Governance and Longevity of Architecturally Embedded Applications

ABSTRACT

This study explores the relationship between the governance of architecturally embedded applications and their longevity in organizational use. Using a contingency-fit logic, it posits that alignment between application governance (distribution of decision rights) and architectural contingencies contributes to sustained durations of organizational use. Architectural contingencies include internal application modularity, application inter-connectedness, and governance of surrounding application clusters. Evidence for the governance-architecture fit hypothesis is derived from analyzing temporally ordered data about applications (n = 225, n = 498, n = 193) from organizations in the media, banking, and utilities sectors over five years. Using a two-step approach with ordinal regression, along with Cox proportional hazards and logistic regression models, our findings indicate that a fit between observed governance and the governance induced by architectural contingencies reduces the likelihood of application decommissioning. The study advances the fields of information systems (IS) governance and architecture by offering a micro-level perspective on the longevity of organizational IS use while also shedding light on the importance of governance choices aligned with unique architectural application properties.

KEYWORDS:

• Digital infrastructure
• Application governance
• Application longevity
• Application discontinuance
• Information-systems architecture
• Modular systems
• Contingency fit
• Survival analysis

Disclosure statement

No potential conflict of interest was reported by the author(s).

Supplementary Material

Supplemental data for this article can be accessed online at https://doi.org/10.1080/07421222.2023.2301169

Notes

^1. Anonymized due to confidentiality agreements with these organizations.

^2. Utilities did not collect intermediate datasets and discontinued/pivoted its enterprise architecture efforts after 2015.

^3. See Online Appendix B for a detailed coding guide.

^4. To account for survivor bias [18] (the problem of focusing on applications that have endured, while neglecting those that have not), we included age as a separate control variable in the regressions. This information was extracted from a separate, reported field in the datasets called the productive date.

^5. Additional exploratory discussions with five application owners, one IT strategist, and one enterprise architect confirmed the relevance of the notion of governance-architecture fits when confronted with the findings of the analysis. Across all companies, we established close rapport with the architecture group to refine, validate, and challenge the findings.

^6. A prominent example of modified legacy software is the Ada-coded US Air Traffic Control system, around which the Federal Aviation Administration (FAA) has built special wrappers and APIs to enable it to interface with more modern external applications [69].

^7. Because there are 0 observations, the standard log-transformation is not available. To test robustness, we also experimented with $\mathit{ln}\left(\mathit{X}+\mathit{k}\right),$ $0.1\le \mathit{k}\le 10$ , to inter-connectedness and the results are robust.

^8. For Media and Utilities, the number of using business units served as a proxy for the scope of use because no other user data were available. For MEDIA, we imputed missing values (n = 5) by drawing on related information, such as the business processes in which the application was used, and validated these imputed values with a key informant. For BANKING, we imputed a low number of missing values (n = 18) with the mean number of users and validated this procedure with a company informant. For UTILITIES, we counted the number of distinct units in a comma-separated field called ‘using business units’ in the raw data. There were no missing values.

^9. Note that this is a simplification that holds true for larger values of inter-connectedness. See Online Appendix C for details on the interpretation of coefficients in the log-arcsinh model.

^10. Banking had followed a consolidation policy for its multiple shadow IT applications within the observation period. We had controlled for the bias of this policy on longevity through a separate binary control and only show non-shadow IT applications in Figure 4.

^11. An illustrative analysis of the decommissioned applications at Media did not reveal a pattern of reverse causality. Instead, it suggested that application governance was stable, whereas diminishing support in the organization was sometimes expressed in the reduced number of business units using the application. In some cases, the additional analysis also revealed that investments were made in an application to prepare for decommissioning.

Additional information

Notes on contributors

Karl Akbari

Karl Akbari is an Assistant Professor at the Department of Business Administration at National Taiwan University of Science and Technology. He received his Ph.D. from the University of Vienna. Dr. Akbari’s research interests are pricing issues using econometric modeling and experimental methods. His research has been published in Marketing Science, Review of Managerial Science, and Schmalenbach Journal of Business Research, among other outlets.

Daniel Fürstenau

Daniel Fürstenau is an Associate Professor in Information Systems at the IT University of Copenhagen. He received his Ph.D. from Freie Universität Berlin. Dr. Fürstenau is an affiliated member of the Einstein Center Digital Future and a digital health researcher at Charité. His research interests include IT management, digital health, and AI-driven innovation. His 2019 Information Systems Research paper received the AoM Organizational Communication and Information Systems best paper runner-up award.

Till J. Winkler

Till J. Winkler is Chair Professor of Information Management at University of Hagen, Germany, and affiliated Associate Professor at the Copenhagen Business School, Denmark. His research on the governance of information technology, service management, and digital health has appeared in Journal of Management Information Systems, Journal of the Association for Information Systems, Journal of Strategic Information Systems, MIS Quarterly Executive, and others. Dr. Winkler is recipient of the Early Career Award of the Association for Information Systems.