Modular Service Design of Information Technology-Enabled Services

The literature has proposed ways to modularize information-technology-enabled services (ITeS) with limited success. We argue that applying design principles (DPs) can address this gap and revitalize the service modularization literature. With a qualitative research study, we develop exemplar DPs and a set of prioritized DPs for ITeS. We contribute to the literature by demonstrating how complex service systems, specifically ITeS, can be modularly designed. Our DPs show how different ITeS design elements or service attribute combinations impact the outcome-driven design of service experience. Based on the findings, we present a modular service design framework and a service design method that adopts DPs to create effective modular ITeS designs. We also offer ways to conceptualize and apply service modularization to improve the adoption of the modular service design by service designers and managers. Graphical Abstract


Introduction
The increasing deployment of technology is altering how individuals interact with organizations; the traditional goodscentric paradigm is now being challenged by the acceleratory growth of service-dominant economic activities that use information-technology-enabled services (ITeS) (Lusch and Nambisan 2015;Peters et al. 2016;Vargo and Lusch 2004). This recent trend has amplified technology-driven enthusiasm, as demonstrated by the explosive growth in customization, integration, intelligence, and globalization. However, this kind of enthusiasm makes designing ITeS increasingly complicated and challenging. ITeS are essentially a socio-technical phenomenon in which services are designed and delivered using all available means to realize value for both providers and service users (Grönroos 2006(Grönroos , 2007Grönroos and Voima 2013).
We define ITeS as any type of service-based activity that utilizes IT to satisfy users' needs and requirements during the consumption process. Examples of ITeS include self-service solutions, such as online banking or shopping services; software as a service, such as online accounting software; and personal mobile services, such as mobile applications. It is worth mentioning that "IT service" is a sub-concept of ITeS. An IT service is generally defined as complex, knowledge-based work that aims to provide business clients with quality service (Jia et al., 2008). Often, this simply involves technical solutions developed by IT industries. In contrast, ITeS include a broader range of services mediated by IT but that are not necessarily only provided digitally.
The recent service design literature has studied how to design ITeS (see, e.g., Maglio et al. 2009;Patrício et al. 2011;Teixeira et al. 2017). For example, Teixeira et al. (2017) examined the challenges related to designing ITeS. Meanwhile, the service modularization literature (see, e.g., Bask et al., 2014;Dörbecker & Böhmann, 2013;Tuunanen and Cassab 2011;Voss and Hsuan 2009) has more specifically attempted to address ITeS design challenges by proposing different ways to reuse, substitute, or vary service elements. However, the service modularization literature has not yet been able to operationalize these concepts for wide adoption by service design researchers or practitioners. For example, Bask et al. (2014) defined a modular service offering as consisting of a standardized base service(s), customized service(s), or combinations thereof. Tuunanen and Cassab (2011), in turn, proposed a way to modularize the service process using concepts derived from software development, namely reuse and variation of service processes. Neither approach has been adopted in practice, and it seems that the service modularization literature has stagnated in recent years. This is a concern as the perceived view is that service modularization allows companies to offer more product variety, improved flexibility, simplified complex systems, enhanced quality, and cost savings (Jose and Tollenaere 2005; van Liere et al. 2004).
Accordingly, we argue that new approaches are needed to support modular service designs and, specifically, the design of ITeS. Specifically, we propose that the concept of design principles (DPs) is a way to efficiently design modular services. Design principles have been widely used in design-focused research on information systems (see, e.g., Hevner et al. 2004;Markus et al. 2002;Walls et al. 1992). In this study, we define DPs as generalizable design guidelines and abstractions that can be applied to develop service-specific solutions. However, the prior literature has yet to offer service design methods to support the development of DPs. We believe that ITeS provide an excellent setting to study this. ITeS are based on digital solutions, which are today designed modularly. 1 In the information systems literature (Gregor and Jones 2007;Hevner et al. 2004;Markus et al. 2002;Walls et al. 1992), DPs have been used to theorize about and generalize the design of IT solutions. Walls et al. (1992) proposed that metalevel user requirements can be used to depict the meta-designs of IT solutions and that such meta-designs can be used to theorize further and generalize efforts. Later, Gregor (2002) described DPs as design decisions and design knowledge intended for manifestation or encapsulation in an IT solution. Following Sein et al. (2011), in our study, we apply Mathiassen and Sørensen's (2008) framework for defining the different types of DPs: adaptive, collaborative, computational, and network.
Accordingly, our research objective is to examine how DPs can be identified and formulated and whether this approach can be formalized as a modular service design (MSD) method for ITeS. Based on this, we developed our research question: How can DPs be identified and formulated to support MSD and, more specifically, for ITeS? We apply a qualitative research approach to answer this. The data collection for this study was done in collaboration with six New Zealand-based companies that develop ITeS, and we interviewed 25 persons who participated in ITeS development. We used cluster analysis with the data to develop exemplar DPs for modular ITeS design and present how to prioritize these for MSD.
Our study contributes to existing knowledge by proposing a new definition for modular service offerings to renew the service modularization literature and an MSD framework, which can be used to develop new MSD methods. We operationalize the framework to develop the primary output of the study, an MSD method, that can be applied to the design of ITeS. The developed MSD method is ready for use by practitioners. We demonstrate how complex service systems, specifically ITeS, can be modularly designed. We also suggest that it is important to consider why specific service design methods work, which the extant literature typically does not discuss. Our study offers an exemplar of how to develop a theory-based and driven service design method. This opens new ways to advance the development of service design methods and the practice of service design.
The following section reviews the literature on ITeS, service design and modularization, and the four general types of DP for ITeS. After that, our research methodology and analysis are presented, followed by the findings. Finally, we conclude by discussing the findings and limitations of the current work and topics for future research.

Rationale for MSD OF ITeS
Services are becoming increasingly important for organizations. Service, as a concept, refers to applying specialized knowledge, skills, and experience to co-create value for both the service user and the provider (Grönroos 2008;Grönroos and Voima 2013;Vargo and Lusch 2004). Recently, there has been a shift toward "a world in which value is the result of an implicit negotiation between the individuals and the firm" (Prahalad and Ramaswamy 2004, p. 7). Enabled by ITeS, the interactions between customers and organizations strongly affect the way companies operate and compete in the market (Vargo and Lusch 2008).
However, the extant service research literature does not sufficiently accommodate the needs related to designing ITeS. Berry and Lampo (2000) highlighted the need to consider how service (re)design should be performed to make services appealing and satisfying. For this purpose, they proposed a framework for different service (re)design approaches that may help to create innovative new services or rejuvenate existing ones. This is a typical approach to service design in the literature. Authors often offer different methods for designing services or service systems (Maglio et al. 2009). However, they do not explain why this method would provide better utility or value to service designers than other methods. Sangiorgi (2014, 2018) proposed an approach that can provide a foundation for comparing service design methods. Yu and Sangiorgi (2018, p. 41) defined service design as "an integrative approach to collaborative and cross-disciplinary service innovation." This definition is based on the designcentered approach to service design (cf. Mager, 2008). Yu and Sangiorgi conceptualized the role of service design in value cocreation and new service development. Sangiorgi (2014, 2018) divided the approach in terms of focus to how services are designed (service design and analysis) and how services are implemented (service development and implementation). This argument is summarized in Table 1.
The framework below depicts objects of service design and links these to different phases of the process. The left side focuses on the value, form, and function of the service and the service experiences and outcomes, while the right side looks at the structure, infrastructure, and process of the service delivery system. Yu and Sangiorgi (2014) also described some facilitators for service design, such as different methods and tools, staff and customer involvement in the service design process, and how the organization's characteristics may impact the service design work.
Our review of service design method literature, 2 which addressed how service design methods manage the integration of "design and analysis" and "development and implementation," found two methods that, at some level, achieve this goal: the multilevel service design method (Patrício et al. 2011) and the management and interaction design for service (MINDS) (Teixeira et al. 2017). The authors looked at different levels of service design and proposed how to apply well-known service design methods (or facilitators), such as blueprinting (Bitner et al. 2008;Shostack 1984) or value constellation mapping (Michel et al. 2008), to understand the potential (or perceived) value for a developed service. The authors also highlighted the importance of the technological aspects of the design (e.g., the service delivery system). The data collection was done with semi-structured interviews with customers and other actors related to the developed services. In a more recent study to prototype a service, Teixeira et al. (2017) integrated more traditional service design methods with those commonly used for IT design, such as affinity diagrams, system navigation mapping, and user experience blueprinting. In Table 2, we have summarized the MINDS service design method using the Yu and Sangiorgi (2014) framework. We can see that MINDS supports developing a service concept and encounter design by focusing on the form and functions of the service.
MINDS pays attention to the service experience design with a specific version of blueprinting and combines that with user interface/wireframe sketching. Similarly, MINDS looks at the service delivery system from the user's navigation process and structure, but the method does not directly support infrastructure design. In addition, the method currently does not provide support contextualizing the use of the method to different organizational settings and work cultures.
However, while the multilevel service design method (Patrício et al. 2011) and MINDS (Teixeira et al. 2017) are a step toward the integration of design and analysis (of a service) and development and implementation (of a service delivery system), we argue that the authors still only investigated instantiations of service systems and, more specifically, how the selected service design methods enabled the design of these systems. What is lacking, and more general in the extant service design literature, is a way to integrate the service concept and encounter better service delivery system design, which becomes more and more important when designing ITeS. Here, an MSD can provide an answer to resolve this difficult problem.
The service modularization literature (see, eg., Bask et al., 2014;Dörbecker & Böhmann, 2013;Tuunanen & Cassab, 2011;Voss & Hsuan, 2009) ; ; has examined how different ITeS design elements (Sheng et al. 2017) or service combinations (Ordanini et al. 2014) impact the design of service experiences. The extant service research defines service experience as phenomenological or process-or outcome-based (Helkkula 2011), and it typically takes the customer's point of view to study it (Teixeira et al. 2012). The service modularization literature, in turn, looks at the service experience as an objective for the service design. Patrício et al. (2011) have been proponents of developing service design methods that support the design of service experiences. Zomerdjk and Voss (2010) described this outcome-driven approach as experience design. In turn, Jaakkola et al. (2015, p. 190) characterized this as a way to orchestrate service elements to design service experiences. We suggest that service modularization enables the reuse of service elements, such as process steps, which can be combined in service implementation (Bask et al. 2014) to improve service experience outcomes. Tuunanen and Cassab (2011) aimed to understand how different service modularization choices impact the likelihood of customers using the service again and the impact of these choices on the utility perceived by the customers. However, the use of IT can dramatically increase or decrease the number of encounters between a person and a firm (e.g., Bitner et al. 2000). It has been found that users' perceptions of service encounters are strongly linked to their satisfaction with overall service quality (Parasuraman et al. 1985;Parasuraman et al. 1991). This can apply to ITeS-related services as well (e.g., Jiang et al. 2002). Users form perceptions of service encounters by evaluating the tangible aids in these encounters, such as the IT interfaces. Service users are increasingly creating their own experiences dynamically and autonomously (Ostrom et al.  Sangiorgi (2014, 2018).

Focus Design and Analysis Development and Implementation
Objects Service concept (value, form and function, experience, and outcomes) Service delivery system (structure, infrastructure, and process) Facilitators Methods and tools, staff and customer involvement, and organizational dimensions 2015). For instance, in an interactive episode of Netflix's television show Black Mirror, viewers can choose different story paths and achieve different service experience outcomes. This highlights the need to examine how ITeS should be designed and how service designers could be assisted in this process.
Although service modularization attracted some interest in the 2010s, its impact on the extant service research literature has been marginal so far. Brax et al. (2017) summarized this situation and proposed research topics to advance the field, including service-specific modularity theories and principles, architectural innovation in services, and modularity in hybrid offerings combining service(s) and tangible product(s). Furthermore, Ordanini et al.'s (2014) call to compare combinations of service attributes with individual service attributes can be echoed for the service modularization literature in general: by understanding how different combinations of service attributes impact the service experience, we can (re)design services to better fulfill the needs and wants of customers, including combinations of the attributes of the architecture, scale, style, shape, and layout of the service as well as functional or aesthetic service features (Sheng et al. 2017). How this can be accomplished remains an open question.
We argue that, by shifting the focus from attempts to modularize different aspects of a service to the DPs of services, we can identify possible ways to generalize the modular design of ITeS and respond to the need for better integration of the service concept and encounter, as well as service delivery system design. For this purpose, we apply Mathiassen and Sørensen's (2008) framework to develop DPs for modular ITeS design. According to Mathiassen and Sørensen (2008), there are four general types of DPs for ITeS. Computational DPs rely on encountered services, and they support repeatable patterns of information processing. Adaptive DPs interpret and transform available and emergent information by adapting information processing patterns to specific contexts. They rely on relationships and allow involved actors to explore and debate interpretations while executing tasks. Networking DPs help actors produce their information about phenomena by following standardized and repeatable patterns of information processing. They connect actors to relevant information sources through IT solutions, such as email systems, search engines, electronic libraries, mobile phones, and SMS messaging. Collaborative DPs support actors in producing information about phenomena within an organization and its environment by interpreting the specific work context (Mathiassen and Sørensen 2008). McKenna et al. (2013) utilized Mathiassen and Sørensen's (2008) framework to study how adaptive and computational service components are linked to self-efficacy, collaborative service components are linked to social influences, and networking service components are linked to facilitating conditions. A follow-up study by McKenna et al. (2018) demonstrated that different service components have more diverse ties than were found in the earlier study. Although these studies provide valuable insights into the design of ITeS, they do not provide a means for developing an MSD method.

Research Methodology
This section presents how we resolved the problem of developing an MSD method using a qualitative research approach. The methodological choices depicted here also provide the foundation for the MSD method summarized and presented later in the findings section. We first applied laddering interviews to enable rich qualitative data collection. With laddering interviews researchers are able to collect in-depth data about individuals' perceptions of products or services and their reasoning for these perceptions (Grunert et al. 2001;Reynolds and Gutman 1988). More specifically, we applied a version of laddering interviewing, which a group of information systems researchers has developed for system design applications in the past two decades (Tuunanen and Peffers 2018). We then performed a cluster analysis to identify clusters in the collected data within each different service type (Peffers et al. 2003). This procedure allowed us to formulate DPs. The data collection and analysis processes are described in more detail below.

Data Collection
We collected data from New Zealand companies developing different ITeS for the firms' service offering(s). We adopted Mathiassen and Sørensen's (2008) four general types of DPs for ITeS as the basis for our theoretical sampling and selection of the participating companies (Patton 1990) to cover all four different types in our data collection. Four of the six companies were selected based on each of the four general types. This was to ensure that the participating companies were specialized in at least one of the general types.
The six participating companies were all small-to mediumsized and operated within New Zealand. Companies 1 and 3 specialize in innovative, online-enabled, web-based service solutions for individual consumers or small businesses, and both had been operating for less than 4 years. Company 2 has 12 years of experience in offering adaptive online marketing consultancy services to the marketplace. It specializes in search engine optimization and tactical implementation of online marketing. Company 4 is an experienced franchise retailer of one of the biggest mobile network services in New Zealand. It specializes in selling network services to individual customers and businesses. Company 5 has 32 years of experience developing systems or online service platforms for business users, such as online work management systems or e-commerce solutions. Company 6 is a travel service wholesaler that uses a combination of online services and computer-based software to operate its business. This company is one of the few travel service wholesalers that has been operating for almost 25 years in New Zealand. The data collection was conducted in 2009.
We followed the examples of Reynolds and Gutman (1988), Peffers et al. (2003), and Tuunanen and Peffers (2018) to conduct our laddering interviews. The laddering interview technique is based on the personal construct theory developed by Kelly (1955). Kelly's aim was to model individuals' belief structures based on personal constructs, which result from individuals' observations and interpretations of events (Pervin 1993), to determine how their perceptions of certain situations impact their experiences in those situations. In other words, he argued that a person has individual multi-dimensional models (i.e., constructs) that describe the attributes and behavior in relation to objects and events, their consequences, and their relationship to a person's values.
Later, Gutman (1982) proposed that product attributes are relevant to consumers because of the consequences derived from consumption behavior. These consequences are relevant to the personal values they help to satisfy for the consumer. A complete sequence of attribute-consequence-value association is referred to as a means-end chain. To study consumers' meansend chains for a given product, Reynolds and Gutman (1988) developed an interview approach called "laddering." In such interviews, participants are typically given a choice or decision task related to a service or product category and then asked to describe which service or product attributes informed their decisions (Modesto Veludo- de-Oliveira et al. 2006). Then, participants are questioned to identify the relevant consequences they experienced from using the service or product (Reynolds and Gutman 1988). Probing questions are asked until the participants describe the final personal values they satisfied by consuming the service or product.
We began the interviews by presenting a list of stimuli (see Appendix A) intended to suggest ideas about possible service applications to enable brainstorming by the participants (Peffers et al. 2003). Following Peffers et al. (2003), we then asked the participants to rank the stimuli in terms of importance. Then, one at a time (for the two highest-ranked stimuli), the interviewer asked each participant to describe the important applications and the desirable attributes (i.e., features) of said applications. The interviewer proceeded to ask the participant to explain why each particular feature was important to elicit the consequences that the participant expected from the feature. The interviewing process continued with a series of "Why?" and "Why would that be important?" questions to determine what end result the subject expected (Reynolds and Gutman 1988). To elicit more concrete system attributes, we asked the participants a series of questions, such as "What about the system makes you think that it would do that?" The data were recorded as attribute-consequence-value chains, as Tuunanen and Peffers (2018) described. An example of the laddering interview process is included in Figure 1. In this study, we selected six participating companies and conducted 25 individual laddering interviews with the companies' employees. The demographic information for the 25 interviewees is presented in Table 3.
We recorded all the data we obtained from the individual laddering interviews in a series of laddering chains. From the 25 interviews we conducted, we recorded 556 chains that we used for data analysis. For each chain, the individuals mentioned approximately 8-14 consequences. In addition, we recorded at least one attribute for each chain, and when a chain branched out, we recorded several attributes. Regarding values, our experiences were similar to those of other studies (e.g., Peffers et al. 2003), and we recorded values for a total of 227 chains.

Data Analysis
Data analysis for the study was done in two phases: data coding and cluster analysis. First, we needed to adapt the rich textual data in the laddering chains to what we could use for the cluster analysis. For this purpose, we performed two iterations of data coding. The two iterations are the interpretation process for analyzing our laddering data. This is considered the most critical step for analyzing laddering data as it will directly influence the content quality and the results (Gengler and Reynolds 1995). To avoid bias, the laddering data were coded by two researchers at the same time. The codes were revised and checked several times by the two researchers. The second phase of the data analysis used cluster analysis to conceptualize the DPs. For this purpose, we conducted hierarchical cluster analysis using Ward's method (Peffers et al. 2003).
The first iteration of data coding involved assigning descriptive codes for service attributes, performance consequences, and values. Three new columns for attribute, consequence, and value codes were inserted into the worksheet containing the laddering chains. If more than one code could be directed from the laddering chain, the ladders were copied into a separate line to form a sub-chain. The ladders that derived the relevant codes are colored accordingly. After the first iteration, we obtained 344 unique attribute codes, 505 unique consequence codes, and 91 unique value codes, which were summarized based on the original words used by the participants. The second data coding iteration concerned the classification of similar attributes, consequences, and values. We aggregated similar codes into smaller sets based on the three new columns for attributes, consequences, and values. First, we sorted the chains in alphabetical order based on the 505 unique consequence codes. These consequence codes were examined based on similarity and were classified into smaller groups. Then, an identical code was developed for each small group of consequence codes. Thereafter, identical codes were aggregated and cross-checked by the two researchers. The same process was performed for attribute and value codes.
Next, the data needed to be converted from text to binary format to enable the application of the hierarchical cluster analysis using Ward's method, in line with Peffers et al. (2003). Therefore, we converted the unique codes of 221 attributes (A), 96 consequences (C), and 21 values (V) into columns of binary numbers (0 and 1). The service attributes were extracted from the features of ITeS tools mentioned by the participants, and, as such, they are specific to certain services. This resulted in a large aggregation of service attribute codes that emerged from the various ITeS being used or developed by the participants. The columns Interview ID, Reference Code, Service Type, and Chain Number were selected from the original laddering chains. With the binary representations of A, C, and V codes, the selected columns formed a binary matrix table for use in the cluster analysis. The binary matrix table contains 424 columns and 556 rows of data. The columns, codes, and data types are described in Appendix B. For laddering chains that did not have a relevant A, C, or V code, we created three extra columns: A0, C0, and V0. The binary columns have a value of 0 or 1, where 0 indicates no such code exists in the laddering chain, while 1 indicates the reverse.
Next, we wanted to develop DPs from the data set. For this purpose, the laddering chains were first clustered based on the following variables: Attribute Codes (A), Consequence Codes (C), Value Codes (V), and Service Type (ST). The ST code was derived from the stimuli selections made by the interview participants, which we recorded during the interviews for each ladder chain. After that, we first generated cluster solutions from two to eight clusters for the initial hierarchical analysis within each ITeS type (i.e., subsets of the data). For the clustering analysis, the laddering chains formed the unit of analysis. To measure distance, we squared the Euclidean distance to measure similarities between laddering chains, as our data were in binary metric format for a hierarchical clustering procedure (Hair et al. 2006). In hierarchical clustering, clusters are nested rather than mutually exclusive as they are formed only by joining existing clusters. Any member of a cluster can trace its membership in an unbroken path to its beginning as a single observation. Finally, we adopted Ward's method as our clustering algorithm. 3 Finally, we used correlation tests between each consequence and value code 4 within each sub-cluster to determine aggregate connections between the constructs using the Kendall tau rank correlation method (Abhi, 2007), which measures the strength of association between the consequences values within each sub-cluster. The correlation test helps elucidate the dependence relationship between the design elements and what impacts them. 4 If we did not find a connection between the consequence code and a value code, we added a not applicable (N/A) note to the table. We began to interpret the codes by developing an analytical tool, keynotes, based on the wording used in the consequence and value codes. These keynotes were then used to develop the DPs. Furthermore, the content within the chains was used to interpret each DP. The consequences reflect the design elements valued by individuals based on their experience with each type of ITeS.

Findings: Prioritized DPs
This section depicts how the applied research methods enabled us to define DPs and prioritize them. Figure 2 presents a summary of how specific DPs fall into each of the four general types. We use the number of laddering chains to indicate priority between the general DP types and order the DPs within each type. In Figure 2, the most important DPs are on the left both within the four general types and within each DP type. Table 4 describes all the DP definitions and the descriptive statistics for each DP. 5 Below, we also present four exemplars of the dataset based on the computational, adaptive, collaborative, and network DPs. We selected two of the highest-ranking examples for adaptive and computational DPs and two of the second highestranking collaborative and network DPs to give some variety.

Computational: Flexible Information Integration
For this DP, the key issues are integration, usability, flexibility, performance, and responsiveness of the service delivery system. This requires standardized information obtained from different units within or across organizations. According to the participants, successful integration is the backbone of the other key aspects because it organizes information from several sources and thus reduces, for example, the risk of inputting the same information several times. It is also recommended to deliver information through one integrated system and update information in real-time within one system or in sync within several systems. Based on these results, we propose the following DP: DP9: Computational designs should focus on flexible information integration with other services.

Adaptive: Usability and Customer-Centricity
For this DP, three important issues emerged from the data: functionality, customer-centricity, and usability. The participants emphasized the role of practical functions in adapting to tasks performed by knowledge workers, such as merging information from different databases. The participants also highlighted usability since they required easy-to-use functions, hassle-free recovery options, and timely responses in their service encounters. Positively perceived usability led participants to think that they could rely on the ITeS they used. Based on these results, we propose the following DP: DP1: Adaptive designs should focus on the usability and customer-centricity of the functionalities.

Collaborative: Accessibility and Knowledge Sharing
For this DP, we observed two service concept issues: accessibility and sharing. The participants indicated that they seek easily accessible ITeS to support distributed work and collaboration. However, end-users often experience transaction overload when interacting with an ITeS. The participants also believed that if an ITeS promotes open knowledge sharing across users (e.g., in organizations), it can create a culture where everyone is encouraged and potentially willing to contribute to a common good. Based on these results, we propose the following DP: DP5: Collaborative designs should focus on accessibility and information sharing to support collaboration.

Network: Constantly Improving Service Efficiency
For this DP, efficiency, mainly cost and resource savings, were the most important aspects. Participants also highlighted that they aim to minimize technical support by constantly improving the quality of services. For example, one participant stated that the goal was to eliminate customer support and, in so doing, decrease costs and increase the efficiency of the service delivery system. Based on these results, we propose the following DP: DP13: Network designs should focus on constant improvement of service efficiency.

General Discussion and Implications for Research and Practice
Our study contributes to the literature by showing how complex service systems, specifically ITeS, can be modularly designed (Maglio et al. 2009). Our DPs show how different ITeS design elements (Sheng et al. 2017) or service attribute combinations (Ordanini et al. 2014) impact the outcome-driven design of service experiences (Zomerdijk and Voss 2010). Our study also responds to the recent call for the advancement of service modularization research (Brax et al. 2017). We offer a new approach to designing services (specifically ITeS), applying DPs as the conceptual framework foundation for service modularization. The implications of our research for future research are summarized in Table 5 and discussed below. Tuunanen and Cassab (2011) argued that it is necessary to understand how ITeS should be designed and proposed service process modularization to achieve this. Although some modularization research has been published in recent years (see, e.g., Brax et al. 2017), the overall impact of this research stream on the service research community has remained modest, and it has not yet significantly impacted service design research or practice. Recent papers published by the Journal of Service Research (Patrício et al. 2011;Teixeira et al. 2017) still focus on combinational solutions that apply sets of well-known service design methods, such as blueprinting (Bitner et al. 2008;Shostack 1984) or value constellation mapping (Michel et al. 2008), to address the complexity of designing ITeS. Although they offer valuable insights, these studies do not adequately explain how to design combinations of service attributes.
Therefore, we argue that we should move beyond traditional ways of conceptualizing service design methods to facilitate the design of service attribute/feature combinations. The application of DPs provides a meaningful foundation for this. However, this solution requires re-thinking the expected outputs from service design activities and how DP-driven modular service design should be performed. Earlier, Sangiorgi (2014, 2018) proposed an approach to compare service design methods (cf. Table 1) based on activities (design and analysis/ development and implementation), objects of the activities (service concept/service delivery system), and facilitators for

Implications for research IMPLICATIONS FOR PRACTICE
Modularization of services should be considered at the DP level to facilitate the design of service attribute/feature combinations. The general DPs provide a theoretically sound way to reuse, substitute, or variate service modules.
The MSD method is ready for use by service designers and managers.
The MSD framework shows how DPs can be integrated into service design methods.
Service designers and managers should initially recognize the standard features, how these can be customized, and how combinations of these can be used for service modularization. We propose that modular service offerings be redefined as standard, customized service features and DPs, or combinations thereof.
The MSD method can be modularized, and service designers can substitute or variate parts of the method.
these (methods and tools/staff and customer involvement/ organizational dimensions). However, the extant literature does not yet offer ways to integrate DPs into service design methods. We argue that DPs can be used to modularize service modules (Brax et al. 2017), and that the four general DPs for ITeS provide a theoretically sound way to accomplish this. We further suggest that these general types can be further divided. Namely, for the service concept object, we propose that adaptive and collaborative DPs should be emphasized. Correspondingly, for the service delivery system object, computational and network DPs should be considered first. Our findings support this argument (cf. Table 4 and Online Appendix 1), which is summarized in Table 6. Still, we also recognize that this issue is likely not straightforward. Moreover, the application of the DPs is contextual, and the emphasis of DPs for each IT-enabled service will vary. Thus, the presented DPs should not be considered general guidance for all ITeS designs but instead instantiation for our MSD method's applicability to support modular service design for ITeS.
However, our conceptualization of the MSD methods (Table  6) offers new possibilities for theorizing about service solutions (Markus et al. 2002) and developing generalizable DPs for ITeS, its more immediate impact is on the further development of service design methods that accommodate the discovery and recognition of DPs. These methods should also enable modularization at the service attribute combination (i.e., feature) level, thus resolving the complex problem of operationalizing service modularization in service design and, more importantly, how to show its benefits in terms of efficacy and efficiency. So far, these have remained unsolved by the service modularization research community (Brax et al. 2017).
However, our study provides the foundation for this work. As a first step forward, we argue that the definition of a modular service offering (Bask et al. 2014) should be reformulated. We suggest that modular service offerings should be defined as standardized sets of base service features and DPs, sets of customized service features and DPs, or combinations thereof. Our study offers an example of how DPs can be formed based on rich laddering interview data. Tuunanen and Peffers (2018) have demonstrated how attribute-level information for service features can be derived from similar datasets. We foresee that the proposed definition of modular service offerings can revitalize the current stagnant state of modularization research. It opens new ways, for example, to study architectural innovation in services and how modularity in hybrid offerings combining service(s) and tangible product(s) can be accomplished (cf. Brax et al. 2017).

Implications for Practice-Presenting an MSD Method
The findings provide general guidance for applying the MSD framework in practice, specifically for the design of ITeS. Table  7 depicts how the MSD framework can be operationalized and how service modularization can be used effectively for service design, especially for ITeS design. Although the literature has argued for the benefits of service modularization (see, e.g., Bask et al. 2017), the industry has been slow to adopt the concept in practice despite the argued benefits of efficiency gains for the design shown in product and software development (Tuunanen and Cassab 2011). Our argument here is that the problem may be in the conceptualization of service modularization by academic researchers.
The presented MSD method is ready for use by service designers. It can be extended further, for example, to develop the analysis approaches to target specific service modules in an IT-enabled service. The firms can use the MSD method to collect data and derive DPs that offer a concise overall view of the most central design aspects for their ITeS. Such a comprehensive view can be beneficial for service designers and managers and decision-makers, for example, by orchestrating the integration of service elements into the design of service experiences (Jaakkola et al. 2015). Our advice for applying this method for service modularization, and the resulting DPs, to a particular study, is first to look at the general composition of an ITeS and then use specific DPs to enhance its design. Of course, this should be done with the help of various methods and techniques to identify the user needs that will inform the features of the ITeS. Our findings provide a starting point for understanding the underlying principles of different types of ITeS and how they should be designed. In this way, our DPs start to bridge the potential "language gap" between the more techsavvy individuals and others.
We also offer a more straightforward definition that focuses on service features and more generalizable DPs. Furthermore, we suggest that practitioners should initially recognize the standard features instead of using several modularization types (like reuse, substitution, or variation). These features are customized and, lastly, combinations of these modularization types. In our view, the more important matter is to understand how to design new services versus discussing the finer details of philosophical differences between, for example, reused and Table 6. The MSD Framework, Based on Bask et al. (2010), Mathiassen and Sørensen (2008), and Sangiorgi (2014, 2018).

Focus
Design and Analysis Development and implementation Objects Service concept (potential value, form and function, experience, outcomes) Service delivery system (structure, infrastructure, and process)

Design principles
Adaptive and collaborative service modules (standard, customized, or combinations thereof) Computational and network service modules (standard, customized, or combinations thereof) Facilitators Methods and tools, staff and customer involvement, organizational dimensions variated service features. Thus, we simply propose that DPs should be the basis for service modularization efforts. We suggest that practitioners use DPs, for example, to recognize standard, customized, or combined service features. These can be later considered for reuse, variation, or substitution to improve the MSD efficiency further. Our MSD method also provides rich information about the actual or perceived use of different types of ITeS. These details can be utilized in tandem with the service modularization approach; developers and providers can look at each service module and evaluate how well the module supports the DPs relevant to the IT-enabled service type in question. In this way, managers can more accurately identify developmental areas and weaknesses in their service concept and delivery systems at a feature or feature set level (cf. Table 7) to improve the customers' service experience. In other words, by recognizing how different service features impact the service experience, we can design better services (Ordanini et al. 2014). Furthermore, our MSD method can provide service designers with ways to achieve a fine-grained view of each service module. By analyzing the laddering data and the attribute-consequence linkages, service designers and managers can gain detailed information about how each attribute related to a particular module matches (or does not match) the desired outcome of the module. In addition, Table 7 also illustrates how the MSD method facilitates different methods and tools to support staff and customer involvement. Furthermore, we argue that the method also enables taking account of organizational needs and customer requirements for the developed ITeS via prioritizing the recognized DPs.
Our MSD method can also be modularized itself, and service designers can substitute or variate parts of the method to fit their organizations or projects. Our approach builds on the work done in applying personal construct theory to understand customers' mental models (Kelly 1955). However, other suitable theories may similarly help guide the development of DPs and MSD methods. We ask the service designers to consider that they pause to think why specific service design methods work and the (theoretical) reasoning for this. This may lead to the design of better service design methods. Our method provides an exemplar of this by purposefully developing a method to support MSD for ITeS. This is a different approach to developing service methods than we usually see in the literature. For example, blueprinting (Bitner et al. 2008) or the more recent MINDS (Teixeira et al. 2019) do not make an argument as to why these methods work. Moreover, these are combinations of earlier work that have been customized to offer a solution to a set of service design problems.
We recognize that these well-known service methods, such as blueprinting, clearly have value as the industry has widely adopted them. Still, what could we achieve if we purposefully, using the literature and research methods, developed theoryingrained methods for service design, and how would this change service design practice? Kurt Lewin argued in the 1940s that "there is nothing more practical than a good theory" (Hunt 1987). We believe that this is also true for service design methods and the practice of service design. The answer is not likely to be just "efficiency gains for service design." Still, we would also learn, for example, why certain service design methods are easier to adopt by service designers and managers and whether the applied service design methods impact the service itself.

Limitations
Our study develops an MSD method that particularly fits the design of ITeS. Furthermore, the developed method applies DPs to conceptualize how modularization can be accomplished in service design in general. We applied a qualitative research approach for development of the method. The data consist of 25 individual interviews with representatives of six organizations in New Zealand. We recognize that our participant sample is limited since it was obtained from a single country. However, New Zealand's multicultural society brings potential richness to the dataset. Another limitation is that the organizations chosen for the case study were selected from the current pool of Develop prioritized computational and/or network design principles that can be used to define ITeS modules (standard, customized, or combinations thereof) Facilitators Apply methods and tools for staff and customer involvement and accounting for your organizational needs and customer requirements: • Laddering interviewing for rich customer data collection • Data coding for developing user need definitions and data constructs • Cluster analysis for developing DPs and aggregated design knowledge • Prioritizing DPs to meet your organizational needs and customer requirements companies in the Auckland region using theoretical sampling. Although the number of interviews (25) conducted can be considered low, it is within the range of the number of interviews conducted in similar studies (Peffers et al. 2003;Peffers and Tuunanen 2005;Tuunanen and Govindji 2016;Tuunanen and Kuo 2015;Tuunanen and Peffers 2018). Thus, we see that our study meets the expectations of similar studies in the literature, and it opens many new avenues for further research. For example, we should study how the context of specific ITeS impacts the application of the MSD method or how to define new, generalizable DPs that can be applied outside of specific ITeS contexts.
Appendix A.

Stimuli List
At the start of each interview session, the interviewee was given a brief description of the four types of ITeS in order to stimulate the discussion. The descriptions of each type of ITeS were as follows.
Network services: Services that deal with instant communication. Such services aim to connect users to relevant information sources through software or hardware. Examples include email systems, broadband services, mobile phones, and SMS messaging. The technologies available for these types of services will provide users with immediate access to relevant information sources.
Collaborative services: Services that rely on relationships. Information is shared on a collaboration platform for distributed work. Examples of collaboration systems include Microsoft sharePoint, Google Docs, and Microsoft Exchange Server.
Adaptive services: Services that create flexible business processes or enhance product customization. Examples of an adaptive service evoked by a customer include the standard task of ordering groceries from a website, paying for the items with a credit card, and choosing a delivery date and time. Another example of an adaptive service is search engine optimization (SEO) for websites, which is the process of improving a website's ranking.
Computational services: Transactional type of information services. These deal with processing operational data, such as transactions, accounts, or customers. An account management system is an example of this type of service.
To stimulate ideas, the participants were asked to talk about and list the types of ITeS with which they have interacted. Then, they were asked to describe a recent problem they encountered with this type of service. Guiding questions for the laddering interviews were as follows.
Guiding Questions: What kinds of ITeS you are using/developing right now? Please describe your experience using ITeS.
What is the first thing that comes to your mind when you see the term "IT-enabled services?" What is the main challenge you have encountered when using these services?
How does the performance of the service affect your work? How complicated do you think the types of ITeS that you are using are? Appendix B.

Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Column Number
Column Name Code