1 Introduction
Economic value networks are reported to be complex, dynamic and adaptive heterarchic systems, insofar as multiple heterogeneous and mostly legally independent companies (the actors or participants of a network) are tied together in highly interconnected and geographically wide-spread value creation processes [1–3]. However, the term ‘heterarchic’ is used diffusely and heterogeneously since years [4]. The concept of heterarchy was first introduced by [5], who disagreed to the previous concept that human brains are structured in a strictly hierarchical order, only being able to process in a sequential logic. Instead, he introduced the concept of a network-like nervous system, that enables parallel task processing. This concept was applied analogously in the field of economics and management, e.g., by Probst [6], who proposed the idea of fluctuating hierarchical relationships, that could adopt their structure according to situational requirements. In this context, the discussion of hierarchic versus heterarchic managerial structures needs to differentiate between a cross-company value network perspective (no central source of influence, thus only few hierarchical characteristics, unless the respective supply chain incorporates, e.g., a monopolistic supplier or customer) and a company-internal perspective (here, the extent of heterarchic attributes differs, depending on the implemented organizational structure).
Purely hierarchical systems have a unique root point that is the sole source of decision power, e.g., the owner or CEO of a company. All system elements, e.g., all company employees, are linked to this superior element, e.g. the respective CEO, as a direct subordinate or indirectly via several hierarchical levels—in this example, the organizational structure of the company. In a hierarchic organisation leadership is based on the managerial power and activeness of this superior role. In contrast to this, heterarchic structures rely on several stakeholders with distributed power who co-operate with each other in manifold ways. In this case, leadership also includes negotiation and decentral autonomy. Instead of a single source of power, decisions could be taken by either one or another organisational unit, or even within a joint team—depending on the problem to be solved in each case and the expertise required therefore. This is practically implemented by means of predefined rules and responsibilities. For example, the sales force of a company might decide on the question whether a certain customer is considered to be strategically important or not. However, the production units might take utilization decisions. Still a central management board or CEO could be the single point of power, having delegated this power to subordinated units. The resulting organisational constellation could be a hybrid between heterarchy and hierarchy, or could even be close to a hierarchical system. This depends on the question to what extent the principal would still be the central source of power behind operationally decentralized (i.e., delegated) decision responsibilities. The less influence the central principal has got over the involved decentral units, the more the organisational structure could be regarded to be a heterarchy. In the absence of a central and singular source of power, centralized (more or less authoritarian) decisions would be substituted by other decision mechanisms, e.g., negotiation, voting or team consensus. According to situational requirements, varying organisational units would co-operate with each other. Respectively, some authors assume the necessity of an overall hierarchical structure that could be supplemented with heterarchic structures on demand, in case a problem would require a high degree of flexibility and local autonomy [7]. This hybrid organizational constellation is called a ‘heterarchic hierarchy’ [8].
Summarizing, the balance between both poles—pure hierarchy or pure heterarchy will depend on the actual distribution of property rights [9]. Practically spoken, an organizational structure with a high degree of decentralization, self-monitoring and autonomy can be assumed to be a predominantly heterarchic system [10]. Accordingly, the acting agents will be responsible for all activities and decisions, and will co-ordinate the major share of their information management and decision taking on their own account. Comparable to the heterarchically negotiated structure of supply chain processes, also the underlying information systems are increasingly able to process data in reticulate structures. This can, for example, be based on software agents with negotiation skills. Agent-based models could, for example, be used for the purpose of selecting supply chain partners, in particular enabling the analysis of agent behaviours throughout the bidding process [11].
From a value network perspective, in particular the availability of capable cross-company influence mechanisms has gained importance, as company success increasingly depends on suppliers, service providers and customers, as well as on other company-external parties, e.g. public authorities. This value network consists of several legally independent companies and can be considered to be a heterarchic structure by definition, as there is no central “supply chain authority”. In this regard, SCM is not fully comparable to the management board or CEO of a company. Instead, companies are negotiating service and goods deliveries among each other, being determined by fair-scattered legal, political, cultural and social, technical, environmental or market-related influence factors.
Many companies have concentrated on their specific core capabilities, having externalized tasks with lower relevance or far distant from their core competences by means of outsourcing practices since many years [12]. As a consequence, the size and the boundary of singular companies have changed [13] towards more and smaller companies that are linked to each other within a specific supply network section by means of delivery contracts. Also, the overall network-wide proportion of tasks being coordinated hierarchically (inside a company) and tasks being coordinated through negotiation and contracting (company internally, or across company borders, at most within dyadic seller-buyer relationships) has shifted towards fewer make-decisions (i.e., hierarchical coordination) to the favour of more buy-decisions (i.e., market-based coordination; cp. in a supply chain context [14]. Furthermore, also the adequate amount of hierarchical coordination is a matter of discussion in company-internal organizational concepts, as more autonomous concepts like e.g. self-directed work teams, seem to handle a complex business better [15]. Altogether, it can be assumed that the extent of a heterarchic distribution of power has increased in current business environments.
In line with this assumption, the present chapter intends to contribute answers to the questions of (1) ‘What is heterarchy and why does it matter in SCM?’ and (2) ‘Could hierarchical layer models, as frequently used in SCM, informatics and in automation, still serve for the purpose of achieving a holistic model of heterarchic systems in the context of digital manufacturing?’. The remainder of the chapter is as follows: Sect. 2 will briefly define SCM from a value network perspective with regard to the characteristics and principles of heterarchic systems. Subsequently, Sect. 3 shortly explains how hierarchical layer models are applied on the SCM and business process level as well as on the underlying information management perspective. Interestingly, the SCM- and the business process layers are undergoing similar shifts towards higher decentral autonomy and increasingly heterarchic structures, as the IT- and automation layers. Accordingly, Sect. 4 discusses the role of service-oriented architectures (SOA) and modern IT techniques, with a specific focus on current smart manufacturing and digital production concepts in industrial environments. We conclude in Sect. 5 with possible managerial implications of these converging trends for supply chain management that is based on a digitized manufacturing environment.
2 Heterarchic Supply Chain Structures for Dynamic Market Conditions
Heterarchic supply chain structure [19]
This network structure represents a heterarchic system in the following sense: every company arranges a comparable picture with regards to its business, positioning its own operations (or eventually even more than one business field) in the central position of the network topology in Fig. 1. This might as well happen unconsciously, in case a company doesn’t explicitly draw this structure, but only implicitly builds respective contracts with customers, suppliers and service providers. The resulting constellation is can’t be as easy described as the frequently drawn assumptions that “supply chains compete with supply chains” [20], as this would assume a supply chain to be an acting entity. This is typically not the case, as it would require a central authority or extensive negotiation processes with the joint target of ‘optimizing the supply chain as a whole entity’. Contrariwise, companies would rather enforce their individual objectives. Still, companies with mutual alliances could factually establish ‘supply chain sections’ (e.g., consisting of two, three or sometimes more supply chain partners) via closely linked supplier-, customer-relationships or within collective companionship contracts. The closer the interdependent integration and managerial cohesion between them, the more the system will show a heterarchic behavior with syndetic supply chain sections (simplified supply chains).
Efficient and sustainable management requires the resolution of occurring conflicts in the face of limited resources and differing targets. However, the control and solution mechanisms that would support a manager to resolve these conflicts are rather different in hierarchical structures compared to heterarchic structures – no matter whether a certain conflict situation applies to a company-internal or to a cross-company setting or both. The more, a system can be considered to consist of comparably independent agents with strong self-interest, the more likely it is that decision rights and required information are distributed within the system. None of the acting agents can be sure to be completely aware of each other’s intentions, objectives and operational action. As a consequence, authoritarian control ratio and related formal decision mechanisms (e.g., linear optimization algorithms within production planning) will fail to act. The same applies to situations characterized by information lacks—a rather usual state of things in highly dynamic and changing environments. As a consequence, managerial efficiency can be assumed to depend on mutual capabilities to exchange relevant information, to negotiate for working co-operation constellations and to resolve occurring conflicts. In doing so, one must be aware of the fact, that even the most efficient settings could be impaired through principal agent deficiencies (in particular hidden intentions, hidden characteristics and hidden action of agents, see for example [21] or more recent [22].
From the focus of each single company respective managerial methods, in particular supply chain management practices [23], have to be applied to adequately design, operate and optimize material-, information- and financial flows. Such practices usually imply integration and coordination issues. Thus, the company-internal value creation and the attached external material supply processes are aligned towards the objective to achieve a delivery performance, a selling price and a flexibility that is competitive in the respective industry segment and/or geographical region. Typical inhibitors are low predictability, incomplete information and opportunism [24, 25], asymmetric market power between companies that are part of the value network [26] and highly fluctuating demand progressions [17] together with extensive and volatile product variety [27].
There is evidence that the concept of heterarchy is neither been sufficiently discussed and understood in an SCM-context, nor has been adequately transferred into managerial practices [28, 29]. Unfortunately, concepts related to suchlike system theory oriented questions often are of abstract, sophisticated nature, which handicaps a transfer into practical usage [30]. In most cases no scientifically proven and practically feasible managerial guidance is provided regarding the question at how to achieve a well-working balance between opposing concepts, such as for instance central control versus decentralized autonomy [15, 31] or the exploration of innovations and new capabilities versus the efficient, reliable and standardized execution and improvement of existing processes [32] or reliable forecasting and planning versus a high instantaneous flexibility [33]. SCM typically describes such coherences as disparate dichotomies and—at best—provides segmentation criteria to differentiate adequate managerial policies.
Logistics and SCM theory have provided several conceptualizations regarding the nature of value networks and their attributes [3, 34] or [35], and have also investigated topological questions. For example, Lambert et al. [36] or Gosling [37] have provided generic value network models, emanating from a focal company and have subsequently modelled attached supplier and customer structures. However, practical circumstances often require a more elaborate consideration: simplified models for example often neglect the fact that each participating company establishes its own subjective value network conceptualization. This regards the valuation of the reciprocal importance within dyadic buyer-seller relationships: For example, a purchaser can rate a certain supplier as low priority ‘C-type’ supplier, whereas the supplier might perceive this customer as important ‘A-type’ customer (and vice versa). Therefore, a value network can be considered to be an aggregation of mutually differing value network notions [19] that are, besides, shifting over time. Further, the particular capability of each participating company to influence its environment according to its market power will be different among the network and may again vary over time according to developing technologies, logistics capabilities, selling and purchasing volumes or supported through further monopolistic (unique) or oligopolistic (rare) market constellations.
Typical performance targets, a company has to meet under such in-transparent and permanently changing conditions through the application of SCM concepts and practices are short lead and reaction times, a competitive economic position, a distinct operational flexibility [38] and a proficient capability to adapt internal processes and related supply flows to structural changes [39]. Relevant challenges for SCM arise from the fact that the coordination of economic value networks is reliant on accurate and exhaustive information [40, 41]. However, in practice the lack of available information and the mutual capability to exert influence upon each other are causing huge constraints, compared to the ideal state.
A further difficulty derives from the fact that value networks are regarded to be complex adaptive systems (CAS, see [31]). Whereas the original development of SCM concepts and practices took place under rather stabile economic circumstances, actual markets are showing turbulences and unpredictability [17]. Therefore, existing SCM methods and tools have to be developed further towards a point where they can be applied easily and provide for a fast and repeated execution within fluctuating conditions, but without losing explanatory power and validity due to oversimplification [42].
‘Traditional’ reductionist approaches to handle erratic changes are typically seeking to smooth turbulences and to best possibly predict residual unavoidable fluctuations. Respective control means require accurate and timely information to increase value network transparency [17]. Corresponding SCM practices are for instance order smoothing [43] or co-operative approaches like collaborative planning concepts or vendor managed inventory agreements. This is however stated to be inopportune, if applied in volatile environments—especially if there are notable impacts on the focal company that originate from tier-2 to tier-n network participants.
Thus, holistic approaches related to CAS and system theory have recently gained more attention also in a SCM context since long [44]. Respective approaches typically assume that a complex dynamic system is characterized through extensive relations and interdependencies, non-dynamic coherences, self-organizing behavior patterns and the occurrence of emergence phenomena. All these attributes also apply to economic value networks. Hence, a main paradigm shift in SCM is that the system behavior can’t be necessarily explained from the analysis of its singular parts, but emerges either due to unexpected external impacts or as an inherent part of the internal system behavior that was neither observable nor predictable from the partial analysis of its single constituents [15]. In addition to simple linear coherences also circular causalities play a major role [45]. This is well-known in systems theory, time and again in principle discussed in SCM related theory, but has scarcely been operationalized for practical SCM transfer [30, 17]. One evident example for this is Ashby’s law of requisite variety [46]: At first, universal sight, Ashby’s proposition is obvious. At second, also generic principles can be deduced for SCM, such as ‘Develop an adequate product variety towards your target markets: not too high, to avoid excessive efforts and complexity, but not too low, to avoid lost sales in favor of competitors.’ However, the third and operationalizing step has remained unsolved: management approaches to quantitatively determine and achieve the ‘adequate complexity or variety’ have not yet reached a sufficient maturity [19].
These coherences have major consequences for SCM that have not yet found a broad propagation in corresponding methods and practices: reductionist partial analyses provide useful means, but may not always be expected to deliver exact and stabile evidence over a long time-span. They rather provide snapshots that are eventually clear and precise but possibly diffuse and faulty. Although the common assumption that logistics capabilities can be classified into demand-management and supply-management capabilities, together with an emphasis on interface and information management proficiency [47] is valid in principle, this approach doesn’t go far enough in complex turbulent environments: An additional body of dynamic capabilities, such as for instance the transfer and practical application of decentralized theoretical approaches (e.g., the long-proposed theory of loosely coupled systems (for a reconceptualization of this concept, refer to [48]) into the context of value networks is strongly required. Further, it is important to develop practical means that support the evaluation and handling of dichotomies in the sense of ‘dynamic, fragile balances’—e.g. between control and emergence, though a main difficulty lies in the fact that a suchlike emergent nature is not reducible to the characteristics of the network elements and therefore can’t be predicted ex ante, but only becomes explainable ex post [15].
Altogether the theoretical base of SCM is contradictory and has not yet been developed to a point where it provides sufficient explanatory power and decision support regarding the design and operation of value creation flows in complex, volatile and unpredictable environments. In practice the proportion of hierarchically coordinated areas in value networks have decreased as a consequence of outsourcing. Further, also the adequate amount of hierarchical coordination is a matter of discussion in company-internal organizational concepts, as more autonomous concepts like e.g. self-directed work teams [15], virtuality concepts [41] agent-based modelling approaches [11], big data concepts [49] or concepts that involve elements of artificial intelligence techniques [42], seem to have the potential to handle a complex and unpredictable business better than conventional SCM- approaches that depend on planning reliability. Obviously managerial means beyond and supplementing traditional hierarchical and reductionist means are needed. In this context it is a noticeable observation, that the relevant body of literature has not only emerged recently: Most theoretical concepts have been introduced years ago (cp. the literature cited in this chapter as a characteristic—though on no account exhaustive—overview). Comparably, the related managerial challenges have been uncovered long ago. Still, operating answers aren’t developed up to a sufficiently mature state. Obviously, ‘hierarchy’ is a useful structural principle under stable and predictable conditions. However, recent dynamic and unforeseeable business conditions can’t be handled well by means of a purely hierarchical perspective. Hence, several authors (for an overview see for instance [50] have proposed frameworks that are based on heterarchic strategies and principles. These contributions claim decision making to be far less centralized and hierarchical, but rather autonomous and linter-linked across independent companies compared to past business models and organizational structures. With regard to question (1) it can be stated that heterarchic concepts definitely matter within supply chain management, once only the issue of operationalizing the abstract concept with regard to concrete applications and competitive business advantages.
3 Hierachical Layer Models in SCM, Information Management and Factory Automation
Evolving data models within the field of informatics [51]
Evolving data models within the field of informatics [51]
An important concept within suchlike layer models is the principle of encapsulation, which is however not fully applicable in a context that also claims to model human behaviours—not only deterministic computational coherences. Still, such models can serve well in order to reduce complexity, and can be applied as a descriptive architecture model for further formalization. There are obvious similarities between data models and the organizational structures with regard to the division of labour in companies and supply chains: a hierarchical topology allows for the rapid processing of well-standardized tasks. This is due to the singular centre of power and the respective uniform chain of commands in suchlike systems. Accordingly, also the communication means in use are unambiguous and well-defined. The disadvantage of hierarchical structures is the inherent difficulty to adapt, together with a low degree of flexibility. These characteristics can be identically observed on both layers—organizational structures and IT-structures. Contrarily, polycentric structures (such as heterarchic systems) are flexible and adaptive towards changes. They allow for fragmented requirement progression with higher performance. Besides, they show advanced capabilities to handle ill-structured problems. On the other hand, theses heterarchic systems are slower and less reliable when it comes to the execution of well-structured and properly defined standard tasks. For example, successful concepts on both layers, IT and SCM, seek to address opportunism problems through reciprocity and both have the inherent ability to solve “flash crowds” through distributed resource allocation [52]; In order to illustrate this assumption for an IT-example, one could refer to e.g., IT-based swarm algorithms. On a supply chain level, respective value network cooperation concepts could serve as illustrating examples: Here, local improvements are often disadvantageous at the superior system level, thus in principle preventing their usage. However, in practice, the additional issues of human decision biases and distributed decision power have to be taken into account. This leads to the effect, that not always the most advisable decision is actually taken and consequently implemented.
Comparable to SCM, also within IT-based networks not only the network size, but primarily the proportion and density of relationships have risen. This goes hand in hand with an increased need for dynamically configurable virtual networks in the course of efficient resource sharing and a focus shift from resource sharing only to content transfer.
The same applies for automation concepts, that also have strictly applied hierarchical layer models in the past ([53]; for a more detailed overview see, for example, the contributions in [54]). Similar as SCM and informatics, also factory automation and industrial robotics have to handle increasing uncertainty and complexity. In this context, in particular industrial machinery and respective control devices (either industrial computers or programmable logic controllers—PLCs) have to be operated and optimized under these changed conditions.
The lowest layer includes equipment components (e.g., a pump, a valve, a drive, a measuring device, …), sensors (either assembled to other components or independent) and actuators (e.g., claws, a drill spindle, a suction device, …). On this field level (also device level), in particular the signal flow between these devices and the transformation of physical quantities into data signals and vice versa the triggering of (end-) effectors via respective data signals has to be managed.
The proximate control layer makes use of sensor information (having originated from the underlying device level) for the purpose of process control and generates control commands to be send to the particular devices (actuators) via the subordinate device level.
Automation pyramid, modified from [53]
Indeed, the hierarchical approach behind the automation pyramid enables the application of heterogeneous technologies within an automation system, e.g., a production plant. However, this approach increasingly fails to act in vague and ambiguous situations.
Currently, due to an increased decentralization of automation and robotics applications the decisional capabilities of automation systems are more and more transferred towards distributed components. Besides, the importance of vertical integration becomes a critical bottleneck, as risen business process and supply chain complexity are demanding a more efficient computational and automation support with less necessity for manual interventions. Especially, the amount and the heterogeneity of information, to be processed in this regard, have reached an immense extent. Furthermore, in the absence of commonly agreed standards, heterogeneous devices (different control device manufactures, several technology generations and maturity levels, manifold operating systems and software applications, diverse country standards etc.) have to be integrated into a well-integrated system. As a consequence, different layer models from several disciplines—informatics, automation, SCM (and most probably further subject areas)—have to be integrated interdisciplinary beyond previous system borders in order to achieve an acceptable adaptability (or technically spoken re-configurability), flexibility and maintainability.
Moreover, current industrial systems need to be resilient on all aforementioned layers: As more and more unforeseen events might impair the system surprisingly, as well technical systems (informatics and automation) as organizational structures (company-internal and related to the cross-company supply chain context) have to be able to rapidly recover from internal failures and external disasters by means of being able to return to the previous, well-performing state or to achieve a new scenario, that is viable under mutated conditions [55]. This, however, requires network-like structures.
The development from purely hierarchical architectures toward a network topology has been repeated in several disciplines. The first to be mentioned in this context is the evolution of database models—at first hierarchical, later with (still hierarchically normalized) relational models, later evolving towards multi-dimensional architectures with enhanced distributed architectures like for instance online-analytical processing techniques (OLAP, e.g., [56]). However even in a distributed OLAP-datawarehouse with highly sophisticated locking and recovery procedures, the basic idea still relies on a hierarchical model. Second, the architectural design of IT-networks and respective communication protocols show a comparable progress that has started with the long-standing hierarchical OSI reference model [57], and has developed towards more recent standards (in particular ISA88-95, NAMUR, TC-184 and ISO; see e.g., [58]). A first decentralisation step was reach in this field with the introduction of client-server architectures, and step by step advanced capabilities at the client-side. Also multiprocessor-systems have provided the capability of less hierarchical information processing in principle. However, these systems are even today widely operated by means of hierarchically control structures and software.
Only slowly, decentralized systems came into broader applications due to advancements in software engineering (e.g., neural network computing or agent-based systems, cp. e.g., [59]) and the increasing need to solve problems that couldn’t be processed easily with common hierarchical systems. However, it must be stated, that these systems generally rely on a hierarchical core structure, when analysing it in the details: As well an x-based software relies on an operating system that is installed on the executing computational unit, and works with hierarchical principles. Moreover, recent simulation studies with regard to Industry 4.0 automation concepts have indicated that within the current state-of-the-art hybrid systems (i.e., a combination of heterarchy and hierarchy) still outperform fully decentralized systems with regard to production planning [60]. Obviously, today even a definitive heterarchy in the end still requires hierarchic roots—at least with regard to the current development stage of Industry 4.0 (also referred to as smart manufacturing, smart automation, digital factory and similar terms). Further research has to be done in order to better clarify this issue.
Last recently, concepts like the Internet of things (IoT) and in German-speaking countries industrial production paradigms like Industry 4.0 show similar tendencies, although it is not yet fully conceivable, towards what technical standards, and to the favour of what novel architectures. The underlying assumption is, that a changed focus away from simple electric signal processing towards web-based semantic services is to be expected [61]. However, apart from the World Wide Web there is no such thing as one unique technology that drives the break-through in digitized automation systems, but rather a bundle of established and novel technologic options that have to be applied in a unique configuration according to situational conditions and requirements [62].
Without any doubt, decentralization is a major trend and is currently implemented in manifold ways, in particular through the use of cyber-physical systems (CPS, enabling real-time data exchange between relevant entities of the automation system [63]. Comparably, pervasive computing concepts are driving the number of (eventually autonomous) network nodes towards an immense amount of ‘smart’ devices. As these decentral entities are able to exchange data and to execute at least simple IT-processing tasks within a network architecture [64]. However, as long as there still remains a central source of authority—here a central server—the true paradigm change towards a heterarchy has not yet fully taken place. When reviewing decentralization concepts more closely, the concept is coupled with heterarchy only on a very loose base: decentralized real-time data processing also provides advantages for fully hierarchical and mixed (‘hybrid’) concepts, for example, when using conventional linear optimization techniques for the purpose of production planning [60].
At the same time many authors are assuming the resolution of purely hierarchical factory automation concepts, in particular at the middle layers of the automation pyramid: previously PLCs and SCADA systems have integrated superior and subordinated layers according to strictly hierarchical algorithms and access modes. Currently, newly emerging technical ‘quasi-standards” like OPC/UA (unified architecture) allow for a more randomized access that may even omit singular automation layers and that at the same time enable higher semantic richness [65]. For example, an industrial robot could be directly controlled without using a PLC and respective middle-layer software.
This raises the question whether the observed trends are essentially contradictory. Revisiting the concept of layer models, this is not necessarily the case: referring to Fig. 1, for example the process level (e.g., looking at production planning and scheduling processes) could be controlled mostly hierarchically, as long as the planning premises are suitable. At the same time, a different layer could be organized more—or even fully—heterarchically, for instance, the communication between physical production devices (machinery, tools, sensors, transportation vehicles, etc.).
As a first tentative assumption that, however, would have to be investigated and validated within further theory-oriented and empiric research, we assume that hierarchical layer models might still have a profound eligibility for the following reasons: to begin with, they could support decision makers to conveniently segment between system partitions that are structured according to an in each case different extent of hierarchy (and respectively heterarchy). Moreover, the general advantage of hierarchical layer models, to partition a complex problem into smaller portions, thus reducing complexity and ensuring interoperability through consequent encapsulation, would remain a beneficial concept also in future systems that have an overall high degree of heterarchic coordination.
A framework for the service enterprise building blocks interoperability [68]
Indeed, the hierarchical approach behind the automation pyramid enables the application of heterogeneous technologies within an automation system, e.g., a production plant. However, this approach increasingly fails to act in vague and ambiguous situations. Frameworks like proposed in [68] allow for a consistent integration of at the one hand supply chain and business processes with at the other hand IT- and automation devices within the so-called ‘service enterprise’ in the ‘service society’ [69]. Moreover, also co-operation between value network partners can be facilitated by means of highly advanced services. This advanced services, in turn enable the evolution of heterarchic structures as well company-internal as across company borders.
The main idea behind SOA is the definition of all tasks, functions and business transactions as independent services. It is most important to optimize the condition-monitoring and the control of the underlying manufacturing devices, i.e., the physical manufacturing activities, in the course of for instance a production process, in order to support the required supply chain agility (and as well other supply chain objectives) best possibly. Here, heterarchic architectures offer promising improvement potentials by means of more autonomous and ‘intelligent’ devices with local decision capabilities [58]. Accordingly, critical success factors are not only the capabilities of the used physical devices (tools, sensors, actuators, measuring instruments, etc.), but also a proficient IT-integration and communication. In the current dynamic market, companies have to be able to collect and analyse unstructured information that is derived from widely distributed channels (i.e., multi-channel interaction with a high amount of customers who generate huge data volumes, see [70]). Subsequently, adequate conclusions, managerial decisions and activities have to be deviated across all layers—from the process level down to the field level. The underlying architecture either could be strictly hierarchical, as currently implemented in the majority of existing enterprise architectures on process-level, IT-level and automation level.
For example, an agile customer-to-order process could require the capability to directly impact a technical device that configures a 3D-printed part: The customer uploads a respective CAD- or printer-file or is allowed to choose from a database, that is offered by the company. This is a feasible business concept also within a hierarchical layer model, but might involve many abstraction layers, and hence be slow due to the high calculation efforts, when re-calculation through all layers strictly hierarchically. A more heterarchic architecture could skip layers that are not needed (because not influenced) for this operations. This however needs a well-substantiated understanding of respective technology impacts. For the (simplified) example of the customer-configured 3D-printed part production, for instance a different geometric structure will, e.g., influence cycle time and material consumption, and hence the order scheduling on super-ordinated layers; a different colour might not have this influence at all.
Altogether, developing theses architectures towards a more heterarchic structure is appealing on all layers as soon, as the existing, strictly hierarchical structure becomes too rigid to handle complexity and fuzziness. However, this requires a deep knowledge among the modelling experts as well regarding the users perspective (process level) as the concerned data structure(s), IT-infrastructure, automation and manufacturing technologies, including adjacent fields of knowledge like material sciences. Regarding question (1), we accordingly assume that heterarchic concepts have a high potential for the improvement of SCM concepts, but have to be developed with suitable cautiousness and diligence. Proceeding, it does not seem to be likely that heterarchy is going to completely substitute hierarchy according to the current state of the art knowledge. This means—concerning question (2)—that hierarchical layer models, as frequently used in SCM, informatics and automation, will still serve for the purpose of achieving a holistic model of heterarchic systems, at least in parts of the system or within highly standardized and predictable applications that don’t require advanced capabilities.
4 The Role of Modern IT for the Management of Heterarchic Supply Chain Structures
Section 4 discusses the role of modern IT techniques, with a specific focus on current smart manufacturing and digital production concepts in industrial environments. As the previous sections have shown, interestingly, the SCM- and the business process layers are undergoing similar shifts towards higher decentral autonomy and increasingly heterarchic structures, as the IT- and automation layers.
Shroff [71] has proposed six elements that constitute the ‘intelligent web’: looking (looking ‘at’ information in terms of sensing and looking ‘for’ information in the sense of searching, especially within the web), listening and learning (i.e., retrieving the most relevant insights from this collective, digitized knowledge), connecting (these information fragments), predicting (what is likely to happen next in the relevant context) and correcting (in the sense of improving). Hence, we perceive major changes in the manner of data access—starting with humans who increasingly navigate through hyperlink-structures instead of sequentially and hierarchically organized information to the point of automatic data access on all levels and also between all levels of the automation pyramid. While stable situations (e.g. manufacturing objectives and conditions) can be handled satisfyingly with information systems that rely on a single entry point, dynamic situations with a huge amount of urgent decentral changes would overstrain conventional systems [60]. This leads to the following conclusion: theoretically spoken, the required capabilities (e.g., following the idea of looking, listening, learning, connecting, predicting and correcting, but also in the course of other capability taxonomies) can be designed independently from the underlying modelling principle—be it hierarchic or heterarchic, or be it a combination (a ‘hybrid’) of both. However, trying to take advantage from advanced technological possibilities, requires the in each use case applicable modelling approach.
Within digital manufacturing, respective considerations typically rest on the conventional automation layer concept [53]. At the device level, more and more perception points occur, due to the increasing types and amounts of sensors. Additionally, the growing extent of interconnectedness allows for more extensive access not even on machine level, but as well on the level of single components or tools, that are only a part of an industrial machine or a robot. Similar to the people layer, not only humans start to look for pieces of information in a hyper-link structure; the same applies to people and also to software systems that are looking for information about the actual condition inside a factory, or even inside a machine (technical or workload condition). At the same time the advancing development of standards (e.g., OPC/UA) allows for cross-level access, and hence more sophisticated logical interferences.
Shroff [70] mentions connectivity—as he calls it ‘the ability to connect the dots’—as a next maturity step. In the context of human cognition this refers to logical reasoning; in the context of industrial manufacturing machinery, this relates to automated logic reasoning mechanisms. Again, such algorithms could be used on all aforementioned levels—be it the ‘smart’ combination of sensor data that has been retrieved from several sensor devices (i.e. intelligent sensor fusion within cyber-physical manufacturing, e.g. [72]) or be it the application of automated reasoning software applications for the purpose of appraising supplier performance [73]. Not only humans, but also connected devices on all levels—as well a singular sensor, a machine or a machine component, as a PLC-device or a MES-system (and so forth) are enabled to access an immense variety and volume of data via internet access. In a manufacturing context this can be limited to factory-internal devices only, but can also be extended to external internet resources.
As searching for relevant and meaningful data is a major challenge [70], digital manufacturing concepts like Industry 4.0 and similar seek to exploit the evolving technological options bet possibly. Accordingly, hierarchic architecture models come under criticism in some ways, despite the obvious advantage of unique control and accountability due to a single point of power and responsibility. On company- or plant-level, the alternative to hierarchic co-ordination has been discussed since long in the context of make-or-buy decisions, and so-called hybrid co-ordination forms, e.g., a joint venture or a franchising system [14, 21, 22, 25].
Nowadays, powerful IT-infrastructure and data processing algorithms have stimulated the same discussion analogously also at lower technology levels, down to the device-layer: comparable to human agents that execute managerial tasks on behalves of a superordinate principal [24], software agents could make use of the advanced data access capabilities and take decentral decision ‘on behalves of’ a smart production system. Comparably to a supply chain or enterprise scenario, where more than one participant owns a certain extent of influence which allows for at least some autonomous decisions based on respective organizational roles or contracts, also technical devices are enabled to de-centrally influence the manufacturing system to the favour of better decision quality, faster decision taking, higher flexibility and reduced transaction cost. De-central decision taking requires knowledge and the ability to judge. As soon, as technical systems are able to judge, the same structural principles can be used, as on the organisational layer. One doesn’t have to wait for the IoT or the internet of services, in order to explore this assumption: as well a simple control loop (e.g., for temperature control) without any learning abilities, and with few sensors (often only one temperature sensing device) is a first example for de-central power that alleviates superordinate layers—here humans that do not need to control the thermal conditions of a e.g., a building manually.
A further factor of immense importance for the discussed progression is the availability and processability of huge amounts of SCM- and manufacturing-related data. This includes as well company-internal from all automation pyramid layers, as company-external data from customers, suppliers or public sources. Whereas well-structured data can be processed comparably easily with conventional algorithms (given that a sufficient processing capacity is available), huge challenges have to be seen in the processing of ill-structured data from heterogeneous sources, which is assumed to be the disproportionally higher share of relevant data in a SCM context [49]. Big data processing and data-driven strategies are a major success factor in digital SCM and manufacturing. According to the mentioned six elements of looking, listening, learning, connecting, predicting and correcting, a significant enabler is the achievement of richer semantics out of collected data sets. Again, hierarchic layer models play a major role in the course of systemizing this advanced data analysis and processing approaches.
Systematizing manufacturing data, cp. [74]
- Proposition 1
First, decisions taking should be assigned to the position where they are resolved with the best available quality. The more ‘intelligent (i.e., capable of decision taking) a system on a lower level of the automation pyramid becomes, the less important the strictly hierarchical line of commands, as the subordinated devices can decide autonomously instead of ‘having to ask’ superordinate layers for appraisal and permission. Does that at the same time mean, that hierarchy will get obsolete with increasingly smart devices? We assume the opposite, as this first proposition also suggests that simple problems would not unavoidably require this high amount of de-central intelligence and autonomy. The assumption rather is, that an adequate segmentation between centralized and de-central allocation of responsibility has to be achieved. Ideally, this can be modelled by means of a hybrid approach that combines conventional system elements with autonomous (software-)agents within a heterogeneous network.
- Proposition 2
Second, it is advantageous to continuously advance the ability to decide autonomously also on lower levels. In the fields of SCM, enterprise organization and business process optimization, this is a well-known fact since long. On technical layers, the future development has only begun with current IT-processing abilities. Still, we assume, that the development will proceed in a similar way also on the technical layers. However, it is important to consider the economic benefits of building an increased amount of intelligence into a manufacturing system or factory: comparable to the necessity of consequently focusing human and organisational company-resources upon the core competencies of a company, also ‘technical factory intelligence’ has to be focused on the gaining of competitive advantages. This generates a huge challenge for current smart factory and digital manufacturing concepts, as the current experience regarding the linkage of technical ‘low-level intelligence’ to the comprehensive business strategy layer is limited to currently available use cases and business scenarios.
- Proposition 3
Third, in an organizational context de-central intelligence seems to be applicable in particular in the case of complex, scattered and quickly changing environments [75]. Taking into account the aforementioned structural similarities and analogies between the different layers and the phenomena that have been discussed in the previous sections of this chapter, this is likely to be applicable also on technical levels
Altogether, modern IT can be expected to be a strong enabler that allows to repeat previous developments from linear and purely hierarchical control structures and organisational principles towards more heterarchic characteristics within system layers that previously didn’t allow for a distinct degree of de-central autonomy. With regard to question (2), hierarchical layer models are not becoming obsolete in the course of this development. Quite the contrary, the explicit linkage of de-central manufacturing intelligence to a well-aligned and comprehensive business focus seems to be a critical success factor for the purpose of achieving a holistic model of heterarchic systems in the context of digital manufacturing.
5 Managerial Implications
In view of the aforementioned theory-based considerations regarding hierarchic and heterarchic organisation structures, SCM has to face a ‘systemic paradigm’ shift. Current geographically distributed, complex, dynamic and unforeseeable business conditions have generated the urgent need for managerial means that are capable of handling these intricate and fragmented conditions. At the same time, it is to an increasing degree possible to augment also low (i.e., technical) layers with ‘automated intelligence’ be it the generic ability to execute software procedures, the elementary capability to store data (e.g., regarding an object’s own characteristics and actual condition) or the advanced ability for reasoning, (machine) learning or agent-based and negotiation-driven automated task execution.
For the field of SCM this leads to the conclusion, that from a current point of view heterarchic structures offer favourable improvements potentials (question (1)). Despite the advances to be expected from heterarchic structures, most likely a ‘hierarchic backbone’ will remain necessary or at least desirable in terms of efficiency in order to achieve a holistic model of heterarchic systems in the context of digital manufacturing (question (2)). According to the amount and variety of existing research from multiple disciplines, this chapter doesn’t claim to present an exhaustive review or survey. Still, we have concretized several essential aspects regarding the more extensive application of heterarchic principles in order to advance SCM practices. Henceforth, we assume the necessity of future research with a respective focus.