1 Introduction

Increasingly, commercial service innovation is enabled by information and communication technologies (ICT). For instance, in the entertainment industry, digital resources such as music and video are delivered via the Internet (Premkumar 2003). The delivery of these resources is a commercial service because customers have to pay for it, or have to accept ads.

Information technology is not only an enabling factor for many services, rather such technology is the actual service. Take for example content storage services (e.g., data storage in the cloud). Here, information technology plays a crucial role in developing and providing commercial services themselves.

To develop ICT components, it is common practice to state first requirements and designs in terms of conceptual models (e.g., cf. the UML set of diagrams). We argue to use such type of conceptual models also for developing the business understanding of a service at hand. For this purpose, we have presented e 3 value (Gordijn and Akkermans 2003) and e 3 service (De Kinderen et al. 2013; Razo-Zapata et al. 2012), and its predecessor serviguration (Baida et al. 2003), which resemble a model-based concept of working, but use a business-oriented terminology.

Commercial services require at least two parties, namely a customer and a supplier. However, often more than one supplier is involved. For instance, to obtain a music stream, a user needs a content provider for the stream itself, but also needs an Internet Service Provider (ISP) for having Internet access. Therefore, services are provided and consumed in networks, which we call service value networks (SVNs) (see e.g., Basole and Rouse 2008; Allee 2002; Razo-Zapata et al. 2012).

SVNs can be manually designed beforehand, e.g., by business developers of participating companies. However, since many services are actually ICT services, a logical next step is to automatically compose SVNs, given a customer need.

In this paper, Sect. 3 explains our vision of SVN composition, in relation to other efforts. In Sect. 4, we discuss e 3 service as a way to semi-automatically compose SVNs. Then, this paper contributes a critical reflection on e 3 service (see Sect. 5). Based on this reflection, an SVN research agenda is presented in Sect. 6, in addition to Gordijn et al. (2012). We illustrate our vision and research vision by means of an illustrative online entertainment example (Sect. 2).

2 Example: The Online Entertainment Marketplace

Based on previous case studies (Gordijn et al. 2011) and new ways of delivering multimedia content (Premkumar 2003), we have developed an educational example to describe the elements within our long-term vision as well as future research challenges with respect to e 3 service.

Our example assumes that a customer wants to compose an entertainment service out of elementary services. For example, as shown in Fig. 1a [cf. the annotated e 3 value notation (Gordijn and Akkermans 2003)], a service bundle may include only basic Internet access. In this way, the bundle consists of only one actor (ISP 2000) performing one single service (Basic Internet), which offers the value object (Internet access) in exchange for the other value object (Fee). In addition, the Internet service may also depend on other services (service enabler) such as SIP Footnote 1 control to implement Voice over IP (VoIP), which is outsourced to a third actor.

Fig. 1
figure 1

Two service value networks modeled as e 3 value models

Full size image

In contrast, as depicted in Fig. 1b, a service bundle may contain services such as basic Internet access, a help desk, a streaming service and an online music service. All these services can potentially be offered by different suppliers, so that a multi-supplier service bundle emerges. Similar to Fig. 1a, services mentioned by Fig. 1b may also depend on service enablers. For instance, “GS Plus” may require software protection from reverse engineering, which can be outsourced to a third actor such as “Security Inc.” Footnote 2

The focus of the models is on what kind of objects must be transferred in order to cover customer needs. (a) An SVN for a customer need demanding FC1 (connection to Internet). (b) An SVN for a more elaborated need demanding FC1, FC2 (Online Basketball), FC3 (FBA press conferences), FC4 (Tech support) and FC5 (Classic recorded concerts).

3 Vision of the Composition of Service Value Networks

As many other systems, SVNs exhibit both inherent (due to the number, existence and nature of internal interactions) and epistemic (because our limited knowledge about their properties) complexity (Basole and Rouse 2008; Sommerville et al. 2012). To address these types of complexity, we have conducted substantial research (Akkermans et al. 2004; Baida et al. 2003; De Kinderen et al. 2013; Razo-Zapata et al. 2012). Nonetheless, the current state of the art of e 3 service is still at quite a distance from our long-term vision, which can be characterized by four elements: a multi-perspective view, integration, market-based composition and resilient SVNs (Gordijn et al. 2012). While the first two elements deal with epistemic complexity, the last two address inherent complexity.

3.1 A Multi-Perspective View of Service Value Networks

Since SVNs represent complex interactions between customer, service suppliers and enablers, they are influenced by at least four perspectives: economic, social, political and technological (Basole and Rouse 2008). In a similar vein, efforts such as the Unified Service Description Language (USDL) focus on three perspectives: business (usually dealing with economic and political issues, too), technical/software and operational/platform (Cardoso et al. 2010; Barros and Oberle 2012). Consequently, given the complexity of integrating several perspectives within SVNs, a multi-perspective integration framework is yet to be developed.

3.1.1 The Business Value Perspective

We assume a critical separation between business value and business process aspects since the former is focused on what commercial services offer and what they want in return; whereas the latter is centered on how the transfers are achieved in terms of operational processes. We separate the ‘what’ from the ‘how’, because to our experience, it is already quite difficult to create a shared and agreed understanding of which commercial services are offered and requested by whom in the first place. E.g., in the online entertainment market it might be important to analyze what objects of economic value (digital resources such as tracks, streams, and money) are transferred among actors as well as the sustainability of composite SVNs.

3.1.2 The Business Process Perspective

As we are interested in fully operational SVNs with automated coordination, we must provide means to allow orchestration and choreography (Papazoglou 2008) of web services that can execute SVNs’ processes. E.g., once a customer decides to acquire an SVN, a cross-organizational business process must start to coordinate the transfer of digital and monetary resources. The business process perspective answers the ‘how’ question as discussed in the previous section.

3.1.3 The ICT Perspective

Since we assume that SVNs can be composed of machine-readable and executable commercial services, an ICT architecture should be specified, preferably a Service Oriented Architecture (SOA) (for instance, cf. web services) (Papazoglou 2008; Pautasso et al. 2008). Nonetheless, it might be also possible that more than one web service is required to implement a single commercial service. E.g., a commercial service offering online music might be implemented by web services such as audio streaming, tracks downloading and audio compression.

SAP’s USDL has a number of modules to describe services, amongst others the service and technical module (Barros and Oberle 2012). The service module describes the service, e.g., in terms of Service, CompositeService and ServiceBundle whereas the technical module describes technical interfaces, e.g., described c.f. WSDL.

Microsoft proposes an ontology and taxonomy (OT) of “software services” (Cohen 2007). The ontology defines the characteristics of services: main purpose, interface, state management, transactions, error handling, security, management/governance, how built. The taxonomy provides a classification of services: Bus Services, Communication Services, Utility Services, Application Services, Entity Services, Capability Services, Activity Services, and Process Services.

3.2 Integration

These perspectives must be integrated to provide a clear blueprint of the composite SVN. Approaches such as e 3 alignment provide a valuable foundation to achieve such integration since it enables to design and analyze inter-organizational business-ICT taking into account four perspectives: strategic, value, process, and IS interactions (Pijpers et al. 2012).

Other research areas such as enterprise interoperability (EI) (van Sinderen 2008), cloud computing (CC) (Mell and Grance 2011), and Business Services Modeling (BSM) (Amsden 2005) can also provide insights to achieve this integration. EI, also referred as enterprise computing, aims at providing to enterprises the ability to interoperate for achieving their business goals while taking care of the cross-organizational, cross-jurisdictional, and cross-domain nature of business collaborations (van Sinderen 2008). In this way, enterprises must not only align business goals with ICT infrastructure internally but also enable effective and efficient inter-enterprise collaborations (van Sinderen 2008). The business perspective is usually addressed as a goal or value modeling problem, whereas the process perspective is mostly solved by applying choreography and orchestration of business processes. Finally, the ICT perspective relies mainly on software (web) services (van Sinderen 2008; Khadka et al. 2013).

CC “is a model for enabling ubiquitous, convenient, on-demand network access to a shared pool of configurable computing resources” (Mell and Grance 2011). Although CC is well supported by academia and industry, there is limited information on how CC deals with business aspects (Commission 2012). Some industrial approaches take a managerial perspective by proposing the use of the IT Infrastructure Library (ITIL V3), which addresses service management as the capability for providing value to customers in the form of (IT) services (IBM 2010). Other approaches propose the use of strategies such as Enterprise Service Bus (ESB) (Chen 2012).

Finally, IBM’s BSM aims at bridging the gap between business requirements and the architecture of IT solutions that are intended to meet them (Amsden 2005). BSM uses “Business Services Specifications” (BSS) that can mediate between business operational requirements (business processes represented as collaborations in UML2) and the final implementation by providing a specification of what those implementations must do to meet the requirements (Amsden 2005). Table 1 summarizes how USDL, OT, e 3 alignment, EI, CC, and BSM can provide insights on SVN’s multiple perspectives as well as on integration issues.

Table 1 Multiple perspectives on SVNs, integration strategies, and current approaches
Full size table

3.3 Market-based composition of SVNs

SVNs might be composed in at least two fashions: hierarchical and self-organized. We have extensively explored hierarchical approaches where customers drive the composition of SVNs in a top-down manner (Baida et al. 2003; De Kinderen et al. 2013; Razo-Zapata et al. 2012). Nonetheless, self-organized approaches might be also suitable for this task. For instance, multi-agent systems can be applied to provide a bottom-up composition (Shoham and Leyton-Brown 2008). In this way, the composition task can be modeled as a coalitional game theory problem in which service suppliers build up coalitions (e.g., service bundles) to achieve a goal (e.g., obtain good payoffs while solving a customer need) (Shoham and Leyton-Brown 2008).

Moreover, we also consider that the decision on whether to apply hierarchical or self-organized composition depends on the nature of the market, i.e. self-organized composition could be applied to highly complex markets where there is a huge diversity of customers and suppliers. For instance, Basole and Rouse have identified that fragmented markets such as retail and telecom are much more complex (i.e. there is a big amount of participants, customers and suppliers) than consolidated markets such as aerospace (Basole and Rouse 2008). In this way, hierarchical approaches may be applied to markets such as aerospace as there are relatively few aerospace suppliers compared to retail and telecom, for which self-organized approaches could be more suitable since the complexity is higher.

3.4 Resilient SVNs

Once SVNs are deployed by means of ICT services, we must monitor its performance to avoid undesirable situations such as drops in Quality of Service (QoS), decrease in value, or unavailable suppliers, which may diminish its performance. Our final goal is to provide SVNs with resilient capabilities to adapt to changes. To this aim, we must deal with at least the following issues:

  • Monitoring We must continuously monitor SVN’s performance to detect changes that can degrade SVN’s quality or value. As changes can cause an SVN to become completely inoperable or partially compromised, we would like to monitor whether value transfers are achieved as expected.

  • Recomposition If an SVN becomes inoperable, recomposition is required to either build up a new fully functional SVN or an operational SVN with “nearly acceptable” performance. E.g., a recomposed SVN could provide a borderline QoS at lower cost, which may still satisfy the customer need.

  • Adaptation Whenever SVN’s performance is affected (due to recomposition or degradation), we must adapt it to meet again customer required specifications. While adaptation may require changing/replacing few services, we believe recomposition may require building up new SVNs.

Finally, we believe that recomposition and adaptation can be addressed either as a design problem (Razo-Zapata et al. 2012) or as a (re)configuration design problem where all possible solutions (recompositions/adaptations) adhere to a common solution template (Motta 1999). For instance, in Razo-Zapata et al. (2010), we have already proposed a mechanism to compose SVNs based on skeletons (templates that capture knowledge about SVN’s structural properties).

4 e 3 service as an Example of SVN Composition

Research around SVNs is vast and deals with issues like value co-creation, business and ICT alignment, modeling, automated composition, among others (Basole and Rouse 2008; Hefley et al. 2010). The e 3 service ontology semi-automatically composes SVNs matching customer needs. The e 3 service consists of two parts: (1) knowledge representation of customer needs and available services, and (2) reasoning about composition.

4.1 Knowledge Representation to Address Epistemic Complexity

Customers consider services from a customer need perspective, whereas suppliers think of services in terms of packages offered to the customer. The e 3 service ontology uses the notion of functional consequence (FC) to match both worlds. A consequence is any kind of result (physiological or psychological) acquired by consuming a valuable resource (tangible or intangible) (De Kinderen et al. 2013; Razo-Zapata et al. 2012). Customers obtain valuable consequences to satisfy their need; suppliers offer one or more consequences as a result of service provisioning.

The e 3 service customer need ontology describes customer needs using terminology from marketing (De Kinderen et al. 2013). Key notions in the ontology are need, consequence and want (De Kinderen et al. 2013). A need is a basic human requirement that can be specified (concretized) by means of one or more consequences, whereas a want is a commercial solution containing consequences that can be provisioned on its own (De Kinderen et al. 2013). In the latest work on e 3 service we focus mainly on functional consequences (FCs), which represent the functional goal that can be achieved through consumption of a valuable resource (Razo-Zapata et al. 2012).

Example In Fig. 1b the customer need “How can I get entertainment while at home?” can be specified by specific FCs such as connection to Internet and online basketball that might be respectively contained in wants such as Internet access and FBA Stream, which are commercial solutions (Razo-Zapata et al. 2012).

The supplier-oriented part of the e 3 service ontology utilizes e 3 value (Gordijn and Akkermans 2003) and considers service suppliers as actors performing value activities (commercial services) that produce service outcomes (value objects), with consequences, which can be offered to customers.

Example Fig. 1b depicts the actor “FBA” performing the value activity “Basketball Pass” that produces the value object “FBA Stream”, which contains FC2 and FC3 (Razo-Zapata et al. 2012).

4.2 Computer-Supported Composition to Address Inherent Complexity

To reason about service bundles, Baida et al. proposed an algorithm that automatically bundles services using customer needs and dependency relationships among services (Baida et al. 2003). De Kinderen et al. used a problem-solving method for matching customer needs to pre-composed multi-supplier service bundles (PCM2) (De Kinderen et al. 2013).

Recently, Razo-Zapata et al. (2012) have applied the so-called Propose-Verify-Critique-Modify (PVCM) problem-solving method to compose SVNs matching customer needs. PVCM removes two important assumptions made by Baida and De Kinderen. First, since Baida’s bundling algorithm focuses on services and their interdependencies, the bundling process is limited by the number of optional services to be bundled (i.e. it assumes a small number of services Footnote 3) (Baida et al. 2003). Second, De Kinderen assumes that service bundles are pre-composed (De Kinderen et al. 2013). In PVCM, the propose subtask automatically bundles services (no pre-composed bundles anymore) matching a customer need and assumes a bigger search space by focusing on FCs rather than services (Razo-Zapata et al. 2012).

5 Critical Reflection on e 3 service

5.1 Multiple Perspectives

Since e 3 service deals mainly with a business value perspective on services, issues regarding business process and ICT perspectives are mostly overlooked. Some efforts have been done to combine e 3 value models with process-oriented approaches (e.g., Schuster and Motal 2009; Fatemi 2012). However, the link between value and ICT perspectives remains highly unexplored (see Sect. 6.2).

In addition, regarding the value perspective, since e 3 service can compose SVNs combining single services with unique pricing schemes, it must also provide ways to compute consistent pricing schemes for bundles composed of several services (see Sect. 6.1). In De Kinderen et al. (2013), a pricing ontology has been proposed for pre-composed bundles, which can be extended to use with automatically composed SVNs.

5.2 Integration

Since both e 3 alignment and e 3 service are compatible with and inspired on e 3 value, it is reasonable to use them together to integrate value, process and ICT perspectives. However, since e 3 alignment deals mostly with the integration of the value and process perspective, some technique(s) must be developed to integrate the ICT perspective (based on web services perhaps).

5.3 Market-Based Composition

The e 3 service ontology currently supports at least three approaches on hierarchical composition of SVNs, namely Serviguration, PCM2 and PVCM (Baida et al. 2003; De Kinderen et al. 2013; Razo-Zapata et al. 2012), which have been applied to different sectors such as music, education, health care, energy among others (Razo-Zapata et al. 2012). Nonetheless, e 3 service may be enriched by self-organized approaches to address issues such as scalability (i.e. how many services can be considered to compose SVNs), which has been previously observed by Serviguration (Baida et al. 2003) and PCVM (Razo-Zapata et al. 2012) (see also Sect. 6.3). In addition, Gordijn et al. (2012) argue that a self-organized approach has also a business consequence, namely that selecting the actors providing the services is not done anymore by a powerful central actor, but by the market itself.

5.4 Resilient SVNs

Although our current vision on resilient SVNs is composed of monitoring, recomposition and adaptation, e 3 service mostly provides solutions only for recomposition since approaches such as PVCM can recompose SVNs based on feedback provided by customers about required FCs (Razo-Zapata et al. 2012). As e 3 service has no mechanisms to address neither monitoring nor adaptation, extensions must be made to fully provide resilient SVNs.

Based on our long-term vision, Table 2 summarizes which issues have been solved, partially solved and unsolved in e 3 service. For instance, the value perspective is solved as e 3 service provides a sound mechanism to model the exchange of valuable resources, which meets our vision (see Sect. 3.1.1).

Table 2 Summary of solved, partially solved and unsolved parts
Full size table

5.5 Other Approaches

First, the e 3 value and e 3 service ontologies can be used as a business development tool. A group of business developers and/or consultants then uses the ontologies as a conceptual modeling instrument, to create a shared understanding of the business case a hand. The Business Model Ontology (BMO) (Osterwalder 2004) (later: Business Model Canvas (BMC) (Osterwalder and Pigneur 2010)) is an approach for business model development. In contrast to e 3 value, BMO takes a particular enterprise as a point of departure, and considers, amongst others, the business network surrounding this enterprise. The e 3 value makes no explicit distinction between suppliers and customers but treats parties as equal citizens, thereby taking a business network perspective. The Resource Event Agent (REA) ontology (Geerts and McCarthy 1999), added with a graphical syntax (Sonnenberg et al. 2011), focuses amongst others on resources. REA provides a generic view on economic activity, in terms of resources. The e 3 service ontology adds to such a generic view a specialization about the notion of services, and modeling and reasoning about multi-supplier service bundles that satisfy a complex customer need.

Second, the e 3 value and e 3 service ontologies may facilitate semi-automatic reasoning about SVNs. Currently, such reasoning is useful for developing and understanding the service catalog of an enterprise (and its partners). Tool support can be used to understand which alternative service bundles are possible given a customer need as these tools can enhance the design of process models based on value exchanges (Razo-Zapata et al. 2012; Fatemi 2012) and the analysis of economic issues such as profitability (Fatemi 2012).

6 Research Agenda

6.1 Pricing Complex Bundles

Theme Multi-perspective.

Description Although service bundles are composed of one or more single services, they are sold for a single price. Therefore, it is necessary to compute final prices for bundles as well as the amount of money to be distributed among each single service within bundles. In addition, some observations have to be taken into account (e.g., De Kinderen et al. 2013; Monroe 1990). First, the price of a single service can be fixed, usage-based or a combination of both. Second, discounts may apply to the prices of single services when bundled.

Example In Fig. 1b the service bundle contains four services (provided by different suppliers) for which the customer must pay some money. Consider for instance that GS Plus, At Your Help and Basic Internet services are offered at a fixed price, whereas Basketball Pass is offered under a usage-based scheme (perhaps per viewed game). In addition, when acquiring Basketball Pass combined with other services, discounts apply. To compute a final price for a given bundle, we must compute all pricing schemes as well as determine under which conditions discounts might apply. Finally, the money being paid by the customer must be repartitioned among suppliers.

Foreseen solution To address this issues we can rely on research previously conducted by De Kinderen et al. who have already proposed a pricing model for pre-composed bundles (De Kinderen et al. 2013), USDL that provides a pricing module for services (Barros and Oberle 2012), or Becker et al. who have designed a language to price bundles containing products and related value-added services (Becker et al. 2011).

6.2 Value and ICT Duality

Theme Multi-perspective and integration

Description Even though the link between value and process perspectives has been largely explored (see Schuster and Motal 2009; Pijpers et al. 2012; Fatemi 2012), the link between value and ICT perspectives is usually overlooked. Nonetheless, value and ICT perspectives must be described and integrated to deliver fully operational SVNs as commercial services will then contain information regarding value, process and ICT issues.

Example In Fig. 1 services offered by suppliers and enablers provide information about value issues such as business to customer (B2C) and business to business (B2B) relationships (value transfers between the customer, suppliers and enablers); nonetheless, there is no information regarding what type of ICT services can be used to realize such services. For instance, there must be information on what kind of ICT services can be used to provide the service “GS Plus” provided by “Green Shark” since many technologies can be used, e.g., web services for streaming or downloading video.

Foreseen solution In Rolland et al. (2010), the authors present an MDA-based approach for linking software services (mainly web services with WSDL and BPEL descriptions) to so-called intentional services that might meet business-oriented requirements such as goals. We can follow a similar approach in which value-oriented service descriptions are linked to operational services by using MDA techniques. For instance, we could define a metamodel capturing concepts of value and ICT models; the metamodel could be used to define model transformations for linking value-oriented services to ICT services.

6.3 Service Bundling Meets Combinatorial Optimization

Theme Market-based composition.

Description To provide e 3 service with a toolkit to compose SVNs for different types of markets, we must first address scalability issues for our hierarchical algorithms (Razo-Zapata 2014). Although our current algorithm can handle a considerable number of services (up to 1000), they cannot deal with high numbers of customer requested functionalities (Razo-Zapata 2014). In Razo-Zapata (2014) we have explained that composing SVNs might take up to 80 s for several requested functionalities.

Example Our current algorithm can handle service catalogs containing 1000 services which provide on average ten FCs; however, once a customer requests more than 15 FCs the composition takes a considerable amount of time.

Foreseen solution Since bundling services aims at building up a combination of services to match a customer’s need, to improve the scalability of our algorithm, we can model the bundling problem as a combinatorial optimization problem. In this way, approaches such as parallel computing (Baase and van Gelder 2000) can be applied. For instance, since in Razo-Zapata (2014) we have already defined a bundling process based on computing and combining service clusters to generate solutions, a next step might be to use parallelism so that service clusters are computed and combined faster.

6.4 SVN Monitoring

Theme Resilient SVNs.

Description As SVNs are exposed to changes due to its complexity, we have to provide ways to monitor their behavior and detect when changes compromise their performance.

Example The SVN in Fig. 1b is exposed to changes in at least three contexts (value, process, and ICT). For instance, suppose that the “Basketball Pass” service is realized by a video streaming web service that suddenly becomes unavailable; as a consequence transferring the resource “FBA Stream” to the customer is not longer possible due to the technical failure. Therefore, a monitoring scheme must allow to detect such failure so that an adaptation strategy can be applied.

Foreseen solution e 3 service can try to reuse previously developed ontologies such as e 3 control (Kartseva et al. 2009) or the so-called Value Monitoring Ontology (VMO) (Silva and Weigand 2012), which both provide methods for monitoring networks of activities based on e 3 value constructs. In addition, the proposed solution must also address detection as it is important to know when SVNs are experiencing deviations from expected behavior, i.e. changes about QoS, decrease in value or unavailable suppliers. For instance, VMO offers the possibility to define monitoring policies for critical activities within an SVN. In our example, it would be possible to define policies for services such as “Basic Internet” and “Basketball Pass”.

By no means, the list of challenges is complete. Issues such as software support for integration and SVN adaptation are not discussed due to space restrictions. In addition, more challenges have already been described in (Gordijn et al. 2012), such as ontological issues for matching customer and suppliers, detailed mapping from customer needs onto specific requirements and web service support for e 3 service. Furthermore, other challenges can be found by considering Service Dominant Logic (SDL) (Vargo and Lusch 2004), e.g., by taking into account the role of the customer as service co-producer.

7 Conclusions

In this paper, we have introduced a number of research challenges with respect to commercial service value networks in general and e 3 service in particular.

The contribution of this paper is threefold. First, we shared our long-term vision on SVN composition, where we also discussed overlaps and connections with other approaches such as USDL, IBM’s Business Service Modeling (BSM), cloud computing, and enterprise interoperability.

Second, we also presented our experiences on value-driven service composition and provided a critical reflection by describing the state of the art of e 3 service. We explained how addressing epistemic complexity has improved our understanding of services and SVNs. We briefly described the two perspectives in e 3 service (customer need and service supplier) that allow us to design, compose and analyze SVNs. Likewise, we also described different composition approaches (Baida et al. 2003; De Kinderen et al. 2013; Razo-Zapata et al. 2012) to handle business to customer (B2C) and business to business (B2B) relationships, which help us to address inherent complexity.

Third, we described a research agenda to push e 3 service closer to our long-term vision, which encompasses four elements: a multi-perspective view on SVNs, integration, market-based composition of SVNs and resilient SVNs.