Differentiation and customer decoupling points: An integrated design approach for mass customization

PDP

The following definition for PDP is proposed: The PDP denominates an operation that transforms a product common to all the products of the same family to a customized or personalized product (finished or not). This transformation can be done by the adjunction of specialized components and/or by the action of a special process. PDP can be multiple, as many attributes can differentiate products among a family. PDPs are not limited to manufacturing operations. Supply, pricing, and sales operation can also differentiate products.

The choice of PDP encompasses the two dimensions of MC optimization: product and process. Product differentiation is either studied on the product portfolio angle (how many different products should be offered) or on the position in the chain angle (should the PDP be moved to the customer). Rajagopalan and Xia (2012) studied the impact of product variety, localization, and pricing decisions on a supply chain with a small academic example. The setting of these three parameters affects the market demand and, therefore, both the global profit and its repartition within the supply chain members. Determination of optimal PDP localization should consider the connection between the physical entities (the products differentiate or not) to the information describing the market (demand for finished product; Zhang et al., 2013).

CODP

The CODP is defined as

the breaking point between productions for stock based on forecast and customization that responds to customer demand. It is also the breaking point between MTS and MTO, namely, activities before CODP are driven by forecast while activities after CODP are driven by real customer order demand. (Ji et al., 2007)

Its position can be at one of the five following stages: design, production, assembly, distribution, and consumer (Martínez-Olvera and Shunk, 2006). The main factors to consider when determining the CODP position are production process, set-up time, production technology cost, customer service level, production utilization rate, and the requirements of work in process for storage conditions and time (Ji et al., 2007). Many optimization problems to define the optimal position of the CODP were proposed in the literature such as Olhager (2003), Ji et al. (2007), Shao and Jia (2008), Jeong (2011), Rafiei and Rabbani (2011), and Abbey et al. (2013). The CODP position should be calculated based on the overall value generated by the system for each beneficiary party (e.g. customers, shareholders, and company). The CODP position influences the customization offer and determines how this offer is being satisfied. It is the main factor in classifying MC systems (Daaboul et al., 2012). It influences the performance of the network via its impact on costs (cost of resources, utilization of resources, cost of materials, inventory cost, production cost, and transportation cost), lead time (set-up time, processing time, order lead time, etc.), and quality (conforming orders, customer’s perceived quality of the product and the service, etc.). Its position should be optimized by taking into consideration not only capacity constraints, cost, and time but also the PDPs and the customization possibilities offered to the customer.

Coupled analysis

The questions of PDP and CODP positioning were often considered as exclusive alternatives. Dobson and Yano (2002) developed nonlinear, integer programming models for jointly optimizing product offering, pricing, MTO versus MTS, and cycle time decisions. Their analysis was not expanded to consider other structures such as assemble-to-order (ATO) or engineer-to-order (ETO). The analysis in this article supports their work in terms of making both decisions simultaneously. Su et al. (2005) considered form and time postponement as exclusive alternatives. They simulated the impact of those configurations on cost, using a simple model (two-stage single-server queuing system with exponential inter-arrival times and exponential service times) and limited instances (only 16 different products). They have concluded that once the number of products increases above some threshold, the time postponement structure is preferred under both performance metrics (cost and waiting time). Hegde et al. (2005) developed a framework to address what degree of customization to offer. Their model included two thresholds, defined as mismatch and manufacturing thresholds. Their results showed that product conformance gradually decreased when the degree of customization exceeded the manufacturing threshold. They have defined the customization offer by taking into consideration the manufacturing capability as a constraint. This article advocates that the decision on the customization offer should be made simultaneously with the decision on the manufacturing system configuration and, in particular, with the CODP determination. According to Ramdas (2009), “Variety creation and variety implementation decisions are inextricably linked, as both impact perceived variety as well as costs.” He includes in variety creation the decision concerning what and how many products to offer. Variety implementation focuses on how a firm’s manufacturing or service delivery processes and supply chain are managed to implement its variety creation strategy. According to Khalaf et al. (2010), in designing an efficient product family, designers have to anticipate the production process and the supply chain costs. They propose a design method in which the product and the supply chain design are considered at the same time. Shahzad and Hadj-Hamou (2012) propose a mixed integer linear program model including logistics and general bill of material constraints for strategic decisions about opening or closing of a market segment. Their work emphasizes the importance of considering not only the product design in decision making but also its production and logistics.

At first, why a value approach?

Cost is not a sufficient variable to base a strategic decision on it. Especially, in the case of a customer-driven strategy such as MC where the main objective is to increase customer satisfaction and better answer his specific individual needs. Yet, as Wang and Tseng (2008) advocated, technical merit is no longer enough to determine customer satisfaction. It relies on “qualitative and subjective factors such as affection, aesthetic appearance, and easy-to-use.”

Value incorporates in its definition tangible and intangible variables and thus permits the inclusion of intangible variables in the performance analysis. These variables include the image of the company, the customer satisfaction, and the service’s perceived quality among others. Moreover, by adopting a value approach, the strategic decision analysis is based not only on one factor, such as cost, but also on all factors affecting value such as delays, quality (of the product and of the process), service level, and customer satisfaction. Taking value as the main decision criterion allows the company to consider the influence of a change, choice, or decision both on its overall generated performance and more particularly on its main partners and its customers (Daaboul et al., 2011). Furthermore, as Ueda et al. (2008) emphasize in their extensive review of the evolution of value, value models enable the integration of the sustainability issue in a system in “which both the overall purpose and individual happiness” are considered.

Value is defined as the amount of satisfaction created by fulfilling a certain physical, biological, or psychological need of a beneficiary party. It is influenced by many criteria such as cost, delay, perceived quality, and perceived price. It can be objective or subjective and is dependent on the circumstances and tied to the specific goals of the beneficiary party. Thus, the value concept enables the inclusion of the customer’s perceived value, the suppliers’ generated value, the retailers’ generated value, and value generated for all partners in the analysis and not only the value for the main manufacturer.

Second, why a value network approach?

Nowadays, competition is between networks of companies, rather than between companies. Nowadays, no innovative product can be put on the market without a well-structured and well-organized network of partners (Zolghadri et al., 2008). The performance of one enterprise influences the performance of all its partners. For example, the performance of a supplier affects the performance of the whole network. Therefore, the decision-making process should be completed by taking all the partners in a network (shareholders, suppliers, retailers, distributors, customers, etc.) into consideration nowadays. Later, this concept of interconnecting businesses has been widely developed (Ivanov, 2009). In addition, the value chain concept, which is extended to value network, has become the “focal point of MC competition” (Fogliatto et al., 2012). Of the recent works advocating the use of value networks in decision making or in performance evaluation are Daaboul et al. (2012), El Fassi et al. (2012), Verdecho et al. (2012), and Macedo and Camarinha-Matos (2013).

Proposed integrated value network approach for MC design

The proposed approach is based on the assumption that each of product design and value network design affects the total generated value, including the enterprise profit and the customer satisfaction. The proposed methodology is based on a value network conceptual model proposed by Daaboul et al. (2013). The proposed conceptual model in Daaboul et al. (2013) is activity based. It is formed of two types of activities: decision and execution activities. An activity uses resources and has as input and output physical and informational flows. It is stimulated by a trigger. The model in brief is formed of eight main concepts. These are as follows: (1) partner or organization (manufacturer, distributor, supplier, retailer, customer, etc.); (2) physical flow (products, semi-finished products, etc.); (3) resources (man, machine, and tools); (4) informational flow; (5) activity (execution and decision activities); (6) financial flow; (7) immaterial variables (state and action variables, state variables enable performance to be measured, and action variables influence or affect performance); and (8) trigger that stimulates an activity. A trigger can be an informational flow (production order, a customer order, and a decision) or a physical flow.

The CODP and PDP positions are action variables. They both affect the performance of the company and the supply chain. They affect costs, lead times, and customer’s perceived quality. The CODP position defines the type of triggers for activities. All activities positioned after the CODP are triggered by the customer order. The PDP position affects the physical flows and the number of variants per product.

The proposed integrated value network-based method for MC design compares different possible simulated scenarios. It consists of five steps:

Value network modeling and simulation: methodology implementation

Identify possible PDPs or CODPs and thus possible MC scenarios

A shoe is formed essentially of an upper, outsole, insole, midsole, eyelets, buckle, laces, and accessories. After analysis and discussions with Alpina, two possible PDPs were identified as shown in Figure 1. The first PDP is at the laced shoe stage, meaning that the customer can only customize the color of the laces. The second PDP is at the upper stage, meaning that the customer may customize the color of their shoe, the type and color of the outsole, and the laces. Other PDP positions are possible, these two positions were chosen for first tests since they represent the two extreme possible positions of PDP: the shoe is a standard or common product until the end stage (first PDP) or is a customized product from the first stages.

OPEN IN VIEWER

Figure 1.

Possible PDP and CODP positions.

The shoe production process consists essentially of seven main processes: outsole production, insole and midsole production, cutting upper parts, stitching upper parts, lasting upper and assembling it to the soles, cleaning the shoes, and adding laces. This process leads to five possible CODPs as shown in Figure 1.

The AS-IS scenario (current situation) is a combination of PDP 1 and CODP 5. It is designated as scenario 1. Based on the possible PDPs and CODPs, nine TO-BE scenarios were identified:

Scenario 1: PDP 1 or CODP 5

Scenario 2: PDP 1 or CODP 1

Scenario 3: PDP 1 or CODP 2

Scenario 4: PDP 1 or CODP 3

Scenario 5: PDP 1 or CODP 4

Scenario 6: PDP 2 or CODP 1

Scenario 7: PDP 2 or CODP 2

Scenario 8: PDP 2 or CODP 3

Scenario 9: PDP 2 or CODP 4

The scenario formed of PDP 2 and CODP 5 is not feasible since it is not possible to have such customization of the shoe and have the CODP positioned at the last stage.

Model and simulate current value network

Collect necessary data

The data collection process which included different departments and experts in Alpina was carried out in four phases as follows:

Phase 1. Excel templates filled out by the company including data on the main activities, physical flows, and the main partners of the network (suppliers, distributors, retailers, and customers). The data collected for the execution activities consisted of name of the activity, description of the activity, execution time, set-up time, partner undergoing the cost, partner collecting income, cost per unit of physical flow, cost of transportation (if the activity of transportation is not modeled by itself), used resources (and quantities), and the inputs and outputs of the activity.

The data collected for the physical flow consisted of name, characteristics of the flow and possible values for each characteristic, partner undergoing the storage cost, and the initial stock for every variant of the flow.

The data collected for the partner consisted of name; type (supplier, retailer, distributor, and customer); and location and for suppliers supplied material, supply strategy, replenishment lead time, supply frequency, minimum quantity to supply, pricing model, and inbound transportation cost.

These are the needed data for modeling an execution activity, a physical flow, and a partner based on the chosen framework.

Phase 2. Collection of data necessary for modeling the decision activities. This was achieved via filling out an excel template with the decision maker including the decision function, variables, constraints, conditions, and actions to be taken based on the decision made. An example of a decision is the supply strategy (when to order and from which supplier).

Phase 3. Collection of production plans via an excel sheet provided by Alpina.

Phase 4. Collection of sales history in any format provided by the company. We collected sales history for the previous 4 years.

It was necessary to have several meetings to collect these data. This was the lengthiest task in the entire case study. Alpina’s value network included 20 partners (Alpina, customer, shop, distribution company, and 16 suppliers); 88 physical flows (e.g. shoe, laces, insole, outsole, leather, glue, thread, and carton box); 100 execution activities; and 30 decision activities.

Develop the influence model

This step consists of modeling the generated value for all partners. In this case study, and due to the lack of all the required data concerning the suppliers and distributors, the analysis is limited to the value generated for Alpina and for the customer. Concerning the value for Alpina, its related objective performance indicators are revenue and cost. Its subjective performance indicators are image or reputation, ranking among competition, service level, customer’s loyalty, and employees well-being. The objective value is measured, and it is equal to Alpina’s revenue from which we deducted the total cost it induced in the network. The subjective value is not measured but integrated via the influence network. This influence network models all variables and their interdependencies and inter-influences.

Customer’s perceived value is more complex to calculate. It is strictly subjective as shown in Figure 2. It is equal to the customer’s perceived value of the product added to the service value and then divided by the perceived price.

OPEN IN VIEWER

Figure 2.

Customer’s perceived value.

To calculate the customer’s perceived quality, we use the quality model proposed by Olson (1977). He differentiates the intrinsic attributes from the extrinsic attributes of quality. The intrinsic attributes concern the product itself and consist of its physical composition. They cannot be changed without changing the nature of the product itself. On the contrary, extrinsic attributes are not part of the physical composition of the product. Based on this model, we define the intrinsic and extrinsic attributes of quality and determine their value equivalent to five evaluations: very poor, poor, average, good, and very good quality. Each of these evaluations has an assigned grade ranging from 1 for very poor quality to 5 for very good quality. Weights of importance for men and for women are also determined. The perceived quality is then equal to

Q P i = ∑ j = 1 n Q A ij × W A ij

where Q P i is the total perceived quality of product i (1 ≤ i ≤ n; i ∈ N ), Q A ij is the quality of attribute j of product i (1 ≤ j ≤ m; j ∈ N ), and W A ij is the weight of importance of attribute j of product i (different for men and women, see Table 1).

OPEN IN VIEWER

Table 1.

Quality model for case study.

	Name	W A ij , men	W A ij , women	Very poor quality	Poor quality	Average quality	Good quality	Very good quality
Intrinsic attributes	Fitting	0.9	0.92	Not fitting for both feet	Fitting for one foot	Fitting more or less	Great fitting for a foot, good fitting for the second	Perfect fitting for both feet
	Fashion ability	0.66	0.74	Not fashionable	Old fashion	Classical	On fashion	Fashion creator
	Thermal comfort	0.88	0.84	>6°C	6°C	5°C	4°C	<4°C
	Flexibility	0.88	0.82	<2 or >8.99	2–2.99 or 8–8.99	3–3.99	4–4.99 or 6–6.99	5–5.99
	Weight	0.76	0.82	>500 g	450–499 g	400–449 g	350–399 g	<350 g
	Stability	0.88	0.78	Unstable	Nondense midsole	Stable midsole	Stable Sole	Dual-density midsole
	Breathability	0.88	0.84	<0.15 mg/cm2 h	0.15–0.29 mg/cm2 h	0.3–0.49 mg/cm2 h	0.5–0.99 mg/cm2 h	>0.99 mg/cm2 h
	Durability	0.74	0.78	<0.5 years	0.5–1 year	1 year	2 years	>3 years
	Material	0.72	0.82	Poor synthetic	Plastic	Full grain side leather	Kidskin (goat)	Pigskin leather, calfskin (young cattle)
	Water proof	0.71	0.74	<20 min	20–39 min	40–59 min	60–80 min	>80 min
Extrinsic attributes	Brand name	0.56	0.54	Unknown brand	Unknown brand	Jello, Reebook, Converse	Alpina, Peko, Nike, Adidas	Deichmann
	Ecologic	0.52	0.5	No natural or recycled material is used	Less than quarter of material is natural and recycled	Half of the shoe material is recycled and natural	Most of the shoe is made out of natural and recycled materials	Shoes made out of natural and recycled materials
	Advice from shop personnel	0.68	0.66	No advice at all	Small talk	Advice about the actual fashion and best sellers	Advice based on few information collected from customer	Advice based on discussion with customer

The quality model for casual shoes shown in the following table was developed with experts from Alpina, from their marketing office in particular.

The service value is affected by the following three main indicators:

The time necessary for customization: customization process indicator (CPI) = (total time of customization or maximum tolerated time for customization)×penalties;

The customization offer: used variety indicator ( UVI ) = ∑ i = 1 n w i × v i / NV

where W_i is the average weight of importance of variant i (1 ≤ i ≤ n; i ∈ N ), V_i is the variant i, and NV is the total number of variants;

The order delay time = penalties×(the time of reception of order − the time of placing the order).

Finally, the perceived price was assumed to be equal to the price of the shoes.

Construct and simulate the model

As a modeling and simulation tool, value network library developed for Arena software (Rockwell) was used. This library is explained in Daaboul et al. (2014). Arena is discrete event simulation software developed by Systems Modeling in the 1980s. After all data are gathered, it takes almost two full working days to create the model. Its validation and refinement takes approximately 1 week.

Validate and refine the simulated model

The simulation resulted in an average of 347 pairs of shoes produced per day, whereas in reality, the average pairs produced per day is equal to 350. The simulation result for a product cycle time is 104.25 min compared to 105.62 as real value. In the end, all the delays were also validated. For example, the shop replenishment delay when having the material needed for production is equal to 3–4 weeks in both simulation and reality results. Otherwise, it is equal to an average of 57 days in simulation results and 60 days on average in reality.

Simulate all predefined TO-BE scenarios and results analysis

The obtained results were validated by Alpina. The results were coherent with other analyses and studies realized at Alpina. As shown in Figure 3, the value for Alpina is higher in the case of PDP at position 1. This is due to the fact that Alpina was not ready to make the necessary changes and investments such as changing its agreements with its suppliers in order to offer MC shoes. While the customer’s perceived value has a less predictable behavior, it is affected by many indicators such as perceived quality, offered customization, and order lead time.

OPEN IN VIEWER

Figure 3.

Simulation results.

The results show that the value for Alpina is higher for PDP 1 than PDP 2, no matter what the position of CODP. The value for Alpina is highest when the CODP is further in the chain (CODP 5). In addition, it reduces dramatically for CODP 1, 2, 3, and 4 compared to CODP 5. This is due to the following:

An increase in the cost of raw materials caused by the increase in the required stock for each material.

Reduced economies of scale as the size of the production batch in the case of MC is equal to 1.

An increase in the cost of usage of resources, since the execution time of an activity increased with increasing set-up time.

A slight increase in transportation costs caused by increased late orders that require faster delivery and, therefore, more expensive.

A slight increase in turnover. As part of the MC, Alpina estimated a 5% increase in sales. However, this increase in sales (an increase of 37,559 Euros) was not sufficient to cover the additional costs.

Increased costs due to suppliers’ constraints: Analyzing the Alpina case study focusing on the product, the process, and the supply chain, one realizes that the main limit to implement MC is the supply chain. This is similar to the case of Milk & Honey (Kieserling, 2011) which had to relocate its entire factory from Italy to China to overcome the limits and constraints of the supply of leather. With its first business model, Milk & Honey was not able to profitably offer mass customized shoes due to the large replenishment time of leather suppliers, their lack of flexibility, the large minimal quantity to order different leather variants, and the high cost of the leather. By moving their facility to China, Milk & Honey operate a factory based on European standards and laws but benefit from a highly flexible supply chain with highly reduced replenishment time and minimal order quantity. The Alpina case study validates the choice of these two companies to move their production to China. Such as Milk & Honey and Selve (Kieserling, 2013), Alpina’s supply of leather is its main limit and constraint to successfully and profitably offer MC.

Scenario 9 (CODP is at position 4 and PDP at position 2) results in a major decrease in the value for Alpina. These results were analyzed. It was found that when stitching the upper parts before receiving the orders, the stock of stitched uppers is high. Moreover, this is an overstock since almost all combinations are available, whereas there are no orders for all possible combinations. Increased cost of stock leads to a decreased profit for Alpina.

The customer’s perceived value is higher for PDP 2 than PDP 1 whatever the position of CODP. This is due to the impact of the customization offer on the customer’s perceived value. Moreover, the customer is willing to wait longer in order to receive a customized shoe. The customer’s perceived value is very low for the combinations CODP 1/PDP 1 and CODP 1/PDP 2; this is a logical result, since the customer is waiting much longer than he or she should to receive a noncustomized shoe. The customer’s perceived value is highest for the combination CODP 1/PDP 2. For CODP 1, the order lead time range announced to the customer was often met, whereas this was not the case for CODP 2 and CODP 3.

This problem is a multiobjective optimization problem. Actually “the solution of a multiobjective optimization problem is found in terms of a discrete approximation of the Pareto front. Then, the decision maker has to select one or more solutions of preference within the Pareto Front” (Zio and Bazzo, 2011). If this task is very hard for the decision maker, he or she can use methods such as level diagram analysis of the Pareto front (Zio and Bazzo, 2011).

The Pareto front for Alpina’s results was built (Figure 4). It shows the dominating combinations of CODP and PDP. These are PDP 2/CODP 1, PDP 2/CODP 3, PDP 2/CODP 2, PDP 1/CODP 3, and PDP 1/CODP 5. The decision maker (Alpina) chose the last combination without hesitating which reflects the current situation. The increase in customer value does not balance the decrease in the value for Alpina. The company wishes at first to negotiate its suppliers’ contracts or find a solution for the nonflexibility and high replenishment lead times for leather supply. Then, based on the negotiation results or solutions found, the analysis will be repeated to simultaneously choose the CODP and PDP positions.

OPEN IN VIEWER

Figure 4.

Scenarios’ performance and Pareto front.