Supply chain design during product development: a systematic literature review

Abstract This paper explores how supply chains can be designed during product development (PD) through systematically reviewing 143 peer-reviewed journal articles written in English. The findings indicate that practitioners can perform 14 supply chain design (SCD) activities during PD. These activities are grouped into levels and areas, and a model is developed that interrelates SCD with nine product characteristics. Therewith, scholars are given a deep insight into the literature on SCD during PD. The paper also provides a framework for developing company- and project-specific checklists that can be used for multiple purposes, including planning, performing, or evaluating SCD activities before, during, or after PD projects. The implementation of the framework, according to the characteristics of a manufacturer and its PD projects, constitutes an opportunity for the creation of resilient supply chains.


Introduction
During the last few decades, manufacturers' supply chains experienced major structural changes due to globalisation, outsourcing, modularisation, and an explosion in product variety (Christopher, Lowson, and Peck 2004;Doran et al. 2020;Wynstra, Weggeman, and Van Weele 2003). The situation has in many cases evolved from sourcing only a few, simple, and separate components from local suppliers, to sourcing large modules/subsystems in many variants from globally dispersed suppliers (Doran et al. 2007;Gunasekaran, Howard, and Squire 2007). However, globalisation and outsourcing introduced the risk of delivery delays and becoming dependent on remote suppliers (Harrison and Van Hoek 2014). Modularisation creates further risk since the transfer of modules/subsystems to suppliers adds complexity to a manufacturer's operations and reduces visibility (Doran et al. 2020). For example, suppliers need to comply with flexibility requirements but must avoid causing supply chain disruptions (Doran et al. 2020). Thus, manufacturers have become increasingly dependent on suppliers and, at the same time, more sensitive to disruptions (Parker, Zsidisin, and Ragatz 2008).
COVID-19 has revealed the sensitivity and vulnerability of global supply chains (Handfield, Graham, and Burns 2020). Many manufacturers are witnessing major disruptions to both upstream and downstream flows in their supply chains, which can all be traced back to the known supply, demand, and control risks (van Hoek 2020). To deal with the economic, social, and human effects of COVID-19, supply chains should be less cost-driven and influenced more by factors such as sustainability, low emissions, and better risk-recovery strategies (Handfield, Graham, and Burns 2020). Thus, there is a call for designing more resilient supply chains (Handfield, Graham, and Burns 2020;Linton and Vakil 2020;van Hoek 2020). This requires the constant evaluation of whether material moves efficiently and effectively through supply chains, compressing the time inventory remains in supply chains, and, when needed, changing the path of materials flows (Handfield, Graham, and Burns 2020). This approach to supply chain design (SCD) is especially important during product development (PD) (Fine 1998;Hilletofth, Reitsma, and Eriksson 2018;Melnyk, Narasimhan, and DeCampos 2014) since an SCD strategy (e.g. physically efficient or market-responsive) should match product characteristics (e.g. functional or innovative) (Fisher 1997).
There are various real-life examples of 'good' SCD during PD. For instance, van Hoek and Chapman (2006) highlight that Microsoft intentionally creates a sense of product scarcity when launching a new Xbox by limiting product availability in the market. Thus, by involving supply chain functions in PD, it becomes possible to plan for a consumer rush for a product, while minimising inventories and express shipments. Furthermore, based on experiences at a gas turbine manufacturer, B€ ackstrand, Johansson, and Ohlson (2014) suggest that supply chain functions can contribute to PD by ensuring that potential product designs result in the point of differentiation being positioned after the customer order decoupling point. This avoids increasing the level of tied-up capital due to needing to either manufacture different product variants based on forecasts or keeping more inventory on hand.
Despite these and many other examples, there is a lack of a coherent overview of the actual SCD 'activities' that can be performed during PD and their interrelations with product characteristics. Closing this knowledge gap should help practitioners to design resilient supply chains while providing scholars with deeper insight into the literature on SCD during PD. Therefore, this paper systematically reviews the literature in order to explore how supply chains can be designed during PD.
The remainder of the paper is organised as follows. First, the methodology of the systematic literature review is discussed in terms of search strategy and procedural steps. Second, the research context is described, and the findings are reported. Third and last, the research is discussed and concluded.

Methodology
This paper examines the literature on SCD during PD by systematically reviewing published peer-reviewed journal articles written in English. Systematic literature reviews (SLRs) are important for building an understanding of theoretical concepts and terminology and identifying areas in which further research would be valuable (Rowley and Slack 2004). To overcome the substantial challenges associated with knowledge development within the supply chain management (SCM) discipline due to differing theoretical views that influence how findings concerning certain phenomena are interpreted, this paper follows Durach, Kembro, and Wieland (2017) six-step SLR methodology: 1. Formulate a research question. 2. Determine inclusion/exclusion criteria. 3. Retrieve a baseline sample of potentially relevant articles. 4. Apply the inclusion/exclusion criteria from step two. 5. Synthesise the articles. 6. Report the results of the SLR.
To make the process of this SLR as replicable as possible, the six steps are presented in detail in Table 1. Table 1. Completion of the six-step SLR methodology.
Step 1: Formulate a research question Since this paper aims at exploring the SCD activities that can be performed during PD, the following research question is formulated: 'How can supply chains be designed during PD?' Step 2: Determine inclusion and exclusion criteria To obtain potentially relevant and important articles that can be subjected to review, Newbert's (2007) five inclusion/exclusion criteria were considered: Search for articles in peer-reviewed scientific journals in English. Ensure substantive relevance by requiring that selected articles contain at least one keyword in their title or abstract. Eliminate substantively irrelevant articles by excluding articles related to very narrow aspects or contexts. Ensure substantive and empirical relevance by reading all remaining abstracts. Further ensure substantive and empirical relevance by reading all remaining articles in their entirety.
Step 3: Retrieve a baseline sample of articles Within the databases ISI Web of Science, Scopus, and EBSCOhost, the search was limited to peer-reviewed journal articles and two groups of keyword combinations were used: (1) (2) ('design for' or 'concurrent engineering' or 'simultaneous engineering') and ('supply chain' or 'logistics' or 'transportation' or 'packaging' or 'purchasing' or 'procurement' or 'storage' or 'distribution' or 'materials handling' or 'commonality' or 'localization' or 'delayed product differentiation') ('product development') and ('coordination' or 'alignment' or 'integration') and ('supply chain' or 'logistics' or 'transportation' or 'packaging' or 'purchasing' or 'procurement' or 'storage' or 'distribution' or 'materials handling' or 'commonality' or 'localization' or 'delayed product differentiation') The first group of keywords was intended to find articles that use concepts such as 'design for x' or 'concurrent engineering', and the second group to gather articles using different terminology. The search was conducted with no constraint placed on journals, disciplines, or date of publication, and the two groups of keywords were applied at the title, abstract, and keyword level. This search resulted in 3409 potentially relevant articles (1106 in ISI Web of Science, 1429 in Scopus, 874 in EBSCOhost). After sorting out duplicates (799) and removing remaining non-journal articles (44), 2566 unique articles were identified (the search was conducted in February 2018) and database alerts were set to make sure that new potentially relevant articles are subjected to review as well.
Step 4: Apply the inclusion/exclusion criteria from step two The inclusion/exclusion criteria ( Step 2) were applied to reduce the baseline sample to a subset of relevant studies. To overcome challenges related to inclusion-criteria bias and selector bias, two scholars individually applied the criteria when reviewing the articles. Cohen's kappa coefficient (j) was used to measure selector bias by rating the level of agreement between the scholars throughout the screening process of the baseline sample. When Cohen's j indicated that the two scholars were in 'almost perfect agreement' (Landis and Koch 1977), the articles in which disagreement occurred were double-checked through a repeated joint evaluation of the respective articles and, if needed, with a third scholar present. The screening process resulted in an initial sample of 109 relevant articles. During the writing process of this paper, the set database alerts added 16 articles to the initial sample.
The reference lists of all articles were checked to identify additional relevant articles. This snowball sampling method added another 18 articles to the sample. Each of the articles that were added to the initial sample of 109 articles were also individually screened by two scholars and, if needed, evaluated with a third scholar present. This resulted in a synthesis sample consisting of 143 articles.
Step 5: Synthesise the articles An aggregative synthesis approach was applied that incorporates both quantitative and qualitative elements (Denyer and Tranfield 2009). The quantitative synthesis focussed on extracting data from articles following an a priori defined coding structure: Publication date. Type of industry. The qualitative synthesis focussed on extracting data from articles following an emergent thematic coding structure. Emergent themes were revealed by three scholars identifying, categorising, and summarising the main contributions of the articles.
Step 6: Report the results Following the advice of Tranfield, Denyer, and Smart (2003), a descriptive analysis of the synthesis sample (n ¼ 143) is first presented and afterwards, a thematic analysis. The descriptive analysis follows the a priori defined coding structure listed in Step 5, while the thematic analysis answers the research questions formulated in Step 1, based on themes that emerged during the qualitative synthesis part of Step 5.

Research context
This part of the SLR aims at showing the context of the literature on SCD during PD. To achieve this, the synthesis sample of 143 articles is dissected by following the a priori defined coding structure listed in Step 5 of Table 1.

Publication date
The first articles that focussed on SCD during PD were published in 1992. Since the appearance of this theme in the academic literature, only one year passed by without a new article being published (1998) (Figure 1). On average, five articles are published on a yearly basis and more than 49% of the articles were published in the last decade. The number of newly published articles peaked in 2017 (10 new articles). On average, seven articles were published on a yearly basis during the last 10 years, indicating that the theme is still actively discussed by scholars.

Type of industry
Regarding the characteristics of the samples used by scholars in the synthesis sample, articles mostly focus on the automotive (24 articles) and computer industries (13 articles). These two industries are probably mostly represented, due to being characterised by a high level of vertical integration, outsourcing, and relatively short product life cycles. Scholars in the sample rarely focus on manufacturers of 'complex products and systems' (Davies and Hobday 2005). For example, the aerospace industry is represented in only two articles. It is also worth noting that scholars often do not disclose the industry under investigation (27 articles) or investigate multiple industries within their article (39 articles).

Findings
This section of the SLR focuses on answering the research question formulated in Table 1 (Step 1). To do so, the SCD activities extracted from the literature are discussed on the basis of key dimensions and themes. This resulted in grouping the SCD activities that can be performed by practitioners during PD into levels and areas ( Figure 2). The SCD activities are listed in Tables 2-5, and it is also discussed how they interrelate with product characteristics.

Strategy
Strategy-level SCD activities span the areas of sourcing, collaboration, and postponement ( Figure 2). First, sourcing refers to finding a balance between internal and external sourcing, a geographical sourcing location, and a balance between single and multiple sourcing. Second, collaboration concerns determining the extent of collaboration with supply chain actors responsible for certain product items (e.g. raw materials, components, modules/subsystems) or workloads. Third and last, postponement refers to finding a balance between forecast and demand-driven activities in the supply chain by positioning the customer order decoupling point.

Sourcing
During PD, practitioners can determine the approach to sourcing products and/or workloads in the supply chain (Table 2). When there are no existing relationships with suppliers, the balance between internal and external sourcing can be determined by prioritising the areas to which internal resources will be devoted. First, this activity answers the 'make or buy' question for each item within a product design. Two classes of items can be roughly distinguished for each item: an in-house made item (internal sourcing) and an outsourced item (external sourcing) (Zolghadri, Baron, and Girard 2008). Second, this activity entails determining whether certain workloads will be (partially) outsourced to suppliers. Examples of these workloads include research and (technology) development (R&D), production, marketing, packaging design, materials handling, and transportation (Johansson 2007;Johansson, Bellgran, and Johansson 2006;Wynstra, Van Weele, and Axelsson 1999;Wynstra, Weggeman, and Van Weele 2003). The balance between internal and external sourcing depends on product characteristics (Fine 2009; Fine, Golany, and Naseraldin 2005; Noori and Georgescu 2008; Primus 2017). For example, Primus (2017) proposes that the outsourcing of items can be an option when developing a short-life-cycle product if its design is highly modular. Additionally, if the manufacturer has the option of outsourcing workloads such as production to low-cost suppliers, then a highly modular design facilitates the outsourcing of a larger proportion of production at low cost (Nepal, Monplaisir, and Famuyiwa 2011). Conversely, a short-life-cycle product can be integral when the manufacturer is primarily responsible for realising the product. However, Ulku and Schmidt (2011) argue that outsourcing can be associated with an integral product design if the technical collaboration penalty is not excessive and suppliers have superior PD capabilities (i.e. capacity and skills). The manufacturer's competitiveness should be considered when determining the balance between internal and external sourcing (Noori and Georgescu 2008). When an item is critical for achieving a competitive advantage, in-house production becomes favourable (Noori and Georgescu 2008). However, when this item also has a short life cycle and there are no issues related to
[ 36,52,66,67,75,76] intellectual property ownership, a modular design can create conditions for outsourcing its production to strategic suppliers (Noori and Georgescu 2008). When an item is not critical for achieving a competitive advantage and has a short life cycle, the balance between internal and external sourcing can be based on the manufacturer's capacity availability and on the suppliers' responsiveness, cost, capacity flexibility, innovativeness, and design know-how (Noori and Georgescu 2008;Wang and Shu 2007). For example, outsourcing is favourable when a supplier has invested in process improvement efforts that decrease the supply lead time (Nepal, Monplaisir, and Famuyiwa 2012). Practitioners can also determine the geographical sourcing location based on product characteristics. Krikke, Bloemhof-Ruwaard, and Van Wassenhove (2003) compare the performance of geographically centralised and decentralised supply chains while considering multiple product designs. The comparison concludes that it is more cost efficient if the manufacturer and other supply chain actors are located in close proximity to each other (Krikke, Bloemhof-Ruwaard, and Van Wassenhove 2003). These actors are ideally also concentrated in one city or geographic region when a product design is integral (Fine, Golany, and Naseraldin 2005). Hong et al. (2018) add that, due to transportation being one of the main contributors to carbon emission, geographically centralised supply chains can significantly reduce a new product's environmental impacts. A geographically centralised supply chain also reduces the Table 4. Node-level SCD activities.

SCD activity (determine) Description
Reference (see supplemental appendix for bibliographic details) The flow layout The manufacturer can determine the flow layout within its node(s) (i.e. manufacturing node or storage node). The flow layout is dependent on manufacturing processes, storage areas, and materials handling operations.
[ 3,23,24,26,34,58,60,73,74] The capability of storage areas The manufacturer can determine the areas within a node (i.e. manufacturing node or storage node) where product items provided by suppliers can be stored before being fed to workstations, where point-of-use storage and WIP is held, and where products are transported to customers.
[23, 58, 60] The capability of materials handling equipment The manufacturer can determine the equipment (e.g. stations and forklifts) needed for materials handling (i.e. moving, storing, packing) within a node (i.e. manufacturing node or storage node).
[ 6,31,36,37,41,44,45,58,88,99,129,134,135,136] The capability of storage nodes The manufacturer can determine the storage network when utilising storage nodes before receiving product items from suppliers and/or before distributing products to customers. [6,15,31,37,55,62,126] The capability of the transportation network The manufacturer can determine the transportation network that support the materials flow in the supply chain. This includes determining the capability of transportation modes and supply and distribution channels.

SCD activity (determine) Description
Reference (see supplemental appendix for bibliographic details) The approach to demand forecasting The manufacturer can forecast known and future demand so that products can be reliably delivered, and customers are (sufficiently) satisfied. [24,55,58,59,60,62,99,120,125,126] The approach to inventory management The manufacturer can determine the levels and locations of stocks throughout the supply chain to maintain an appropriate trade-off between customer service and costs. [19,22,24,26,31,34,38,40,44,45,48,49,52,53,54,58,59,60,67,75,76,88,91,111,125,127,129,136,141,143] The approach to capacity planning The manufacturer can determine the production (and transportation) capacity needed in the supply chain to meet changing demands for new and old products. 3,6,15,24,26,36,40,44,48,49,58,62,91,126,134,135,136,143] negative implications of product variety on operational performance when dealing with a modular product family (Salvador, Forza, and Rungtusanatham 2002). This means that variety in a product family is ideally reduced when a supply chain is geographically decentralised (Nielsen and Holmstr€ om 1995). Pero et al. (2010) explain that global sourcing is typically required when an item is innovative, whereas local sourcing is typically associated with less innovative items. Alternatively, van Hoek and Chapman (2007) argue that global sourcing might only work for basic and established items and that local sourcing can make testing and initial production easier and faster for the manufacturer. During PD, practitioners can also balance single and multiple sourcing. This can be done by comparing the costs and benefits obtained from the use of economies of scale associated with single sourcing (e.g. single supplier, manufacturing plant, warehouse, distribution centre), with the costs and benefits of multiple sourcing (Dowlatshahi 1996). This comparison can be based on aspects such as transportation costs, availability of transportation modes, the manufacturer's ability to provide speedy and reliable deliveries and services, and the ability of supply chain actors to supply materials on time (Dowlatshahi 1996(Dowlatshahi , 1999Pham and Yenradee 2017). Krikke, Bloemhof-Ruwaard, and Van Wassenhove (2003) explain that single sourcing is more cost efficient than multiple sourcing, due to economies of scale lowering the fixed costs per unit. Furthermore, Claypool, Norman, and Needy (2014) propose that single sourcing can be applied when the manufacturer procures items that are inexpensive to buy and store. Additionally, Di Benedetto et al. (2003) highlight that single or limited sourcing is typically associated with items that have a relatively high level of quality and innovativeness. A main driver for this specific sourcing mode is optimising the strategic size of the supply base (Di Benedetto et al. 2003). For example, reducing the number of suppliers facilitates better rates, services, and relationships with the remaining suppliers (Di Benedetto et al. 2003;Dowlatshahi 1996;Guy and Dale 1993;Wynstra, Weggeman, and Van Weele 2003).

Collaboration
Practitioners can determine how to collaborate with other supply chain actors during PD (Table 2). First, the importance of collaborating with actors within the manufacturer's own network (e.g. 'sister' manufacturing plants) should be recognised. This can be done by promoting a common 'language' about strategic network goals, in order to support collaboration in such a way that a balance between control and freedom is reached (Arellano, Rebolledo, and Tao 2019). Second, the approach to collaboration with suppliers outside the manufacturer's own network should be determined. Supplier collaboration (i.e. supplier development as well as involvement in the manufacturer's PD process) can be critical to the latter's growth and very existence (Song and Di Benedetto 2008). Therefore, practitioners should determine with which supplier tiers to collaborate. It may be desirable to interact directly with lower-tier suppliers when they are responsible for strategically critical items, or when tier-one suppliers do not have the skills to undertake the required coordination (Wynstra, Van Weele, and Axelsson 1999). Furthermore, when PD projects are associated with technological uncertainty, it is beneficial to explore potential sources of complementary resources and capabilities embedded further up the manufacturer's supply chains (Xiao et al. 2019).
Collaborations can differ between PD projects. When there is a high level of conflicts of interest related to a fear of information leakage during a PD project, collaboration with supply chain actors can be limited to avoid knowledge spill-over (Melander and Tell 2019). Limited collaboration is also advised when a PD project is associated with technological uncertainty (Mikkelsen and Johnsen 2019). This makes it possible to break ties with suppliers after a PD project and to limit the sharing of sensitive information. When there is a high level of uncertainty, the manufacturer should also be protected from opportunistic behaviour from suppliers, and at the same time, create opportunities for dynamic changes that are necessary due to uncertainty (Melander and Lakemond 2015). Furthermore, when suppliers work on products associated with a low level of innovativeness, supplier relationships can be managed according to R&D technical requirements and supplier cost propositions (Picaud-Bello et al. 2019). When little collaboration is needed, a manufacturer and its suppliers may be culturally different, have few close organisational ties, and modest electronic connectivity (Fine, Golany, and Naseraldin 2005;Ulku and Schmidt 2011). Limited collaboration may also be appropriate when outsourcing workloads associated with highly modular products (Fine, Golany, and Naseraldin 2005). In this scenario, the need for vertical integration is reduced due to modularisation facilitating the relocation of value-adding activities in the supply chain (Pero et al. 2010). This makes it possible for the manufacturer and suppliers to provide feedback on a specific aspect of the product or process when discussing improvements (Droge, Vickery, and Jacobs 2012). However, product modularity can make the manufacturer more dependent on suppliers (Nepal, Monplaisir, and Famuyiwa 2012), as it may need the help of knowledgeable suppliers to appropriately modularise the allocated design (Salvador, Forza, and Rungtusanatham 2002;Ye et al. 2018).
Closer collaboration is typically more appropriate when suppliers are responsible for workloads associated with products that have a high level of integrality, quality, variety, or innovativeness (Di Benedetto et al. 2003;Fine, Golany, and Naseraldin 2005;Kou, Chiang, and Chiang 2018;Lau and Yam 2005;Luzzini et al. 2015;Nielsen and Holmstr€ om 1995;Pero et al. 2010;Ulku and Schmidt 2011;Wasti and Liker 1997;Zolghadri, Baron, and Girard 2008). For example, when products are highly innovative, a manufacturer may need to rely heavily on suppliers to develop a clear understanding of certain product technologies (Picaud-Bello et al. 2019;Ragatz, Handfield, and Petersen 2002;Song and Thieme 2009). A highly innovative product may also require suppliers to tailor their production system for the development of the product (Song and Di Benedetto 2008). It takes time and effort to achieve close collaboration with suppliers (Simms and Trott 2014). For example, it may be necessary to establish cross-functional teams, be linked electronically through integrated IT systems, have a common business and social culture, and a common or interlocking ownership (Fine, Golany, and Naseraldin 2005; Jimenez-Jimenez, Mart ınez-Costa, and Rodriguez 2019; Noori and Georgescu 2008;Pero et al. 2010;Petersen, Handfield, and Ragatz 2005;Ulku and Schmidt 2011).
Suppliers can be integrated at varying stages of the PD process (Petersen, Handfield, and Ragatz 2003;Wynstra, Van Weele, and Axelsson 1999;Wynstra, Weggeman, and Van Weele 2003). During less predictable PD projects, early involvement may be difficult to achieve, because there is less certainty about which suppliers to use (Eisenhardt and Tabrizi 1995). By contrast, during more predictable projects, it is possible and even preferable to involve them early (Eisenhardt and Tabrizi 1995). For example, when a supplier possesses high levels of expertise in a technology that needs to be adapted to a particular PD initiative, it can be involved very early in the PD process, so as to benefit from its expertise (Ragatz, Handfield, and Petersen 2002). When suppliers are invited to contribute to the design stages of the PD process, the manufacturer-supplier alliance should prioritise addressing those items that have a short life cycle and are critical for achieving a competitive advantage (Noori and Georgescu 2008). Sharing design work with suppliers is especially relevant when a manufacturer wants to shorten its development lead time (Wasti and Liker 1997). Even so, suppliers can slow down PD projects (especially critical suppliers) due to uncooperative attitudes or not prioritising the manufacturer (Primo and Amundson 2002).

Postponement
During PD, practitioners can also find a balance between forecast and demand-driven activities in the supply chain by positioning the customer order decoupling point (CODP) ( Table 2). The location where customisation is performed in the supply chain affects how the manufacturer balances the levels of inventory and service (Lee, Billington, and Carter 1993). Therefore, CODP placement should not harm the intended product quality and should lead to appropriate lead times, and optimally allocate inventory, so as to minimise production cost (Kristianto et al. 2012). The position of the CODP is affected by the manufacturing strategy (e.g. make-to-stock, make-to-order, configure-to-order) (Kristianto and Helo 2010). For example, when operating in a configureto-order environment, the manufacturer can delay the configuration of final products until orders arrive and customer requirements become known (Kristianto and Helo 2010). Consequently, the CODP can be positioned by placing inventory at the module level at various places in the supply chain (Lee and Sasser 1995). This action requires an appropriate degree of product modularisation. Hsu and Wang (2004) explain that, even though this approach to (re)designing the products or processes may require extra investments in order to arrange the common items for use in some of these products, there is a great advantage to risk-pooling effects. For example, delaying customisation until customer orders arrive may increase the manufacturer's flexibility to respond to changes in the mix of demands from different market segments (Hsu and Wang 2004). In addition to being able to manage product variety and improve its responsiveness to orders, the manufacturer can reduce its inventory investments (Lee, Billington, and Carter 1993).

Network
Network-level SCD activities model the manufacturer's supply chain as a network in which nodes represent the supply chain actors (Figure 2). Generally speaking, there are three types of nodes that can be managed by the manufacturer: supplier nodes (where items or workloads are controlled by suppliers), manufacturing nodes (where the manufacturer creates sub-assemblies and/or products) and storage nodes (where the manufacturer or a supplier stores inbound items or outbound products). The nodes are interlinked, which refers to the connections that the manufacturer creates with other actors to support the materials flow (i.e. inbound flow of items, flow of WIP, outbound flow of products to customer nodes, return flows) in the supply chain.

Supplier nodes
When outsourcing items or workloads during PD, it is necessary to carefully analyse, assess, and nominate suppliers based on their capabilities (e.g. Chiu and Lin 2016;Fixson 2005;Homfeldt et al. 2017;Kao 2006;Simms and Trott 2014;Wang, Huang, and Dismukes 2004) (Table 3). Since new suppliers typically need to pass a sequence of quality assessments and certifications before being involved in PD, manufacturers tend to collaborate with existing suppliers when the duration of PD is expected to be short (Lau et al. 2018;Simms and Trott 2014). However, when deciding to involve new suppliers, principal and lower-tier suppliers can be selected from a pool of potential options based on a wide range of criteria (Wang and Shu 2007;Wynstra, Van Weele, and Axelsson 1999). The supplier selection criteria depend on product characteristics. For example, when an outsourced item is critical for achieving a competitive advantage and its life cycle short, suppliers should be able to innovate and be technologically capable (Brewer and Arnette 2017;Di Benedetto et al. 2003;Noori and Georgescu 2008). Alternatively, when short-life-cycle items have a low impact on competitive advantage, supplier selection is a trade-off between production costs and market mediation costs (Noori and Georgescu 2008).
When a manufacturer wants to collaborate with suppliers on a product associated with a high level of innovativeness, it typically needs to search for suppliers outside its existing supply network (Mikkelsen and Johnsen 2019). In this scenario, the innovative capabilities of suppliers are important, and new suppliers can be allowed to circumvent the standard supplier approval process in order to fast track their involvement (Mikkelsen and Johnsen 2019). The involvement of a new supplier originating from a different industry in PD is also possible, but this requires a leap of faith from both the manufacturer and the supplier (Mikkelsen and Johnsen 2019). In general, when a manufacturer's products are of high quality and highly innovative, suppliers should match these characteristics (Arntzen et al. 1995;Brewer and Arnette 2017;Chiu and Okudan 2014;Di Benedetto et al. 2003;Fine 2009;Nellore and Balachandra 2001;Sharifi, Ismail, and Reid 2006;Wang and Shu 2007). Alternatively, when a manufacturer wants to collaborate with suppliers on a product associated with a low level of innovativeness, it can typically find suppliers within its existing supply network (Mikkelsen and Johnsen 2019). In this scenario, cost, quality, and delivery are critical supplier capabilities, and suppliers should continuously be assessed based on their efficiency and contributions (Mikkelsen and Johnsen 2019).
Within a certain product design, it is advisable to replace long-lead-time items with standard items (Mather 1992;van Hoek and Chapman 2007). Using standard and simple items results in an abundance of potential suppliers (Dowlatshahi 1992;Mather 1992;van Hoek and Chapman 2007). Being able to choose from a large pool of suppliers makes it possible for a manufacturer to reduce lead times and product and inventory costs (Dowlatshahi 1992;Wynstra, Weggeman, and Van Weele 2003). Additionally, Pero et al. (2010) argue that product modularisation makes it possible to reduce the number of suppliers needed. However, when a product is modular, manufacturers may need to add a supplier tier that can take responsibility for module assembly, which increases the risk of longer delivery lead times (Lau and Yam 2005).
Suppliers can be selected during various stages of the PD process. During less predictable PD projects, manufacturers may favour last-minute supplier selection, so as to maintain design flexibility later in the PD process (Eisenhardt and Tabrizi 1995). In contrast, during more predictable PD projects, suppliers are likely to be selected early on (Eisenhardt and Tabrizi 1995). To optimally match the characteristics of suppliers and products designs, the suppliers can be selected once the structure of the product design(s) is listed in a bill of materials (BOM) (e.g. Graves and Willems 2005;Zolghadri, Baron, and Girard 2008). The BOM is typically created during the system-level design and/or detail design stage and can include information on the costs and sourcing constraints of items (Pham and Yenradee 2017). Gokhan, Needy, and Norman (2010) propose that the BOM can be used to decide which design alternative to manufacture for each item and which supplier to select, based on the impact a potential supplier has on lead times and costs related to production, transportation, and the introduction of suppliers (i.e. verification, training, validation costs). Yan, Yu, and Cheng (2003) show that information regarding quantities of items in the BOM can be used to select suppliers based on their capacity limits. Claypool, Norman, and Needy (2014) use a BOM-based model to select an optimal design alternative and supplier for each item. This model considers time-to-market risks, supplier reliability, and strategic exposure risk (i.e. the risk of the manufacturer being too reliant on one or a few suppliers). Feng, Wang, and Wang (2001) show that the BOM can be used to jointly set tolerances and select suppliers for each item. The BOM can also include information regarding the density, cost, carbon footprint, and amount of carbon rebate of items (Hong et al. 2018). This makes it possible to evaluate the environmental impact of a product when selecting certain suppliers (Chu, Su, and Chen 2012;Hong et al. 2018).
The BOM can also be used to determine the level of product modularity while selecting suppliers based, for example, on exchange rates, cost, lead times, and their compatibility with the manufacturer (Elmaraghy and Mahmoudi 2009;Famuyiwa 2011, 2012). Fine, Golany, and Naseraldin (2005) show that the BOM can be used to simultaneously evaluate the optimal level of product modularity and the performance of potential suppliers, based on five conflicting objectives: quality, cost, lead time, partnership, and dependency. Huang, Zhang, and Liang (2005) add that module suppliers can be selected while determining the optimal modular product design, based on the level of commonality among variants in a product platform. Here, module suppliers can be selected according to their level of flexibility (i.e. how many items as part of a variant module they can produce and supply) (Huang, Zhang, and Lo 2007). Similarly, Zhang, Huang, and Rungtusanatham (2008) propose that the BOM can be used to evaluate whether a potential supplier is flexible enough when there is a high level of customisation in platform-based product variants.

Manufacturing nodes
A manufacturer can be part of a larger network of manufacturing plants that are based in different locations and owned by the same company. Therefore, when the manufacturer opts for in-house production, it can produce its new product at multiple (sister) plants within this network. This set of manufacturing plants can offer different levels of service at different costs (Table 3). During PD, practitioners can determine the manufacturing network based on a variety of criteria, including capacity limits, cost, and lead time (Wang and Shu 2007). It is frequently suggested that this network should be established at an early stage of the PD process (e.g. Fixson 2005; Fujita, Amaya, and Akai 2013; Badinelli 1999a, 1999b;Nepal, Monplaisir, and Famuyiwa 2011). Johansson (2007) provides an example of the manufacturer assessing where to locate production during the planning stage. Other scholars specify that this activity can be more properly performed when the product design is listed in a BOM during the system-level design and/or detail design stage. Due to listing quantities and quality levels of items, the BOM can be used to determine the number, location, capacity, and type of manufacturing plants based on their throughput limits (Yadav et al. 2008;Yadav et al. 2011;Yan, Yu, and Cheng 2003). For example, the BOM can be used to calculate how many manufacturing plants an item or product should be manufactured, where they should be located, and what technologies and capacities each should have (Arntzen et al. 1995).
In addition to listing the quantities and performance levels of items, a BOM can include the impacts of items on overall product cost, finished-goods production, overall throughput capacity, and key capital decisions (Pham and Yenradee 2017). This makes it possible to consider the investment required to acquire (e.g. investment of land), build (e.g. installation cost), and maintain (e.g. production cost) infrastructure for any potential manufacturing plant during its life cycle (Pham and Yenradee 2017). Including cost-related information in the BOM also makes it possible to select manufacturing plants based on cost of goods sold and the inventory holding costs for safety stock and pipeline stock (Graves and Willems 2005). Elmaraghy and Mahmoudi (2009) further explain that the BOM can be used to determine the optimal modular product design and manufacturing network while considering global exchange rates and costs related to production, inventory, and transportation.

Storage nodes
Before a manufacturer receives items from suppliers or distributes products to customers, items and products are typically held at storage points (e.g. warehouses or distribution centres) (Arntzen et al. 1995;van Hoek and Chapman 2007). The manufacturer can be in the position of owning or controlling this network of storage points. If so, practitioners can determine the number, location, capacity, and degree of automation of these storage points during PD (Arntzen et al. 1995;Chiu and Lin 2016;Jafarian and Bashiri 2014;Khan, Christopher, and Creazza 2012) (Table 3). For example, a centralised and highly automated distribution centre can reduce product and delivery lead times (Khan, Christopher, and Creazza 2012). This activity also includes deciding which storage point serves which of the manufacturer's plant(s) or customer(s) for each type of order and product. Similar to the selection of suppliers and 'sister' manufacturing plants, the product BOM can be used to determine the network of storage points in the supply chain (e.g. Fujita, Amaya, and Akai 2013; Jafarian and Bashiri 2014). The BOM typically becomes available during the system-level design or detail design stage, and Yan, Yu, and Cheng (2003) show that it can be used to assess the throughput limits and fixed opening and operating cost of storage points. Similarly, Elmaraghy and Mahmoudi (2009) use the BOM to select storage points while considering the impact of global market currency exchange rates and costs related to inventory and transportation.

Nodal links
A transportation network is needed to support the materials flow (i.e. inbound flow of items, flow of WIP, outbound flow of products) in a manufacturer's supply chain. This network connects the manufacturer (i.e. one or multiple storage point(s) and/or manufacturing plant(s)) with other supply chain actors (i.e. suppliers and customers) and may need to be determined during PD (Table 3). For example, practitioners can assess the supply and distribution channels in terms of their length in distance and time (Arntzen et al. 1995;Dowlatshahi 1999;Fine 2009;Fujita, Amaya, and Akai 2013;. Additionally, they can assess potential transportation modes based, for example, on their freight rate, traffic rate, shipment size, exchange rate, wage rate, and legal and ethical requirements (e.g. Arntzen et al. 1995;Dowlatshahi 1996Dowlatshahi , 1999. The appropriateness of a transportation mode depends on product characteristics (Khan, Christopher, and Creazza 2012), including the product's physical properties (e.g. mass, volume, density, substance effects) and quality (e.g. fragility) (Dowlatshahi 1996(Dowlatshahi , 1999. This means that the constraints of transportation modes should be considered when designing a product and that new transportation modes may be needed due to new product characteristics (Dowlatshahi 1996(Dowlatshahi , 1999. The determination of the transportation network can span multiple stages of the PD process. For example, the location to which a product should be delivered can be identified during the planning stage of the PD process (Johansson 2007). The appropriateness of supply and distribution channels depends on the product characteristics such as product price and quantities (Yadav et al. 2008(Yadav et al. , 2011. Therefore, scholars argue that the transportation network can be defined once the product BOM has been created. BOM information such as quantities of items can be used to assess the costs of transporting a product in the supply chain (Pham and Yenradee 2017;Yan, Yu, and Cheng 2003) and to assess the transportation capacity needed in the supply chain (Fujita, Amaya, and Akai 2013). The BOM can also include information regarding the carbon rebate of items, in order to reduce costs by appropriately routeing items and products in the supply chain (Hong et al. 2018). The inclusion of costrelated information in the BOM also makes it possible to select a transportation mode while considering the cost of goods sold and the inventory holding costs for safety stock and cycle stock (Graves and Willems 2005). The BOM can also be used to jointly select the transportation modes and the optimal configuration of platform products, based on the degree of commonality among variants within a product family (Huang, Zhang, and Liang 2005).

Node
Node-level SCD activities span the areas of (materials) flow layout, storage areas, materials handling equipment, and the packaging ( Figure 2). First, the flow layout concerns the design of a node's floor plan. Second, storage areas refer to those places within a node where items and products are stored. Third, materials handling equipment concerns any tool used to aid the moving, storing, and packing of items and products. Fourth and finally, packaging refers to the process of designing, evaluating, and/or producing the packaging for inbound items, WIP, and outbound products.

Flow layout
Practitioners can optimise the flow layout (i.e. flow of items, WIP, and products) within a manufacturer's node(s) (i.e. storage node and/or manufacturing node) during PD (Appelqvist, Lehtonen, and Kokkonen 2004) (Table 4). The flow layout is the design of a node's floor plan that aims at improving efficiency by appropriately arranging the manufacturing processes, storage areas, and materials handling operations (Dowlatshahi 1994;Fine 2009;Lee 1999Lee , 2001. First, within a manufacturing plant, the flow layout depends on the layout of manufacturing processes (e.g. job shop or cellular). The layout of manufacturing processes should be based on product quality (e.g. stability, robustness) (Dowlatshahi 1994;Fine 2009). Separate considerations of the product design and the layout of the manufacturing process may play a role in reducing the materials handling cost (Lee 1999(Lee , 2001. Therefore, Lee (1999) proposes that an item should be designed specifically for materials handling. For example, making minor design modifications makes it possible to select the right operation sequence that reduces the distance over which items have to be transferred (Lee 1999). However, the most desirable way is to simultaneously design items and the layout of manufacturing processes for materials handling (Lee 1999(Lee , 2001. This approach can result in significant cost savings due to a simplified materials flow (Lee 1999(Lee , 2001. Therefore, when setting or changing a product's physical properties (e.g. geometries), the flow layout within a manufacturing plant can be optimised by evaluating different potential manufacturing process layouts (Appelqvist, Lehtonen, and Kokkonen 2004). This evaluation can already take place during the concept-development stage in the PD process (Johansson 2007;. By the same token, products should be designed so they can flow efficiently through the various manufacturing processes (Dowlatshahi 1996(Dowlatshahi , 1999. For example, when developing products and processes, practitioners can assess the number and size of workstations and the space available for the display of items at these workstations (Johansson 2007). Second, the flow layout also depends on how various storage areas within a node (e.g. storage point and/or manufacturing plant) are connected with each other . Third and last, the flow layout depends on the position and layout of materials handling stations (e.g. circular, single-row, multi-row) (Lee 1999(Lee , 2001. Therefore, within a certain station, items should be routed according to how their physical properties, tolerances, and processes impact the precedence constraints of each item (or among items) (Lee 1999(Lee , 2001).

Storage areas
During PD, practitioners can determine the areas within a node where items provided by suppliers can be stored before being fed to workstations, where point-of-use storage and WIP is held, and where products are stored before being transported to customers (Dowlatshahi 1994;Johansson 2007; (Table 4). This can include determining the storage areas' principles (e.g. FIFO/ LIFO, dedicated/not dedicated), layout, and capacity (Johansson 2007;. Several storage areas may be needed for locating point-of-use storage and WIP inventory within a production area (Dowlatshahi 1994;. Outside the production area, additional separate storage areas may be needed to hold WIP and product inventory. Johansson (2007) adds that room should be made within these storage areas for new items and products during the design stage (e.g. system-level design or detail design) in the PD process.

Materials handling equipment
The equipment (e.g. stations, forklifts) for materials handling (i.e. moving, storing, packing) within a node can be selected and/or evaluated by practitioners during PD based, for example, on its degree of automation, cost, and load/unload capacity (Atmani and Dutta 1996;Dowlatshahi 1994;Gupta and Dutta 1994; (Table 4). The choice of materials handling equipment affects the ideal layout of manufacturing processes (Lee 1999(Lee , 2001. For example, a materials handling robot imposes a circular layout, while an automated guided vehicle imposes a linear single-row layout or multi-row process layout (Lee 1999(Lee , 2001. Other requirements, such as work environment regulations should also be considered (Johansson 2007). For example, certain types of equipment may not be allowed in certain areas (e.g. no forklifts in the production area) (Johansson 2007). The appropriateness and amount of handling equipment needed are affected by product characteristics such as quality (e.g. tolerances), quantities, life cycle (e.g. future redesigns), and physical properties (e.g. weight, size) ???? ( Appelqvist, Lehtonen, and Kokkonen 2004;Dowlatshahi 1994Dowlatshahi , 1996Dowlatshahi , 1999Gupta and Dutta 1994;Lee 1999Lee , 2001. Therefore, the handling requirements of new products should be determined (Johansson, Bellgran, and Johansson 2006). For example, Johansson (2007) states that equipment needed to stack items at workstations can be evaluated when developing products and processes. At the same time, the optimal location and requirements of handling stations can be assessed (Johansson 2007;).

Packaging
During PD, practitioners can determine the packaging for inbound items, WIP, and outbound products in terms, for example, of its physical dimensions, strength, shape, ease of handling, materials, transportation efficiency, and cost (Dowlatshahi 1996(Dowlatshahi , 1999van Hoek and Chapman 2007) (Table 4). Simms and Trott (2014) advise manufacturers to use existing packaging formats and technologies where possible. Nonetheless, it may be necessary to assess whether there is a need for new or special packaging. For example, special packaging may be needed when inbound items are sensitive and at risk of damage . This indicates that the appropriateness of a type of packaging depends on the design of a product. Consequently, to optimally benefit from a certain type of packaging, there may be a need to improve product quality (e.g. reduced fragility) by redesigning parts more effectively (Dowlatshahi 1996). Practitioners can consider packaging throughout the PD process (Bramklev 2010). During the planning stage, the product mission statement can be supplemented by a packaging mission statement (Bramklev 2010). When developing product concepts, it is possible to estimate the type of packaging needed and how many products will fit the packaging (Johansson, Bellgran, and Johansson 2006). Packaging is important for the support of the product and can take over certain product functions (e.g. protection) for parts of the product life cycle, especially during the production and distribution of the product (Bramklev 2010). Therefore, during the same stage of the PD process, the function decomposition between product and package can be decided upon (Bramklev 2010). Packaging design or selection can be determined together with product layout during the system-level design stage (Bramklev 2010). Afterwards, packaging materials can be chosen during the detail design stage (Bramklev 2010). When testing and refining a product, the packaging should be tested as well (e.g. shock levels, fragility) (Dowlatshahi 1996(Dowlatshahi , 1999. Johansson (2007) adds that prototype packaging can be tested before testing the serial production packaging. When evaluating initial production outputs during production ramp-up, it may be needed to respond to demands for alterations of both the product design and the packaging design (Bramklev 2010).

Planning
Planning-level SCD activities cover the areas of demand planning, inventory management, and capacity planning in the supply chain ( Figure 2). First, demand planning is the process of forecasting demand (known and future) so that products can be reliably delivered, and customers are (sufficiently) satisfied. Second, inventory management evolves around determining the levels and locations of stocks throughout the supply chain, so as to maintain an appropriate trade-off between customer service and costs. This includes the location and sizing of stocks in the supply chain and the management of cycle stock and spare-parts inventories to support production. Third and last, capacity planning refers to the process of determining the production (and transportation) capacity needed in the supply chain to meet changing demands for its products.

Demand planning
During PD, practitioners can forecast the number of products requested by customers at a particular time interval to ensure that products are delivered reliably and on-time (Pham and Yenradee 2017) (Table 5). To create an accurate forecast during PD, van Hoek andChapman (2006, 2007) argue that it is necessary to flag forecasting differences between various functions. Furthermore, customer demand information needs to be continuously updated (e.g. in an ERP system) and translated into the overall capacity plan (Johansson, Bellgran, and Johansson 2006;. The forecasting of known and future demand can take many shapes and sizes during PD. Johansson (2007) explains that the required production volume can be estimated during the planning stage. The level of known and future demand affects how a product is ideally designed. For example, a higher level of product variety can be acceptable when the level of customer demand is low (Thonemann and Bradley 2002). Furthermore, additional information such as demand variability, forecast error, and product seasonality can be used to evaluate a potential product's design lead times, cost, and time to market (Dowlatshahi 1996;Khan, Christopher, and Creazza 2012). Finally, in addition to identifying demand for the new product under development, practitioners may need to assess when old products are in demand, periods when both the old and new products are in demand, and periods when only new products are in demand in the marketplace (Jafarian and Bashiri 2014).

Inventory management
Practitioners can also determine the levels and locations of safety stocks for new products throughout the supply chain, in order to maintain an appropriate trade-off between customer service and costs (e.g. Fine 2009;Hatch and Badinelli 1999a;Hatch and Badinelli 1999b;Huang, Zhang, and Liang 2005;Wang and Shu 2007) (Table 5). In the manufacturer's supply chain, each node is a potential location for holding safety stock inventory (Nepal, Monplaisir, and Famuyiwa 2011). If needed, spare assets for the repair of returned failed products can also be kept on-hand (Hatch and Badinelli 1999a). Manufacturers typically use an ERP system to support inventory management within the supply chain (Appelqvist, Lehtonen, and Kokkonen 2004;Chiu and Lin 2016;Jafarian and Bashiri 2014;Johansson 2007;Johansson, Bellgran, and Johansson 2006;Kao 2006;Nellore and Balachandra 2001).
Inventory management should be considered when designing a product. For example, to ensure appropriate product lead times, products should ideally be designed in such a way as to draw upon available inventory in the supply chain as much as possible (Wang and Shu 2007). By considering inventory management at an early stage, inventory levels can also be reduced by minimising product variety (Sharifi, Ismail, and Reid 2006). Furthermore, Dowlatshahi (1996) explains that a high-quality product design in terms of stability and robustness reduces inventory levels, due to increased efficiency in the manufacturing process (e.g. reduced set-up time, predictable and stable lead times, faster throughput time). In turn, the approach to inventory management should be tailored to the new product under development. For example, it is necessary to account for inventory turnover rates, seasonal demand, and seasonal inventory (Dowlatshahi 1996(Dowlatshahi , 1999Johansson, Bellgran, and Johansson 2006;Zolghadri, Baron, and Girard 2008). Furthermore, when a product is critical for achieving a competitive advantage and has a short life cycle, the physical efficiency of just-in-time models can be combined with the responsiveness dictated by the product (Noori and Georgescu 2008). Alternatively, the manufacturer's supply chain can be geared towards small batches when a product is modular, has a low impact on competitive advantage, or has a short life cycle (Noori and Georgescu 2008). A modular product design also makes it possible to decentralise safety stocks (Elmaraghy and Mahmoudi 2009). For example, in a configure-to-order environment, product modularisation enables the allocation of module inventory at various nodes in the supply chain (Kristianto and Helo 2010;Lee and Sasser 1995). The decentralisation of stocks combined with product modularisation makes a manufacturer more responsive to orders and results in lower inventory investments (Lee, Billington, and Carter 1993). Less investment is needed due to lower inventory levels at individual nodes and there is less of a need to correct for stock imbalances between nodes (Lee and Sasser 1995). Therefore, even though product modularisation may require extra investment to arrange the common parts for use in some of these products, there is a great advantage in risk-pooling (Hsu and Wang 2004). This approach to inventory management is especially appropriate when a new market trend renders customer satisfaction a manufacturer's main objective (Elmaraghy and Mahmoudi 2009).
To execute an inventory policy, order quantities for items at suppliers can be set based, for example, on procurement lead times (Dowlatshahi 1992;Huang, Zhang, and Lo 2007;Wang, Huang, and Dismukes 2004;Yan, Yu, and Cheng 2003;Zhang, Huang, and Rungtusanatham 2008). Claypool, Norman, and Needy (2014) add that optimal order quantities for each item can be specified when the product BOM comes available during the detail design stage. Gokhan, Needy, and Norman (2010) add that this can be done by considering the impacts of order quantities on profits over the product life cycle. It is also important to set the delivery and collection frequencies of items in the supply chain (Johansson 2007;Noori and Georgescu 2008). These frequencies, the volumes supplied and delivered in the supply chain, and the production schedules influence each other, which should be considered during PD . Focussing on the materials flow both within as well as to and from a node (e.g. manufacturing plant) during PD, practitioners can also decide whether items will be supplied continuously, in batches, kitted, or sequentially (Johansson 2007). Towards the end of the PD process, safety stocks can be reduced when the materials flow becomes stable (Johansson 2007). Finally, the introduction of a new product may need to be planned simultaneously with the elimination of an old product during PD. Therefore, the inventory of the product assortment should be updated according to product life cycles, and inventory of obsolete products should be properly phased out during the production ramp-up stage of PD (Hilletofth, Ericsson, and Lumsden 2010;Hilletofth and Eriksson 2011).

Capacity planning
Practitioners can determine the size and location of capacity needed in the supply chain for the new product, based on a certain service level (Arntzen et al. 1995;Fixson 2005;Yan, Yu, and Cheng 2003) (Table 5). Capacity planning entails determining the product quantities that need to be processed by various actors in the manufacturer's supply chain, based, for example, on capacity restrictions (Appelqvist, Lehtonen, and Kokkonen 2004;Hatch and Badinelli 1999a;Yadav et al. 2008;Yadav et al. 2011;Yan, Yu, and Cheng 2003). Capacity decisions should be based on product characteristics. For example, the manufacturer's supply chain can be geared towards capacity-planning flexibility and reliable and fast logistics when a product is modular, has a low impact on competitive advantage, or a short life cycle (Noori and Georgescu 2008). When a product is of strategic importance to the manufacturer, capacity decisions should be based on the overall planning objectives in the determined horizon(s) and strategic business goals. Capacity planning is also influenced by the product's production characteristics, which include lead times, set-up time, throughput time, and the length/size of the production run (Dowlatshahi 1996(Dowlatshahi , 1999. Capacity can be planned during various stages of the PD process. Johansson (2007) explains that the production start can be estimated during the planning stage, and describes that the production pace and the ramp-up requirements of customers can be identified during the concept development stage. The transportation capacity of items and products to and from the manufacturer's node(s) can also be planned according to a certain service level early in the PD process (Fixson 2005;Yadav et al. 2008;Yan, Yu, and Cheng 2003). This means that it is necessary to plan transportation according to how fast products should be delivered to customers (Khan, Christopher, and Creazza 2012;van Hoek and Chapman 2007). Gokhan, Needy, and Norman (2010) show that the necessary transportation capacity can be planned between supply chain actors while evaluating the overall lead time of various product design alternatives during detail design. Focussing on the same stage, Chiu and Lin (2016) propose that future capacity requirements can be anticipated by using material consumption and time consumption data. Finally, the introduction of a new product may also need to be planned simultaneously with the elimination of an old product. Therefore, capacity planning can aim at updating the product assortment according to product life cycles and properly phasing out obsolete products during the production ramp-up stage of the PD process (Hilletofth, Ericsson, and Lumsden 2010;Hilletofth and Eriksson 2011).

Summary of the findings
Through systematically reviewing the literature, 14 SCD activities are identified (Tables 2-5) and grouped into levels and areas ( Figure 2). Strategy-level SCD activities (Table 2) cover the areas of sourcing, collaboration, and postponement. Network-level SCD activities (Table 3) cover the areas of supplier nodes, storage nodes, manufacturing nodes, and nodal links. Node-level SCD activities (Table 4) cover the areas of flow layout, storage areas, materials handling equipment, and packaging. Planning-level SCD activities (Table 5) cover the areas of demand planning, inventory management, and capacity planning. Together, the four levels of SCD activities aim at optimally designing the materials flow from the supplier(s) to the customer(s) in a manufacturer's supply chain (internal as well as external). The reviewed literature also shows that SCD interrelates with nine product characteristics, which include life cycle, modularity, criticality, variety, innovativeness, cost, quality, complexity, and physical properties (Table 6).
In order to depict the interrelations between SCD and these product characteristics, a model is developed and presented in Figure 3. This model provides an understanding of which product characteristics to consider during SCD, which SCD areas to consider when determining product characteristics, or which SCD areas and product characteristics to concurrently determine. Figure 3 shows that SCD interrelates with various product characteristics, which increases the overall complexity of PD projects. Furthermore, the SCD activities discussed in Section 4 and listed in Tables 2-5 are both holistic and generic. This means that not all SCD activities are relevant for every manufacturer or PD project, and those that are must be adapted locally and linked to PD. Therefore, Figure 4 highlights a framework that enables practitioners and scholars to perform or research SCD. The framework aims at developing company-and project-specific checklists that can be used for multiple purposes, including planning, performing, or evaluating SCD activities before, during, or after PD projects.

Discussion
The first column of the checklist(s), 'Area', describes to which higher-level grouping (e.g. sourcing) SCD activities belong. This increases the navigability of the checklist, especially when managing many SCD activities. The second column of the checklist(s), 'Activity', lists the SCD activities that  are relevant during PD. Tables 2-5 can act as input for this column and the level at which to describe SCD activities depends on the checklist's purpose. The third column of the checklist(s), 'Timing', shows when SCD activities are relevant throughout the PD process. For example, timing can refer to generic PD process phases (e.g. planning, concept development, system-level design, detail design, testing and refinement, production ramp-up [Ulrich and Eppinger 2016]). The fourth column of the checklist(s), 'Product characteristic', highlights which key characteristics (e.g. product life cycle) to consider when performing an SCD activity. The fifth and final column of the checklist(s), 'Description', shows how SCD activities are or should be performed. This column can be used to plan, track and evaluate the performance of SCD activities and, ultimately, to detail lessons for enhancing the performance of SCD activities in future PD projects. Before, during, or after PD projects, relevant stakeholders (e.g. procurement managers, transportation engineers, packaging designers, manufacturing engineers) can be brought together by organising joint workshops that aim at completing or revising the checklist(s) by incorporating what has been learned from past experiences. Some PD projects (e.g. radical innovation projects) may require more workshops due to their unique nature. Thus, the framework visualised in Figure 4 can constitute a foundation for conducting workshops at manufacturers, with the aim of developing company-and project-specific frameworks for SCD during PD.

Conclusion
There is a call for more resilient supply chains (Handfield, Graham, and Burns 2020;Linton and Vakil 2020;van Hoek 2020), which requires assessing how material moves through supply chains, compressing the time inventory remains in supply chains, and optimising the path of materials flows (Handfield, Graham, and Burns 2020). The paper explores how this approach to SCD can be adopted during PD. More specifically, on the basis of data collected through systematically reviewing the literature, 14 SCD activities are presented (Tables 2-5) and interrelated with nine product characteristics ( Figure 3). Furthermore, Figure 4 provides a framework for developing company-and project-specific checklists that can be used for multiple purposes, including planning, performing, or evaluating SCD activities before, during, or after PD projects.

Theoretical and practical contribution
This paper contributes to both theory and practice. Regarding the theoretical contribution, the paper contributes to knowledge by providing a coherent overview of the SCD activities that can be performed during PD. Another contribution is the development of a model that shows the interrelations between SCD and product characteristics. This provides a deep insight into the literature on SCD during PD. Regarding the practical contribution, this paper proposes a framework for SCD during PD. The implementation of the framework, according to the characteristics of a manufacturer and its PD projects, constitutes an opportunity for the creation of resilient SCD. For example, supply chain functions can reflect on how they can contribute to PD, and product designers can reflect on how they can contribute to SCD. This knowledge may also help process owners and PD project leaders to manage SCD activities throughout the PD process.

Recommendation for future research
Since the articles reviewed in this paper tend to focus on manufacturers of consumer goods, more research is needed on how manufacturers of 'complex products and systems' (Davies and Hobday 2005) design their supply chains during PD. Future research on SCD during PD can further benefit from borrowing concepts and ideas from the strategic management literature. Here, scholars propose that strategy work is based on activities that are intentional or unintentional, routine or non-routine, and formal or informal (Tidstr€ om and Rajala 2016; Tsoukas 2010). Furthermore, activities can Table 6. Product characteristics that interrelate with SCD.

Product characteristic
Description Reference (e.g.) Life cycle The amount of time a product can compete in an industry before becoming obsolete and requiring replacement. Nepal, Monplaisir, and Famuyiwa (2011); Noori and Georgescu (2008); Primus (2017) Modularity The amount and strength of the relationships among items in a product design. Fine, Golany, and Naseraldin (2005); Nepal, Monplaisir, and Famuyiwa (2011); Ulku and Schmidt (2011) Criticality The importance of a product for the realisation of a competitive advantage. Noori and Georgescu (2008); Wang and Shu (2007); Wynstra, Van Weele, and Axelsson (1999) Variety The number of product variants that belong to the same product family, based on their characteristics. Lee and Sasser (1995); Nielsen and Holmstr€ om (1995); Salvador, Forza, and Rungtusanatham (2002) Innovativeness The extent to which a product is new to an industry or the manufacturer.
Melander and Tell (2019) The monetary worth of a product. Claypool, Norman, and Needy (2014); Graves and Willems (2005); Pham and Yenradee (2017) Quality The robustness and stability of a product, and to what extent a product meets customer needs and standards.
Brewer and Arnette (2017); Di Benedetto et al. (2003); Kristianto et al. (2012) Complexity The number of items listed in a product BOM, and the uniqueness of a product. Mather (1992); van Hoek and Chapman (2007) The substance effects, mass, volume, and density of a product.
Appelqvist, Lehtonen, and Kokkonen (2004); Dowlatshahi (1999); Lee (2001) result in realised or unrealised manifestations of strategy (Mintzberg and Waters 1985;Mirabeau and Maguire 2014). By regarding SCD as 'strategy work', SCM scholars can create a contextualised understanding of SCD during PD. More specifically, future research can explore the very ways in which planned and unplanned SCD activities unfold, and whether they result in realised or unrealised manifestations of SCD strategy. This will inevitably be associated with incoherence, inconsistency, conflict, and dilemma (Blackler 1993), which are phenomena that offer major learning opportunities for SCM scholars and practitioners.

Disclosure statement
No potential conflict of interest was reported by the author(s).