What is digital transformation? A survey on the perceptions of decision-makers in business

The availability and increasing use of digital technologies have led to various and sometimes disruptive changes in organizations and their environments (Gurbaxani and Dunkle 2019; Robey et al. 2013; Yayla and Lei 2020). In this context, the term “digital transformation” first emerged in academic research in the early 2000s, gained relevance in the early 2010s, and has continued to increase in significance in the 2020s (Hanelt et al. 2021). The current state of DT research can best be described as fragmented, spanning different disciplines and including different views and research approaches. This is illustrated by the numerous reviews published in leading journals, which have viewed DT research from procedural (Vial 2019), organizational change (Hanelt et al. 2021), innovation and entrepreneurship (Nadkarni and Prügl 2021), and multidisciplinary (Verhoef et al. 2021), perspectives. Table 1 summarizes these major reviews and specifies their conceptual contributions to the DT literature.

Based on the insights presented in Table 1, we drew two major conclusions regarding DT, which then informed the design of our research approach and data analysis.

First, the increasing diffusion and use of digital technologies is the central factor motivating the need for DT.¹ The increasing ubiquity and deployment of digital technologies has led to the changes in market environments, customer expectations, global value networks, and rules of competition that define our current environment (Hanelt et al. 2021; Nadkarni and Prügl 2021; Verhoef et al. 2021; Vial 2019).

Second, DT refers to changes in value creation at various levels of analysis (i.e., individual, organizational, societal), but the organizational level is dominant.² The papers summarized in Table 1 provide insights into the aspects of value creation influenced and altered through organizational DT. As an example, Verhoef et al. (2021) understand a change in an organization’s business model as the distinguishing aspect of DT compared to the related concepts digitization and digitalization. This view, however, contrasts with other perspectives, which also include further alterations affecting organizational value creation as DT, such as changes in organizational capabilities, structures, and processes (Nadkarni and Prügl 2021), general organizational change (Hanelt et al. 2021), and other significant changes to organizational properties (Vial 2019).

In contrast to the reviews presented in Table 1, some authors even challenge the assumption that DT should be considered a distinguishable and novel concept; for example, Chen and King (2022) write that “DT might be the latest in a succession of names, tied to the context of their time, signifying the increasing importance of information technology” (p. 403). This radical view is related to the challenge to strike a balance between avoiding the reinvention of the wheel and capitalizing on opportunities (e.g., those related to research funding) when dealing with new phenomena—or, at least, developments that appear new at first glance. Numerous scientific findings (e.g., Chen and King 2022; Mertens and Wiener 2018, Steininger et al. 2009) indicate that the IS field has a tendency to deal with fad-like topics that resist sustainable nomenclatures and conceptualizations of the discipline’s phenomena. Considering this fact, it is even more important to deal with the nature of DT and how it is different from related phenomena.³

The concepts of organizational transformation and IT-enabled organizational transformation are commonly identified as the conceptual roots of DT (Ismail et al. 2017; Nadkarni and Prügl 2021; Vial 2019; Wessel et al. 2021). The concept of IT-enabled organizational transformation (ITOT) received increasing attention in media outlets focusing on business in the early 1990s and was defined by the Corporate Renewing Centre at INSEAD as a “fundamental change in organizational logic, which resulted in or was caused by a fundamental shift in behaviors” (Muzyka et al. 1995, p. 348). This highlights that IT’s impact on organizations was already significant several decades ago, with practitioners describing the information revolution as one of three central drivers of organizational transformation (Muzyka et al. 1995). As outlined in reviews by Wessel et al. (2021) and Vom Brocke et al. (2021), following studies on the transformative impact of enterprise resource planning (ERP) systems in the 1990s, the IS community developed the concept of ITOT to explore the prominent role of digital technologies as a driver of organizational transformation. Within the broad body of knowledge accumulated under the ITOT term, Venkatraman’s (1994) seminal framework was among the first to provide a clear conceptualization of IT-enabled business transformation. This framework is divided into two dimensions—degree of business transformation and range of potential benefits—and proposes five levels of IT-enabled business transformation (Fig. 1).

Venkatraman (1994) describes the characteristics of each level with corresponding examples. The two levels denoting a lower degree of business transformation are categorized as evolutionary and the three levels with a higher degree as revolutionary (Venkatraman 1994). Specifically, higher levels of transformation (Fig. 1, from left lower corner to right upper corner) are characterized by a broader scope of business transformation. Consequently, Mertens et al. (2017) indicate that the revolutionary levels should be considered as DT, while the evolutionary levels should not. What follows is that according to Mertens et al. (2017) the model can be used to derive DT from specific levels of Venkatraman’s (1994) framework. It should be noted, however, that this does not imply that ITOT and DT are equivalent. Venkatraman’s framework is one of the most cited papers in the field of IT-enabled business transformation (Vom Brocke et al. 2021). Although the framework was published around three decades ago, its ongoing relevance is demonstrated by its recent use as an analytical tool (Vom Brocke et al. 2021) and a methodological foundation (Blanka et al. 2022), as well as by its inclusion in recent editions of seminal IS textbooks (Hess 2022; Mertens et al. 2017).

While the discussion in the literature has been primarily informed by conceptual theorizations and extensive studies of relevant literature (Table 1), we propose that empirically studying practitioners’ perspectives of what constitutes DT can provide fresh insights, especially through developing a new, empirically grounded perspective on the defining characteristics of DT. In what follows, we deem a purely conceptual understanding of DT to be insufficient, and while our survey study was informed by the existing literature reviewed in Sect. 2, we intend to complement prior papers with empirical data. Specifically, we collected survey data on two aspects: (1) the congruence between current DT definitions in the scientific literature and their understanding by business decision-makers, and (2) the evaluation of concrete DT examples which we identified in the scientific literature by business decision-makers. Generally, definitions constitute an essential foundation for progress in a specific research domain and are crucial for communicating meaning among researchers, and between researchers and practitioners (Hodgson 2019). However, the highly dynamic nature of many IS phenomena inherently poses challenges with regard to providing stable conceptualizations for emerging phenomena such as DT. Importantly, the literature provides a plethora of definitions for DT, as illustrated by Gong and Ribiere’s (2021) recent analysis identifying 146 definitions of DT in different publication outlets. In line with the results of previous analyses of DT definitions (e.g., Vial 2019), Gong and Ribiere (2021) indicate that existing definitions may not be scientifically sound. Indeed, one major problem in defining DT lies in the broadening of definitional terms, a phenomenon known as conceptual stretching (Osigweh 1989). When it comes to DT, this leads to the problem that almost every phenomenon related to digital technologies with even minimal consequences—and hence little potential for change—is considered a DT phenomenon. Tellingly, Gong and Ribiere (2021) argue:

This conceptual ambiguity creates another challenge: distinguishing DT from related phenomena. As a solution to this challenge, Vial (2019) and Gong and Ribiere (2021) used a semantic decomposition process originally described by Akmajian et al. (2017) to identify the semantic primitives of existing definitions. Next, we describe our research approach and methodology in detail.

A semantic primitive is a term or concept innately understood by most people which cannot be conceptualized in simpler ways (Goddard and Wierzbicka 2014). After establishing four primitives (i.e., target entity, scope, means, expected outcome), Vial (2019) defined DT as “a process that aims to improve an entity by triggering significant changes to its properties through combinations of information, computing, communication, and connectivity technologies” (p. 118). Gong and Ribiere (2021) developed six primitives (i.e., nature, scope, target entity, means, expected outcome, impact) and further evaluated their description of DT through expert interviews, ultimately developing the following DT definition: “a fundamental change process, enabled by the innovative use of digital technologies accompanied by the strategic leverage of key resources and capabilities, aiming to radically improve an entity” (p. 12). These semantic primitives provide a meta-structure of DT definitions and thus form a frame of reference for different approaches to distinguishing DT from related phenomena. Table 2 summarizes the meanings of the identified primitives.

Regarding the primitive nature, DT is typically understood as an instance of organizational change. The nature of this change has been framed differently—for example, as transformation (Demirkan et al. 2016), process (Verhoef et al. 2021; Vial 2019), or unspecified change (Hess et al. 2016; Legner et al. 2017). The primitive scope reflects the extent of change implemented by an organization. If used in a DT definition, it tends to indicate a major change rather than a minor one. This is best illustrated by authors imposing the necessity for changes to be “fundamental” (Gong and Ribiere 2021, p. 9), “profound” (Demirkan et al. 2016, p. 14), “significant” (Vial 2019, p. 118), or “radical” (Westerman et al. 2011, p. 5; Westerman et al. 2014, p. 108; Karagiannaki et al. 2017, p. 4). Regarding the primitive target entity, Gong and Ribiere (2021) identified individual organizations, business networks, and industries as common target entities of DT definitions. Various definitions include organizations as target entities (Kane et al. 2015; Verhoef et al. 2021; Wessel et al. 2021), while others either do not specify a target entity (Andriole 2017; Legner et al. 2017) or intentionally use the term “entity” and therefore deliberately remain abstract (Gong and Ribiere 2021; Vial 2019).

Identifying DT through the primitive of means is a common aspect of DT definitions. While Vial’s (2019) understanding of this primitive mainly focuses on digital technologies as means, Gong and Ribiere’s (2021) primitive adopts a broader approach reflecting the strategies, resources, and capabilities involved in creating the change. This demarcation is commonly based on the technological constructs used (e.g., certain sets of technology categories; Fitzgerald et al. 2013; Vial 2019) or more abstract technology constructs, such as (new) digital technologies (Fitzgerald et al. 2013; Verhoef et al. 2021; Wessel et al. 2021) or IT (Legner et al. 2017; Li et al. 2018). Further demarcations have also been made based on the capabilities involved, such as innovative use of digital technologies (Gong and Ribiere 2021) or adoption of disruptive technologies (Nadkarni and Prügl 2021).

Various authors have also conceptualized DT based on the primitives expected outcome and impact. In addition to vague definitions of expected outcomes, such as “improvements” (Fitzgerald et al. 2013, p. 2; Vial 2019, p. 120; Westerman et al. 2011, p. 37), more concrete outcomes have also been described, including “(re)defining an organization’s value proposition” (Wessel et al. 2021, p. 102), creating “new business models” (Fitzgerald et al. 2013, p. 2; Verhoef et al. 2021, p. 891), “changed products or organizational structures or […] automation of processes” (Hess et al. 2016, p. 124), and changed tasks (Legner et al. 2017). Wessel et al. (2021) identify the emergence of a new organizational identity as a defining result of DT compared with ITOT. Other approaches are, again, more vague in their identified impacts—for example, “appropriat[ing] more value for the firm” (Verhoef et al. 2021, p. 889) or ensuring effective competition in a digital world (Kane et al. 2015). Comparison of the primitives in existing definitions reveals significant differences regarding the necessity of means, expected outcome, and impact to distinguish between mere organizational change and DT.

To investigate the congruence of understanding of DT among business decision-makers with that of academics based on (1) definitions and (2) real-world examples, we took a three-stage approach when developing our survey (Fig. 2). First, we derived a set of candidate definitions and examples from the academic literature. Second, we reduced our set of candidate definitions and examples using a pre-study and expert interviews, aiming to develop a diverse set of distinct definitions and a set of contextually complete examples for DT spread evenly over the five levels of IT-enabled business transformation presented by Venkatraman (1994; Fig. 1). Third, we delivered our main survey study using our final set of definitions and examples (Tables 3 and 4).

We derived our candidate set of definitions based on Vial (2019), who analyzed a total of 282 works and thus represents the most thorough and broad review of the DT field to date. We established the final set of definitions following a collaborative process aimed at identifying a set of diverse definitory approaches that did not exceed practical limits (i.e., the time required to respond to the questions). This process included the authors of this paper and six additional people with significant expertise in the DT field. The final set (see Table 3) consisted of definitions derived from four sources: eight definitions identified by Vial (2019) as well as Vial’s (2019) unified definition of DT; Venkatraman’s (1994) definitions of the five levels of IT-enabled business transformation; and two definitions from recently published papers that focus on conceptually clarifying the DT concept (Verhoef et al. 2021; Wessel et al. 2021).

Regarding our candidate set of real-world DT examples, we analyzed all papers published between January 2015 and June 2020 that cited Venkatraman’s (1994) framework and included real-world examples of DT.⁴ We identified 54 examples distributed across the five levels specified by Venkatraman (1994): (i) localized exploitation: 17; (ii) internal integration: 10; (iii) business process redesign: 7; (iv) business network redesign: 12; and (v) business scope redefinition: 8. Mertens et al. (2017) indicate that levels three to five should be considered as DT, while levels one and two should not. However, because it is a major goal of our study to reveal possible discrepancies in the perception of the DT phenomenon between academia and practice, we included all five levels in our study.

We derived the final set of examples of DT through a structured two-stage process including a pre-study with DT research experts aiming to select a set of five examples of IT-induced change that (i) provided sufficient contextual information to allow for a clear integration into one of the five levels defined by Venkatraman (1994) and (ii) represented a clear example of each level. In the first stage, we checked each example for the inclusion of information about the context, technology used, and expected outcome. This process reduced the number of examples from 54 to 24. In the second stage, six DT experts from research and practice completed an online survey. After showing each expert the definition of each IT-enabled business transformation level (Venkatraman 1994), we tasked them with ranking the remaining examples of the candidate set in terms of how holistically and understandably they represented the respective level. This resulted in the final set of examples presented in Table 4.

Using the definitions and examples presented in Tables 3 and 4, we developed our online survey, which consisted of three question blocks: (1) demographic data, (2) DT examples, and (3) DT definitions. In the first block, we prompted participants to provide information on their age, gender, highest level of education, industry of current employment, position within their organization, and number of years of professional experience. In the second block, we presented the examples to participants in a randomized order, each on a single page, and requested that they indicate via a radio button whether or not each could be classified as an instance of DT. Furthermore, in an open-text field, participants were prompted to provide the text snippets from each example that were most relevant to their decision. In the third block, we presented the definitions to the participants in randomized order, each on a single page. As in the second block, participants were first prompted to indicate via a radio button whether the presented description constituted an appropriate DT definition and were then asked to provide the text snippets from the definition that led to their decision. We deliberately presented the DT examples before the DT definitions to avoid any bias in respondents’ categorization. Data were collected in English in the UK in 2021. We implemented the survey using the online platform SurveyMonkey. The provider allows the survey to be played out only to participants who meet certain criteria. Our chosen criteria were the following: range of age: 18–99; country: UK; professional position: business decision-maker (i.e., a position including managerial decision-making responsibilities); company size: more than ten employees. An incentive was paid to the participants. In Appendix 1, we show sample screenshots from the second and third block of the survey.

A total of 2156 participants participated in the survey, and 1253 questionnaires were completed in full. Of these, we discarded 724 after a screening process because of data quality issues. First, we excluded persons who filled the free-text fields with meaningless character strings. Second, we also excluded participants who had an unrealistically short completion time (< 4 min).⁵ A possible reason for the high number of participants who had to be excluded could be the relatively high effort involved in the completion of the survey because participants had to give responses via radio buttons and had to use drag and drop to indicate text snippets. Importantly, based on our strict exclusion criteria we secure the quality of our data. The following analyses are based on a final sample of N = 529 business decision-makers.⁶

In our analysis of the survey data, we first focused on the DT definitions. We investigated the distribution of participants’ categorizations of the presented definitions (i.e., whether each was a suitable DT definition) to obtain a clearer picture of participants’ perceptions of the DT definitions. To analyze and group the decisive terms reported, we conducted an in-depth examination of the text snippets in the context of the primitives identified by Vial (2019) and Gong and Ribiere (2021) as presented in Table 2 and consequently the primitives’ relevance for DT definitions from a business decision-making perspective. For that purpose, we coded each of the 16 DT definitions for the applied primitives in advance (Table 5). In the coding process, the second and third author of this paper independently analyzed each definition regarding the coverage of each particular primitive (Table 2). In 90 out of the 96 (16 definitions × 6 primitives) cases there was agreement among the two coders, and in the remaining six cases agreement was achieved through discussion with the first author of the paper. The final coding result is presented in Table 2. After receiving the responses, all text snippets that participants indicated as relevant to their assessment were analyzed with a fuzzy algorithm regarding the corresponding primitives, thus identifying their influence on the assessment. In the present research context, application of a fuzzy algorithm means to reduce the respective terms to their word stems.

Next, we focused on the examples for DT scenarios from corporate practice. As with the definitions, we analyzed the data on the distribution of participants’ categorization of the examples as DT or not DT. We also analyzed the text snippets reported as decisive for the assessments. Various search operators such as ″*″ and ″?″ were applied for the search strings in the analysis to account for different spellings and typos.

Respondents’ gender distribution was almost balanced: 53% female (n = 281) and 47% male (n = 246). Two participants did not specify their gender. The age distribution is shown in Fig. 3 (M = 40.15 years, SD = 12.13 years).

The distribution of respondents’ years of professional experience is illustrated in Fig. 4 (M = 15.66 years, SD = 10.55 years).

The vast majority of respondents (74%) had completed higher education; 45% reported a bachelor’s degree as their highest degree, 25% a master’s degree, and 4% a doctorate. Respondents’ hierarchical positions within their organizations were as follows: 16% held a C-level position, 54% were employed in middle management, and the remaining employees had some other form of managerial decision-making responsibility (e.g., in teams).

Participants were prompted to indicate whether each of the 16 presented descriptions would constitute an appropriate DT definition. Figure 5 shows the results. Interestingly, although we carefully and deliberately selected the 16 DT definitions from the scientific literature for our main survey study (see Sect. 3.3), the share of respondents stating that a definition was not appropriate ranged from 12% (Def_10) to 48% (Def_13). Thus, we observe a significant discrepancy between the scientific literature (see the column “Source” in Table 3) and practice (i.e., business decision-makers’ perceptions and corresponding survey responses).

Using this approach, we obtained further results regarding DT definitions. First, regarding the primitives’ relevance for the overall assessment (i.e., suitable DT definition and no suitable DT definition), we found: the primitive means was relevant to 34% of all assessments of definitions addressing this primitive, whereas impact was relevant to only 15%. The remaining results within these maximum and minimum values were: nature (22%), expected outcome (22%), target entity (19%), and scope (18%). Second, regarding responses stating that a definition was suitable, means was the decisive primitive, relevant to 40%; impact was only relevant in 17% of cases. The remaining results within these maximum and minimum values were: expected outcome (24%), nature (23%), scope (20%), and target entity (20%). Third, regarding decisions that a definition was not suitable, nature, means, and expected outcome were relevant for 16% and impact for only 8%, and we observed 15% for target entity and 12% for scope. Appendix 2 summarizes the detailed results.

Figure 6 illustrates the results of the assessment of whether each scenario describes an example of DT. As previously mentioned, the literature (e.g., Mertens et al. 2017) suggests distinguishing between evolutionary and revolutionary levels of DT based on Venkatraman’s (1994) framework (Fig. 1). Thus, theoretically, one would expect that Ex_1 and Ex_2 (evolutionary levels) would be assessed predominantly as no DT and Ex_3, Ex_4, and Ex_5 (revolutionary levels) as DT. However, for the majority of respondents all five examples constituted DT scenarios. It is even more surprising that Ex_1 and Ex_2 were labeled DT by the largest shares of respondents (86% and 88%, respectively), when we expected that the three examples of the revolutionary levels (Ex_3: 83%, Ex_4: 83%, Ex_5: 73%) would have the largest proportion.

In sum, we observe a significant discrepancy between the scientific literature (see the column “Source” in Table 4) and business decision-makers’ perceptions and their corresponding survey assessments.

To statistically test the significance of the differences between the five scenarios, each of which represents a particular level of IT-enabled business transformation (Venkatraman 1994), we carried out an ANOVA using the Brown-Forsythe method, which performs the ANOVA for the group medians instead of the group means to yield more robust results (Derrick et al. 2018). The results presented in Table 6 show a significant difference between the examples: F(4, 2514) = 12.71, p < .001. The corresponding statistics are presented in Tables 6 and 7.

Subsequently, we performed a post hoc test to show which examples were significantly different. The results presented in Table 8 illustrate significant differences in the assessment of examples 1 through 4 compared with example 5 (see the column ″p_bonf″ in which we present the Bonferroni-corrected p-values for pairwise comparisons).

A t-test also showed a significant difference (p < 0.001) between the two categories of examples, namely the evolutionary levels (Ex_1 and Ex_2) and the revolutionary levels (Ex_3, Ex_4, and Ex_5).

The results of our analysis of the text snippets participants reported as decisive or relevant for their assessments of the examples are illustrated in Fig. 7, which shows their frequency and the corresponding outcome of the assessment. Note that only the terms with the highest frequency of mentions are shown in Fig. 7; in the total bar, all votes are included. In the following, we describe the results, structured along the five examples.

For Ex_1, the terms reported as most relevant for the assessments were automat (automated, automation), autonomous, laser, sensor, handling technolog (technology, technologies), and guided vehicle. Interestingly, all terms except sensor (indicator for DT) and guided vehicle (indicator for no DT) were relevant for both answer decisions.

For Ex_2, the decisive terms for the assessments were integrated database, CAD?CAM (combined), linking, CAD, CAM, and efficient. The only term which was relevant for both answer decisions was integrated database. All other terms show a strict separation. However, the combined mention of CAD and CAM was an indication of DT for a large group of respondents, while the separated terms indicated the opposite for a small share.

For Ex_3, the relevant terms were information?system, pulling data, workstation?based, increase*productivity, process, assembly?line, dedicated case?manager, and case?based?process. Again, there was a small overlap for the terms workstation?based and increase*productivity.

For Ex_4, the most relevant terms were interconnected information system, directly scheduling, external organi?ation (organization, organisation), trucks of external organi?ations (organizations, organisations), port, and scheduling*appointment.

Finally, for Ex_5, the most relevant terms were utili?es IT (utilises, utilizes), change, service?based, business?model, and machine?manufacturer. All terms except the latter were indicated for both decisions (i.e., DT and not DT).

Our results also contradict the argument that a certain degree of innovativeness in the technologies used can be considered a precondition for DT (see the list of the most relevant terms in Fig. 7). The mere mention of technology terms (e.g., automation, autonomous, laser, sensor, integrated database, CAD/CAM) frequently led participants to classify an example scenario as DT, independent from the fact whether they are novel or not. More specifically, although the technologies included in Ex_1 and Ex_2 reflect established IT that is not commonly understood as novel, respondents commonly described these technologies as key reasons why they categorized the examples as DT.