and management perspective to advance the ML management discourse. The results extend the literature by introducing not yet discussed but relevant resources, describing six direct and indirect effects of resources on the ML lifecycle, and revealing the resources’contextual properties. Furthermore, the framework is useful Accepted after two revisions by Stefan Stieglitz. S. Duda 􀀂 P. Hofmann 􀀂 N. Urbach 􀀂 F. V¨olter (&) Fraunhofer Institute for Applied Information Technology FIT Branch Business & Information Systems Engineering, Wittelsbacherring 10, 95444 Bayreuth, Germany e-mail: fabiane.voelter@ﬁ[t.fraunhofer.de](t.fraunhofer.de) S. Duda e-mail: sebastian.duda@ﬁ[t.fraunhofer.de](t.fraunhofer.de) P. Hofmann e-mail: [p.hofmann@appliedai.de](p.hofmann@appliedai.de) N. Urbach e-mail: nils.urbach@ﬁ[m-rc.de](m-rc.de) [S. Duda](S. Duda) 􀀂 P. Hofmann 􀀂 F. V¨olter FIM Research Center, University of Bayreuth, Wittelsbacher Ring 10, 95444 Bayreuth, Germany in practice to support organizational decision-making and contextualize monopolization tendencies. Keywords ML management 􀀂 Machine learning lifecycle 􀀂 Artiﬁcial intelligence 􀀂 Resource-based view 􀀂 Design science research 1 Introduction The current momentum in Machine Learning (ML) development and adoption is making companies to reﬂecton their positioning and the associated conﬁguration of their resource bases. Some companies try to stand out with leading ML models (e.g., OpenAI with ChatGPT) or to capture the market with resource-integrating service platform offerings (Geske et al. 2021) . Others heavily invest in data collection as data is a critical resource when training an ML model (Mikalef and Gupta 2021) . We can even P. Hofmann appliedAI Initiative GmbH, Freddie-Mercury-Straße 5, 80797 Munich, Germany N. Urbach Frankfurt University of Applied Sciences, Nibelungenplatz 1, 60318 Frankfurt am Main, Germany A. Zwickel University of Bayreuth, Universit¨atsstraße 30, 95445 Bayreuth, Germany e-mail: [amelie-zwickel@hotmail.de](amelie-zwickel@hotmail.de) observe that infrastructure resources are no longer necessarily a commodity, but companies are scrambling to develop and deploy speciﬁc hardware, such as tensor processing units (Jouppi et al. 2018) . As with other digital technologies, companies face once again the challenge of ﬁnding their place in the market and, correspondingly, conﬁguring their resource base. Above all, ML applications are most notable in that their performance evolution is non-deterministic because ML ‘‘is a subﬁeld of AI, which tries to acquire knowledge by extracting patterns from raw data and solve some problems using this knowledge’’ (Giray 2021, p. 2) . Recent research has acknowledged ML development’s speciﬁcity (Giray 2021; Iansiti and Lakhani 2020

FAQs

Why does resource allocation matter for the machine learning lifecycle?

Developing ML applications depends on an organization’s resource base. When organizations lack awareness of relevant resources and their lifecycle effects, allocations can become inefficient and interdependencies may be mishandled.

What does the proposed framework connect?

The framework interweaves relevant resources with procedural and technical dependencies within the ML lifecycle, bridging software engineering and management perspectives.

How was the framework developed and evaluated?

The study applies the design science research paradigm, building on a literature review and an interview study to develop and rigorously evaluate the framework.

Aug 4, 2026

Uploaded by Maeve