Conclusions about the hypotheses

In this paragraph I draw conclusions upon the hypotheses. For each hypothesis, I will summarize findings from Chapter 4, and explain these findings within the context of this and prior research which is examined in Chapter 2.

Before I start with the pattern matching, I would like to reflect on the term customization. The term is dominant in this thesis; it is the most important term in the overall research question and in the posed hypotheses as well. In this thesis, the term is used within two different contexts. In the first context, it is used as a strategy for suppliers, which is based on two characteristics. The first characteristic is the point in the production cycle of consumer involvement in specifying the product; the second is the type of modularity employed (Duray et al., 2000). In the second context it is used as the possibilities for consumers to interact with suppliers to create or specify a product. When these two contexts are combined, it fits the customization definition which is used for this thesis formulated in section 2.2.1. Customization as a strategy is commonly known as mass customization and is more generic of nature, while customization by consumers when interacting with the supplier is more specific.

H1: Larger variety enables higher levels of customization

The first hypothesis states that a large product variety enables more possibilities for customizing products. Based on the literature review I assume that the characteristics of digital products in combination with consumer-supplier interaction over the Internet make it possible to apply versioning, personalization and bundling of digital products. An increase in variety should increase consumer-supplier interaction at the design stage of the operations level to create customized digital products delivered over the Internet. The first hypothesis which resulted from the literature review was the following:

H1: The larger the product variety, the higher the level of customization.

I start with a discussion of the proposed pattern for the first hypothesis versus the analyzed pattern in the cases. I finish this section with a discussion on the relation between a large product variety and the degrees and possibilities of customization based on the data analysis in Chapter 4, and discuss the differences between the cases.

Both cases support hypothesis H1 on customization on the general or strategic level. Both cases have a large product catalogue, and that large catalogue supports a high level of customization. A high level of customization is the case when consumers are involved early in the production cycle (Duray et al., 2000). Last.fm as well as Pandora Media can be classified as the involver mass customization classification because consumers are involved early in the production cycle, and the type of modularity employed is during the assembly and use stages. The modules are not manufactured or changed by consumers, but arranged or combined according to consumer specification. When variety is high, modularity is an enabler for mass customization. Modularity at the assembly and use stages seems to be a suitable strategy for suppliers of digital products.

The cases confirm that a large variety of digital products leads to a high level of customization in terms of interaction possibilities for consumers with the supplier, only the factor trust was not influenced by any of the variety factors. An overview of the influence of product variety on customization that can be identified based on the case studies is summarized in Table 5.1. When both cases have similar results on the influence of variety on customization, it can be recognized in the second column of the table. The specific factors of consumer involvement and the type of modularity employed are summarized between the brackets in behind the factors in the last column.

To proof the influence of increased product variety on customization, I will use the data analysis of Chapter 4 on product variety. For each variety factor that influences customization, I will start with the influence on consumer involvement and modularity. This is followed by the influence on the interaction possibilities for the consumer.

Both cases have a high value of the multiple use metric, marked with the plus (+) sign at Table 4.4 and Table 4.11. Because this metric is high, components are shared across products, which enable modularity at the assembly and use stages in the production cycle. Besides using the same modules on multiple products, Last.fm makes use of the submitted metadata from modules that are not available at the supplier, which makes it easier to make better automatic recommendations.

The interface complexity is low for both cases. This makes it easy for the consumer to start interacting with the system, and to be involved early in the design process. Both cases illustrate that it leads to consumer’s requests to be uniquely designed into the finished product. The Last.fm case study shows that the low interface complexity has a positive influence on customization because it enables consumer involvement by means of the ability to specify new product features, and to select features from listings. Both cases explain that the interface enables various customization possibilities in terms of interaction. The Last.fm case study shows that the interface can support the use of defaults, by offering default versions that can be customized. Both cases show that it can increase consumer expertise, Last.fm by having a community whose members can interact with each other on the Last.fm website and Pandora by providing feedback on why a particular track is playing or has been played. Both case studies also show that the interface can provide the possibility to make direct recommendations. At Last.fm it is possible to recommend a particular track, album or artist; at Pandora it is possible to recommend a product by sharing it with other consumers.

Product variety

Case

Influence on customization

Multiple use

Both

Type of modularity employed (components are shared around products)

Last.fm

Automatic recommendation

Pandora

Interface complexity

Both

Consumer involvement (consumer’s requests are uniquely designed into the finished product), direct recommendation, consumer expertise

Last.fm

Consumer involvement (consumers can specify new product features, consumers can select features from listings) defaults

Pandora

Platform efficiency

Both

Type of modularity employed (options can be added to a standard product, components are shared across products, products are designed around a common core technology), automatic recommendation, collaborative co-design

Last.fm

Pandora

Type of modularity employed (products have interchangeable features and options)

Table 5.1: Influences product variety on customization, matched from cases.

The companies from both case studies have an efficient platform. An efficient platform has a positive influence on customization by means of the type of modularity employed during the assembly and use stages because products are designed around a common core technology. Both case studies also illustrate that an efficient platform makes it easy to add a module to an existing product, as in bus modularity (Ulrich and Tung, 1991, in Pine, 1993), and makes it easy to share components across products. The case study at Pandora shows that an efficient platform leads to products that have interchangeable features and options by moving modules to other products. In terms of specific customization possibilities, both cases show that an efficient platform leads to automatic recommendation and collaborative co-design. Automatic recommendations play an important role in both case studies, because it helps discovering new music, which is one of the reasons the companies exist.

It can be concluded that larger product variety leads to higher levels of customization, because larger product variety enables modular products and consumer interaction. Variety is a tool for customization, because it can create a link between the consumer and the product (Svensson and Jensen, 2001). The data analysis proves that the product variety for both cases is high. The factors that determine product variety have a positive influence on almost all of the customization factors, for both customization as a strategy and customization as interaction possibilities for consumers with the supplier. The Last.fm case study shows that the large perceived variety enables more customization possibilities than the Pandora case study. This difference can be explained by the more varied assortment of Last.fm.

H2: Higher levels of customization enables larger variety

The second hypothesis also supposes a relationship between variety and customization, but where customization was the dependent variable for the first hypothesis, it is variety that is the dependent variable for hypothesis H2. Again, this relationship is a positive one. The second hypothesis is the following:

H2: The higher the level of customization, the larger the variety.

Customization

Case

Influence on variety

Consumer involvement

Both

Last.fm

Pandora

Type of modularity employed

Both

Multiple use (components are shared across products)

Last.fm

Pandora

Multiple use (products have interchangeable features and options)

Attribute vs. alternative

Both

Interface complexity

Last.fm

Pandora

Default version

Both

Last.fm

Interface complexity

Pandora

Consumer expertise

Both

Last.fm

Pandora

Automatic recommendation

Both

Multiple use, interface complexity

Last.fm

Pandora

Direct recommendation

Both

Multiple use, interface complexity

Last.fm

Pandora

Trust building

Both

Last.fm

Pandora

Collaborative co-design

Both

Last.fm

Multiple use

Pandora

Table 5.2: Influences customization on increased product variety, matched from cases.

Like the previous hypothesis, I start with a discussion of the proposed pattern for the second hypothesis versus the analyzed pattern in the cases. I finish this section with a discussion on the relation between customization and the influence on variety based on the data analysis in Chapter 4 and then I discuss the differences between the cases.

Both cases partly support hypothesis H2. The case studies illustrate that six of the nine identified customization factors increase variety on two out of the three variety factors. The multiple use and interface complexity factors are influenced in a positive manner, which are similar with the proposed pattern of hypothesis H2. However, the case studies can not show that any of the customization factors can increase variety on the platform efficiency factor. Modularity increases variety on the multiple usages of modules. All specific customization factors except for consumer expertise and trust building increase variety on the multiple use factor or the interface complexity factor, some specific customization factors influence variety even on both factors. Table 5.2 summarizes the influence of customization on an increase in variety. The specific factors belonging to the type of modularity employed are placed between brackets after the influenced variety factor.

Consumer involvement, as one of the characteristics to classify a mass customizer, does not have any influence on variety. The second characteristic to classify a mass customizer, the type of modularity employed, does have an influence on variety. Both cases show that the sharing of components across products during the assembly and use phases lead to an increase in product variety. It increases variety on the multiple use factor. The Pandora case study shows that the interchangeability of features and options also increases product variety. When modules are moved to or shared over a range of different products, they are used more often.

The second group of customization factors are the interaction possibilities for the consumer with the supplier. The cases show that not all these factors increase variety, and that these factors do not increase variety on all variety factors. One of the customization factors that do increase variety is presenting choice sets by attribute rather than by alternative. Both cases present their choice sets by attribute instead of by alternative when the choice set is large, which makes the interface less complex. Automatic and direct recommendations increase variety by increasing the multiple usages of modules. Both case studies also illustrate that these recommendations make the interface less complex, because new digital products and modules can be discovered more easily, which again can increase variety. The case study at Last.fm shows that the usage of a default version makes the interface less complex. The last customization factor that increases variety is collaborative co-design. The case study at Last.fm shows that collaborative co-design leads to musical neighbours, which enables the consumer to discover new music. Collaborative co-design increases variety by increasing the multiple usages of modules.

Based on the data analysis in Chapter 4, it can be concluded that the type of modularity employed, presenting large choice set by attribute instead of by alternative, default versions, automatic as well as direct recommendations, and collaborative co-design increase variety on two out of the three factors: multiple use and complexity. Consumer expertise and trust do not have any influence on an increase in variety, and platform efficiency is not influenced by any of the customization factors. Hypothesis H2 is supported by the cases, but not by all the customization factors on all of the variety factors. The Pandora case demonstrates more influence of the type of modularity employed, while the Last.fm case study explains more influence of default versions and collaborative co-design.

H3.1: Larger variety increases complexity

The third hypothesis is split up in two. The first part of hypothesis H3 suggests a positive relationship between an increase in variety and complexity. The literature on variety induced complexity unveiled the drawbacks of variety. The first part of the third hypothesis is the following:

H3.1: The larger the variety, the larger the complexity.

The cases can neither prove nor reject the proposed hypothesis H3.1. Two out of the three variety factors have an influence on the perceived complexity. According to theory, a complex interface leads to longer interaction with the supplier. However, both cases explain that the interface is not complex, and it is desired to increase the interaction with the supplier. The only factor that increases complexity based on the cases is the platform efficiency. An inefficient platform leads to more consumers leaving the platform. It is interesting to notice that according to theory, consumers leave because of a complex interface. The cases demonstrate, however, that consumers leave because of an inefficient platform, but they also leave because there is not enough variety. Table 5.3 summarizes the influence of product variety on complexity, based on the data analysis in Chapter 4.

Product variety

Case

Influence on complexity

Multiple use

Both

Last.fm

Pandora

Interface complexity

Both

Last.fm

Average interaction length of time

Pandora

Platform efficiency

Both

Abortion rate

Last.fm

Pandora

Table 5.3: Influence of an increased product variety on complexity, matched from cases.

Both cases confirm that an ineffective platform can lead to the abortion from the service. The case study at Last.fm shows that the usage of client software makes the platform less efficient, and the case study at Pandora illustrates that slow Internet connections leads to consumers leaving the service. The case study at Last.fm explains that some features are complex because they are hard to find. This increases the complexity because it increases the average interaction length of time.

Two out of the three variety factors have influence on complexity. The most interesting is the interface complexity. The Last.fm case study shows that a complex interface leads to an increase in complexity by increasing the average interaction length of time, while at the same time both case studies show that it is the purpose to increase the average interaction length of time. This interesting result can be explained by the purpose of both companies. They try to let the consumer discover new digital products, instead of letting them find their digital product they are searching as fast and easy as possible. According to theory, an interface that is not complex reduces complexity because it decreases the average interaction length of time. However, the cases show that an interface that is not complex, is not decreasing the interaction length of time, but is encouraging consumers to interact more with the system. However, an unnecessary increase of the interaction length of time is not desired.

H3.2: Larger complexity increases search costs

The second part of the third hypothesis extends the influence of variety. Hypothesis H3.1 proposes the relationship between increased variety and complexity, hypothesis H3.2 suggests a positive relationship between larger complexity and higher search costs. The second part of the third hypothesis is the following:

H3.2: The larger the complexity, the higher the search costs.

The cases support hypothesis H3.2 to some extent. It is true that the average interaction length of time is related to the number of alternatives searched. According to theory, the average interaction length of time should be minimized to reduce complexity (Blecker et al., 2006), and the number of alternatives searched should be minimized to reduce search costs (Blecker et al., 2006; Kurniawan et al., 2006). The cases demonstrate however that it is not the purpose to decrease the interaction length of time, but to increase interaction with the consumer. Table 5.4 summarizes the influence of complexity on search costs, based on the data analysis in Chapter 4.

The cases show that only one of the complexity factors has an influence on search costs. Both cases show that an increase in the average interaction length of time leads to an increase in the number of alternatives searched. According to theory, this leads to higher search costs. The case study at Last.fm explains that the average interaction length of time has a negative influence on the product accuracy. The hard to find features on the Last.fm website make it more difficult to find products more accurately. The cases do not indicate that the abortion rate leads to an increase in complexity, which therefore shows no relation with search costs. Both case studies can not confirm a relationship between any of the complexity factors and search agents.

Complexity

Case

Influence on search costs

Average interaction length of time

Both

Number of alternatives searched

Last.fm

Product accuracy

Pandora

Abortion rate

Both

Last.fm

Pandora

Table 5.4: Influence of complexity on search costs, matched from cases.

H4: Higher levels of customization decrease search costs

The last hypothesis of the research framework suggests a relationship between customization and search costs. This hypothesis is the only hypothesis that suggests a negative relationship between two themes of the framework. The literature review suggests that high levels of customization should lower the search costs for consumers. The fourth and last hypothesis is the following:

H4: The higher the level of customization, the lower the search costs.

The cases support hypothesis H4, albeit not completely. Except for two customization factors, the data analysis of Chapter 4 confirms that customization decreases search costs on all three search costs factors. Both case studies show results that are similar for most customization factors. There are some differences on the factors attribute vs. alternative and default versions. A summary of the influence of customization on the lowering of search costs is given in Table 5.5.

The case studies illustrate that consumer involvement, as being one of the characteristics to classify a mass customizer, decreases search costs for the consumer. Because consumer’s requests are uniquely designed into the finished product, the product accuracy is improved, and the number of alternatives searched is decreased. The Last.fm case study shows that this can be achieved because consumers can customize the way recommendations are being made. Both case studies explain that when consumers are providing feedback, it results in presenting more accurate products to consumers. By providing feedback, the supplier learns about consumer preferences, which results in more accurate products, while the number of alternatives searched is being decreased.

Both case studies confirm that presenting large choice sets by attribute instead of by alternative decreases search costs for consumers. However, the case studies show deviations in the way it affects search costs. The Last.fm case study illustrates that it leads to a decrease of the number of alternatives searched by consumers in order to accurately find products of their choice. Last.fm does it by using tags and related tags, which is a very accurate way to find or discover new products. The Pandora case study explains that presenting attribute information to consumers about why a particular track is playing decreases search costs, because consumers learn their musical preferences that way, which is a possible substitute for search agents.

The Last.fm case study proofs that default versions decrease the number of alternatives searched. Last.fm offers a default version for recommendations, which can be customized. This is in accordance with Dellaert and Stremersch (2004) who argue that companies should offer a default version that consumers can use as a starting point for mass customization to minimize complexity.

Both case studies confirm that consumer expertise leads to better product accuracy. Pandora uses attributes to explain why tracks are being played. This consumer expertise lets consumers find or discover new products more accurately. The Last.fm case study shows that consumer groups make it easier for members of these groups to find products, or more precisely modules of products, more accurately. At the same time it decreases the number of alternatives searched.

Both case studies confirm that recommendations decrease search costs. Direct and automatic recommendations decrease the number of alternatives searched, and both types of recommendations lead to a better product accuracy. There is a difference at the case studies in how automatic and direct recommendations are presented. At Last.fm, recommendations are available on the personal page of the consumer, while Pandora sends e-mail messages to recommend new or changed digital products.

Customization

Case

Influence on search costs

Consumer involvement

Both

Number of alternatives searched, product accuracy

Last.fm

Pandora

Type of modularity employed

Both

Last.fm

Pandora

Attribute vs. alternative

Both

Last.fm

Product accuracy, number of alternatives searched

Pandora

Search agents

Default version

Both

Last.fm

Number of alternatives searched

Pandora

Consumer expertise

Both

Number of alternatives searched, product accuracy

Last.fm

Pandora

Automatic recommendation

Both

Number of alternatives searched, product accuracy

Last.fm

Pandora

Direct recommendation

Both

Number of alternatives searched, product accuracy

Last.fm

Pandora

Trust building

Both

Last.fm

Pandora

Collaborative co-design

Both

Product accuracy

Last.fm

Pandora

Table 5.5: Influence of customization on decreasing search costs, matched from cases.

Both cases proof that collaborative co-design reduces search costs. At Last.fm it reduces search costs because it leads to consumers having a personal profile. Once consumers do have a personal profile, it becomes possible to make better recommendations based on musical neighbours. The Pandora case study confirms that collaborative co-design is improving product accuracy due to consumers who are collaboratively giving feedback to the system, which leads to changes in the products. It has some similarities with the consumer involvement factor, because consumers’ requests are designed into the finished product. The difference is that it is not done uniquely, but collaboratively.

The case studies can not proof that trust and the type of modularity employed have influence on search costs. It can be argued that trust decreases search costs, by reducing uncertainties about the behaviour of the provider. The Last.fm case study shows however that it does not reduce uncertainties about the behaviour of the provider, but it results in which products are more popular compared to other products.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: