Improvement of management practices in high value added manufacturing networks using a complex adaptive system perspective
The traditional theory of production in operations management has lacked attention to the high-value-added manufacturing industries major segment. This industrial segment face an extreme competitive context where products are large, heavily engineered, highly customized and very expensive, their demand is very difficult to predict, the production lead time is much greater than the delivery lead time than customers expect. Also, holding some critical components stock inventories, WIP inventories, or finished goods inventory is many times impossible both financially and even physically. Thus, traditional order fulfilment strategies (i.e. ETO, MTO, ATO, MTS) are sometimes not a viable possibility. Also, as responsiveness and agility have become important competitive attributes in addition to quality, variety and price, this has lead to many companies to attempt to compete simultaneously in the three domains of product, process, and supply networks. In this sense, the literature states that, novel investigations that profound in the understanding of the interdependencies among the product family design (PF), process design and order fulfilment strategies (P&O), and supply network design (SN), in the high-value-added manufacturing extreme customization environment, are highly desirable both from an academic as well as from a practitioner point of view.
This thesis addresses this challenge, first of all, through the development of a systemic qualitative conceptual model, rooted in a literature research and an empirical multi-case study research performed, that allows the characterization of nine key strategic and tactical problems-issues of high-value-added manufacturing networks. These problems-issues are related to designing geographical configurations of supply networks for customized product orders, setting strategies to deal with different customer behaviours resulting from market changes, establishing inventories’ position and replenishment policies in the supply network, defining the planning period and number of planning points of the supply network, and setting order fulfilment strategies for different product families. Secondly, this thesis provides an empirically grounded, and empirically validated, agent-based decision support simulator (DSS) that quantitative implements the qualitative model in order to dynamically analyse how the various PF-P&O-SN and environment constructs interact. Therefore, how the decisions made within each entity impact over other elements in terms of local and global supply network performance.
Beyond these two contributions, the virtual learning-by-doing nature of the strategic and tactical dynamic DSS has been used for designing and analyzing high-value-added manufacturing network what-if scenarios, in order to extent current managerial practices. These experiments build on under-attended or directly unattended management issues in the literature, related to the nine key problems-issues that drive the conceptual model. Thus, alternative product family structures (i.e. integrated product structures vs. modular product structures), processes & operations strategies (i.e. MTO, STO and ATO order fulfilment strategies, alternative reconfiguration and substitution criteria, and network capacity levels) and market scenarios (i.e. different demand volumes, customer lead times and product variety levels) have been tested and a set of statistically significant generic management guidelines are provided. These guidelines propose novel high-value-added manufacturing managerial heuristics to better deal with customized market scenarios that have proven, at least virtually, to improve key performance indicators of this industrial segment, i.e. supply networks delivery time, product-machine costs, percentage of rejected orders and number of orphan units. Finally, by using complex systems modelling methods, such as the multi-agent model described in the thesis, outcomes of the system under a significant range of possible agent behavioural rules and environmental events can be further explored, and thus, improved levels of functioning can be found.