1
The optimal pricing model for outsourcing CRM
in the Dutch E-Commerce market
2
The optimal pricing model for outsourcing CRM
in the Dutch E-Commerce market
Frans Ruben KramerUniversity of Groningen Faculty of Economics and Business
Master thesis MSc Marketing Intelligence & MSc Marketing Management
08‐01‐2016 Westerhavenstraat 6 9718AL, Groningen Tel: +316 23 62 30 00 E‐mail: f.r.kramer.1@student.rug.nl Student number: s1867830 Supervisors University of Groningen
3
Abstract
The shopping behavior of consumers has shifted from physical stores towards online channels, which has disruptive consequences for the way that companies sell their products and maintain their client helpdesk. The purpose of this study is to find the most valued pricing model for outsourcing customer relationship management (CRM) activities. This study adds knowledge to the research field of CRM‐outsourcing from the value based costs perspective, especially in the E‐Commerce market. With the application of a choice based conjoint analysis, the results show that the most valued pricing model depends on the total number of contacts. With a limited amount of contact moments between the client and the company, a fixed pricing model is preferred, whereas a variable pricing model is preferred when there are more contact moments between the client and the company. Furthermore, higher amounts of add‐on sales possibilities appears to lower the utility for all types of pricing models. No moderating effect for add‐on sales was found.
4
Table of contents
1. Introduction ... 5 2. Theoretical framework ... 9 2.1 Value‐based pricing ... 9 2.2 Pricing models ... 92.2.1 Fixed pricing model... 10
2.2.2 Variable pricing models ... 12
2.2.3 Mixed pricing model ... 14
2.3 Moderators ... 15
2.3.1 Company size / number of contacts ... 15
2.3.2 Cross‐ and upsell possibilities ... 15
2.4 Conceptual model ... 17
3. Methodology ... 18
3.1 Method ... 18
3.2 Model specification ... 18
3.3 Procedure ... 19
3.4 Attributes and levels ... 20
3.5 Choice design ... 23
3.6 Sample ... 23
4. Results ... 25
4.1 Descriptive statistics ... 25
4.1.1 Exclusion of no‐choice answers ... 25
4.1.2 Sample description ... 25
4.2 Model selection and validation ... 26
4.2.1 Main effects ... 26
4.2.2 Interaction effects ... 28
4.3 Model estimation ... 30
4.4 Model fit ... 32
4.5 Interpretation... 33
4.6 Comparing different pricing models ... 35
5. Discussion and conclusion ... 38
6. Limitations and recommendations ... 41
7. References ... 42
5
1. Introduction
The E‐Commerce business is growing at a high rate, with thousands of new competitors entering the market every year with their web shop. Last summer, for the first time, there were more web shops than physical shops in the Netherlands (van der Ploeg, 2015), indicating its relevance and importance. The expectation is that this trend will continue and the amount of web shops will only increase more in both absolute and relative terms (Van der Ploeg, 2015). This is confirmed by Verhoef et al., (2015). They state that the shopping behavior of consumers has shifted from physical stores towards online channels. Online channels became very dominant and the rise of online players is disruptive for the market. Consumers behave differently due to all these changes in consuming behavior.
These customers have an increasing demand of services provided by a company and require to get their solution instantly (Tax & Brown, 2012; Wilson et al., 2012). If the solutions provided by the client helpdesk are perceived as slow or insufficient, customer satisfaction will drop dramatically (Harris, 2013; Hurley, 2015). Furthermore, the odds of churning (leaving a company to acquire the service of a rival company) increase when the provided service is inadequate (Ascarza et al., 2015; Dror et al., 2012). These trends show the impact and the importance of a well‐functioning client helpdesk.
Since E‐Commerce companies are often relatively young, they perceive their customer service as a necessity and not as an opportunity to enhance their business (Wright, 2014). The customer service in the E‐Commerce market is still in its infancy compared to for example the ‘traditional’ markets of Energy and Telecom (Papaioannou et al., 2014). These traditional businesses usually have highly developed customer relationship management (CRM) in which the client helpdesk not only comes up with a solution for their customer, but also generates additional income by actively chasing cross and upsell possibilities (Chen & Popovich, 2003; Jasmand et al., 2012; Richards & Jones, 2008).
6 Multiple studies have been conducted on value‐based pricing (e.g., Gneiser, 2010; Hollensen, 2015). This will be an important guide to investigate the different pricing models. Value‐based pricing will be extensively discussed in the ‘theoretical background’ chapter. Currently, the CRM market has reached a point in which competition is mainly based on price. The valuation of the suppliers service is predominantly based on costs and the price of the competition. In short, a calculation is made of the predicted costs and an additional margin is added to determine the final price, while controlling for the price of competitors. Right now, the question is what price an outsourcing party is willing to pay for outsourcing its CRM. More specifically, what is the perceived value of their client contact? With this research, a solid base should be created which can be used to determine the value of the services of client helpdesks in the E‐Commerce market.
Recently, some research by Hasija et al., (2008) was conducted to investigate the most profitable pricing model for the outsourcing party. However, this research focused on the most preferred price models with the optimal coordination of capacity density as a goal. This was purely based on costs and not so much on other drivers of perceived value in pricing attributes for the outsourcer. No research has been conducted the other way around. No academic papers have been published to investigate what the most optimal pricing model for an outsourcer could be from the value‐based pricing point of view.
This research is relevant in multiple ways. First, it has practical relevance for suppliers in the CRM market to determine the optimal perceived value of their service by the (potential) outsourcers. By identifying which type of pricing models are preferred, the suppliers can anticipate on the needs of their (potential) customers and they will have enhanced opportunities compared to the current situation to acquire new clients. Furthermore, understanding the drivers of the pricing preferences should provide a supplier with more detailed knowledge about pricing decisions of the outsourcing party. In this way, it is possible for the suppliers to create a more personalized offer based on these drivers of price which should lead to increased likelihood of actually acquiring (new) customers.
7 pricing solution with regard to capacity density. They found no universally preferred pricing solution and suggested that other models and contract types in another context with different goals should be investigated as well.
Ren & Zhou (2008) investigated two important aspects in the outsourcing decisions. First they studied the optimal capacity density. Their second main topic was the amount of effort a supplier had to exert to achieve the agreed service quality. Together, these two topics lead to a most preferred model in which the coordination between both outsourcer and supplier should be optimal. The link between value‐based pricing and the optimal pricing model for the CRM outsourcing decision in the E‐Commerce market has not been researched yet. This way, a gap in the literature is addressed.
The aim of this research will be to investigate which pricing strategies and models are optimal from a value‐based perspective. More specifically, by investigating the perceived value of multiple pricing models by the means of a conjoint analysis, the most preferred pricing model should be found.
Ultimately, the goal of this paper is to answer the following research question: What is the
most valued pricing model for outsourcing CRM activities in the Dutch E-Commerce market?
To answer this main research question in a satisfying way, it is necessary to subdivide it into multiple sub questions. In the end, the answers of these sub questions should lead to an answer on the research question. These sub‐questions are based on the literature review in the next chapter and qualitative interviews with relevant stakeholders. The sub questions will be translated into hypotheses in the chapter ‘Theoretical background’.
The next sub questions were derived:
‐ What is the perceived value of fixed‐ and variable price models?
‐ What is the influence of the total number of contacts between a customer and the outsourcing company on the perceived value?
‐ Is the possibility to cross‐ and up sell an important driver of perceived value?
8 contact moments (less than three) between the client and the company, a fixed pricing model is preferred, whereas a variable pricing model is preferred when there are more than three contact moments between the client and the company. Furthermore, higher amounts of add‐on sales possibilities appears to lower the utility for all types of pricing models. No moderating effect for add‐on sales was found.
9
2. Theoretical framework
In this chapter, relevant insights from the literature are presented. First, value‐based pricing will be analyzed. Next, the link between value‐based pricing and different pricing models for outsourcing is reviewed. Multiple pricing models are compared to each other and hypotheses are formed. Finally, a conceptual model is presented. This model is based on the literature review, as well as the qualitative interviews with the stakeholders of this research. 2.1 Value-based pricing
Price is one of the most important Ps of the marketing mix. It has a high impact on the financial result of a company in both absolute and relative terms compared to the other instruments of the marketing mix (Hinterhuber, 2004). Pricing strategies can be divided into three main groups: cost‐based pricing, competition‐based pricing and value‐based pricing (Hinterhuber, 2008). Historically, cost‐based pricing has been the standard (Noble & Gruca, 1999). More recently, value‐based pricing is recognized as superior compared to cost‐based and competition based. For example, Ingenbleek et al. (2013) found a correlation between pricing based on customer preferences and the success of an introduction of a new product or service. This correlation was not found with the cost‐based or competition‐based strategy. Furthermore, Liozu & Hinterhuber (2013) found a positive relationship between value‐based pricing and firm performance, whereas no such relationship was found for cost‐ based or competition‐based pricing.
In the rest of this paper, different pricing models will be reviewed and analyzed from the value‐based pricing point of view. The definition of value‐based pricing used in this paper is based on Hollensen (2015). He defined value‐based pricing as: “When prices are based on the value of a product as perceived from the customer’s perspective. The perceived value determines the customer’s willingness to pay and thus the maximum price a company can charge for its product”.
2.2 Pricing models
10 models for outsourcing CRM were chosen. Both the literature review and interviews are used to derive the hypotheses for the perceived value of these different pricing models. In these models, a main distinction is made between fixed price and variable price models. Both types of price models (fixed and variable) will be analyzed in detail and hypotheses are formed regarding the perceived value of a specific pricing model. Especially the pricing models with a variable component will be extensively discussed since a lot of interesting pricing models are available for the outsourcer.
2.2.1 Fixed pricing model
In the history of pricing models, a fixed total price for providing a certain service is one of the most used models (Allon & Federgruen, 2005; Taylor, 2012). In this paper, fixed price is defined as: “a type of contract providing for a price that is not subject to adjustment on the basis of the contractor’s cost experience in performing the contract” (Nash et al., 1998). A fixed price contract has its advantages and disadvantages from the value‐based point of view. An advantage is that fixed price contracts places the maximum risk and full responsibility on the supplier for all costs and resulting profit or loss. They provide maximum incentive for the supplier to control costs and perform effectively and efficiently and impose a minimum administrative burden upon the contracting parties unless changes are issued or unforeseen events occur during performance (Nash et al., 1998).
From the value‐based perspective, this could be a highly relevant model for the outsourcer since costs are one of the major drivers of value perception (Christopher & Gattorna, 2005).
11
Costs
Quantity
Fixed price model
Costs
Quantity
Variable price model
Figure 1 Graphical representation of a fixed pricing model Figure 2 Graphical representation of a variable pricing model However, this type of contract also brings some significant downsides. First of all, an accurate calculation of the total number of contacts with a customer is required to predict the exact costs of a CRM operation (Taylor, 2012). Workforce management and planning are some of the critical factors for a successful and profitable CRM department (Baron & Milner, 2009). If an outsourcing company is not able to make a reliable prediction of the total number of contacts, suppliers are hesitant to get into a fixed price deal (Taylor, 2012). Security regarding a reliable prediction is a major driver of value for an outsourcer and if they perceive security as (partially) absent, they will not concern an inflexible model, like the fixed price model, as a serious option (Ulaga & Eggert, 2006). A second disadvantage with a fixed price model is the absence of a variable incentive for the supplier. Why would a supplier put extra effort in its operations if there is no potential gain for them? Obviously, there should be a good fit between both parties before an agreement could be one of mutual gain. Since this type of model cannot guarantee such a mutual gain, this model is expected to be less interesting for the outsourcer. Therefore, the next hypothesis is formed:
H1: The use of a fixed price revenue model will be perceived as less valuable
12
2.2.2 Variable pricing models
The focus in this paper will be on the most relevant combinations for the outsourcing of CRM (if possible, related to the E‐Commerce market). For the variable price, a very basic definition is used: ‘”the price of an organization that varies with the amount of work performed’” (Nash et al., 1998). However, ‘the amount of work performed’ is vaguely noted since it can be viewed from different perspectives. These perspectives will be elaborated later on in this chapter. The variable price model also has its advantages and disadvantages. One of the advantages is that the outsourcer and the supplier can make an agreement that is as personalized as possible for both parties from both the cost and the revenue point of view, which is often a highly valued characteristic for the outsourcer (Krasnikov et al., 2009). A disadvantage of a variable price agreement could be that during the process, the costs are rising higher than expected, which could lead to exceeded budgets for both (Erceg et al., 2000). Obviously, this would lead to less perceived value for the outsourcer. In figure 2, a simplistic graphical representation of a variable price model is shown.
Overall, both the literature (e.g., Aksin, De Véricourt & Karaesmen, 2008; Ren & Zhou, 2008) and interviews with relevant stakeholders show that variable pricing models are valued most by outsourcers. First, the flexibility compared to fixed and mixed (a combination of a fixed and variable pricing model) pricing models is an important driver of value for the outsourcer (Ren & Zhou, 2008). Second, the outsourcer pays per unit or solution which he thinks is relevant. Therefore, he will only pay for services that he actually uses and which are of significant importance. Lastly, the incentive for the supplier is the highest in this type of model. Since the supplier only gets paid per unit, it is in the best interest of both parties that the supplier puts in the maximum amount of effort. Therefore, the following hypothesis is derived.
H2a: The use of a variable pricing model will be perceived as most valuable pricing
model
2.2.2.1 Price per contact (plus solution) (ppc+s)
13 first‐time‐fix has become more and more important. Therefore, the focus on a solution in this type of contract could be of high value for outsourcers. In short, the outsourcer pays the supplier a specific amount for each contact that is served and resolved (Ren & Zhou, 2008). These authors also find that the ‘”ppc+s contract can not only coordinate staffing level, but also motivate the call center to exert effort to improve service quality”’. Thus an incentive is integrated in this model, which is beneficial for both parties if the goals stated in the contract are met.
A second important value for the outsourcer of this model is the staffing of the supplier. The supplier gets its revenue based on the total number of contacts resolved so the supplier’s goal will be to avoid any missed contact points. Therefore, the staffing should be optimal to gain the maximum income for the supplier, which is what the outsourcer pursues.
In short, this model offers a more suitable solution with regard to the strategic goals of the outsourcer, thus adding more value for them. Since a staffing decision and a variable incentive is integrated, it is expected that the perceived value for such a model is higher than for example other traditional variable price models, like price per contact and price per client. Therefore, the next hypothesis is:
H2c: The use of a ppc+s model will be perceived as more valuable compared
to other variable pricing models 2.2.2.2 Bonus tariff
14
Costs
Quantity
Variable price model with
bonus tariff
Costs
Quantity
Mixed price model
Figure 3 Variable pricing model with a bonus tariff
H2b: The use of a pricing model with a bonus tariff model will be perceived as more
valuable compared to other pricing models
2.2.3 Mixed pricing model
A mixed model is a combination of a fixed and a variable component. With this model there are numerous options available regarding the ratio between fixed and variable. The focus of this paragraph is not on the best possible ratio between these two but rather on the mix of values that the outsourcer can derive from this model. It brings together both the advantages, but also some of the disadvantages of the fixed pricing model and the variable pricing model (Feng & Lu, 2012). The fixed component provides the outsourcer with certainty regarding a large part of his costs while the variable component (e.g., price per contact, price per client or price per solution) includes an incentive for extra effort of the supplier (Aksin et al., 2008). Furthermore, this model is easy to implement which makes it popular in practice (Ren & Zhou, 2008). Figure 4 shows a representation of a mixed price model.
In short, this model appears to offer extra benefits when compared to the fixed price model. It is expected that this model overall is more valued than the fixed price model. Therefore, the next hypothesis is: H3: The use of a mixed model will be perceived as more valuable
compared to fixed pricing models but less valuable compared to variable pricing models
15 2.3 Moderators
2.3.1 Company size / number of contacts
Another relevant effect that determines whether or not to outsource, is the size of the company. Multiple studies have shown that company size is one of the most important drivers of the outsourcing decision (Embleton & Wright, 1998; Hättönen & Eriksson, 2009). A positive correlation is the most common between these two variables: the larger the company, the higher the likelihood of outsourcing (Zhu, Hsu & Lillie, 2001).
Multiple definitions of company size are present in the literature. One of the many definitions of company size is based on the annual revenue and employees (Arlow & Gannon, 1982). Yet, according to Aldor‐Noiman et al. (2009), there appears to be a strong correlation between the size of the outsourcing company and the total number of contacts. Therefore, company size is defined as the total number of contacts with a customer in a certain time period. More specifically: ‘the total numbers of contacts of a company with its
customers in a one year timeframe’.
According to multiple stakeholders and studies performed by Allon & Federgruen (2005) and Taylor (2012), the total number of contacts with a customer will influence the most preferred contract type. In addition, Aksin, Armony & Mehotra (2007) found that the total number of contacts had a moderating effect on the type of contract chosen. They argued that outsourcers who had a larger amount of contacts with their customers, preferred a more personalized contract. This implies that companies with a higher amount of contacts, prefer a variable pricing model over a mixed‐ or fixed pricing model. Hence, it is expected that the number of contacts will have a moderating effect on the perceived value of different pricing models.
Therefore, the following hypothesis is formed:
H4: The total number of contacts of a company with its customers moderates the
effect on the perceived value of different pricing models such that companies with a higher total number of contacts value variable pricing models over a mixed- or fixed pricing model
2.3.2 Cross- and upsell possibilities
16 company to actively cross‐ and upsell. However, cross‐ and upsell opportunities can have both negative and positive effects. Cross‐ and upselling can lead to a significant increase of workload for agents (Gurvich et al., 2009). Furthermore, agents should possess the capabilities of recognizing an opportunity and actually selling an additional service to a customer. If the balance between time and costs is not right, a company should reconsider if they wish to cross‐ and upsell via their client helpdesk (Aksin & Harker, 1999). However, in this study, this factor will be controlled for, since the assumption can be made that employees are capable (because of the multiple trainings they received) of determining whether it is effective or not to inform a certain customer about the extra services a company is offering.
Another factor that should be taken into account is the occupancy rate (Byers & So, 2007). When the occupation in the call‐center is below a certain threshold, there should be more time available to actively chase for add‐on sales. This factor will be controlled for since the workforce management is constantly recalculating the optimal number of employees in the call‐center.
In conclusion, multiple authors state the importance of cross‐and upsell opportunities and abilities to decide whether or not to outsource (Aksin & Harker, 2009; Gurvich et al., 2009; Kotwal, 2004). Therefore, the next hypothesis is derived:
H5: Cross- and upsell possibilities increase the perceived value of pricing models
Furthermore, Gurvich et al., (2009) argued that the possibility of cross‐ and upselling also influenced the type of contract chosen between outsourcer and supplier. More specifically, if a possibility to cross‐ and upsell was present, the outsourcer preferred a model with a flexible component and an additional bonus over inflexible contracts. This implies that there is a moderating effect between the possibility to cross‐ and upsell and the most valued contract by an outsourcer. It is expected that the presence of cross‐ and upsell possibilities increases the likelihood of choosing variable pricing models over fixed and mixed pricing models. Therefore, in addition to H5, the following hypothesis is derived:
H6: The preference for a variable pricing model compared to mixed or fixed pricing
17 2.4 Conceptual model
The final conceptual model resulting from the hypotheses is presented in figure 5.
Figure 5 Conceptual model
H1‐H3 H4
H6 H
18
3. Methodology
3.1 Method
Based on this study, the preferences towards differences in pricing models of Dutch E‐ Commerce businesses should be found. Participants are asked to choose their most valued and relevant pricing model out of several alternatives. In order to observe their most valued pricing model, a choice‐based conjoint analysis is selected. Therefore, the dependent variable is a choice made by the respondents.
In this research, price is the main factor which is analyzed. According to Eggers & Sattler (2011), when price sensitivity or willingness‐to‐pay (WTP) is measured, a ‘no‐choice’ option should be part of the choice set since it could provide valuable additional information to the outsourcers. Furthermore, if the ‘no‐choice’ option is included in this research, the realism increases (Eggers & Sattler, 2011). Therefore, the ‘no‐choice’ is included in the choice sets.
3.2 Model specification
In this research, alternatives in the choice sets are chosen based on the overall utilities of the attributes (Elrod et al., 1992). This can be translated into the following formula, in which participant has a utility towards pricing model :
= + (1)
In which:
: Represents the systematic utility component (rational utility) of participant for pricing model
: Stochastic utility component (error term)
Furthermore, it is assumed that each pricing model is a composition of multiple pricing attributes (Elrod et al., 1992). The second assumption is that participants will at first attach part‐worth utility towards all attributes. When the part‐worth utilities are found, tests will be conducted on the pricing attributes to check whether they can be translated into linear variables. The sum of all these part‐worth utilities represents the systematic utility of participant for pricing model which can be translated into the next formula:
= ∑ (2)
19 : Number of attributes
: Dummy which indicates a specific attribute level for product : Part‐worth utility of consumer for attribute
Next, the assumption can be made that participants will choose the alternative which is the most preferred. This can be translated into the next conditions:
> , > and > ,
In which A, B, C and D represent the three pricing models and the ‘no‐choice option’ in the choice sets. Since the dependent variable can exhibit more than two states (i.e., fixed, variable, mixed or no‐choice), the multinomial Logit (MNL) model can be used to predict the probability of choosing alternative from choice set (Louviere & Islam, 2008). With the next formula, a link is made between the utility and the probability of choosing a certain pricing model:
( | ) =∑ ( )
( ) (3)
In which:
: Number of alternatives in the choice set (= 4) : Specific choice‐set
Together, these formulas should ultimately translate the choices of the participants to absolute numbers which can be used to estimate the fit of the conceptual model. The pricing model with the highest part‐worth utility will be preferred most compared to the other alternatives.
3.3 Procedure
A survey is created in which respondents first answer general questions about themselves and their company. In this stage, respondents also elaborate on the size of their company by answering questions regarding the total revenue, the total number of contacts and the average number of contacts per customer. By comparing the results of these answers to the outcomes of the conjoint analysis, H4 regarding the moderator ‘size’ (i.e. the total number of
20 The moderator of total number of contacts will be measured by first translating the answers of the respondents towards scale points. Table 1 presents the possible answers of the respondents and the translation of these answers towards scale points.
Table 1 Scale points of total number of contacts
Total number of contacts (per year) Scale point
Between 0‐1 time per year 1 Between 1‐3 times per year 2 Between 3‐5 times per year 3 Between 5‐10 times per year 4 More than 10 times per year 5
Next, the scale point of the outcome is multiplied with the fixed price, the variable price and the add‐on sales.
3.4 Attributes and levels
To measure whether the variables and moderators included in the conceptual model have explanatory value regarding the optimal pricing model, all these variables need to be included in the model. The main effects that will be measured are coded as: fixed price ( ), variable price ( ), variable unit in the model (price per contact, solution or client) ( ), the add‐on Sales ( ), the bonus tariff ( ) and the none‐option ( ).
Next, to measure the moderating effects on the preferred pricing models, multiple interactions have to be created. The pricing model is separated into multiple pricing attributes (fixed, variable and add‐on prices). Therefore, two extra variables have to be created for the effect of add‐on sales and three extra variables for the effect of total number of contacts. These are the interactions between fixed price and add‐on sales ( ∗ ), variable price and add‐on sales ( ∗ ), fixed price and total number of contacts ( ∗ ), variable price and total number of contacts ( ∗ ) and finally, add‐on sales and the total number of contacts ( ∗ ). If all variables were combined into one model, this would lead to the next formula:
= + + + + + + ( ∗ ) + ( ∗ ) +
21 In which:
: Rational utility for pricing model : The effect of fixed price on
: The effect of variable price on : The effect of the variable unit on : The effect of add‐on sales on : The effect of a bonus tariff on : The effect of the none‐option on
: The interaction effect between fixed price and add‐on sales on : The interaction effect between variable price and add‐on sales on : The interaction effect between fixed price and total nr. of contacts on
: The interaction effect between variable price and total nr. of contacts on : The interaction effect between add‐on sales and total nr. of contacts on
The respondents are presented with ten choice sets in which they have to state their most valued pricing model. An example of a choice set is shown in table 2.
Table 2, Example of a choice set
Alternative 1 Alternative 2 Alternative 3 Alternative 4
Fixed Price €100.000,‐ per year €200.000,‐ per year €0,‐ With these alternatives I would not outsource Variable price €5,‐ per client €5,‐ per
solution €10,‐ per contact Bonus 20% increase of the fixed price after 10.000 clients 20% increase of the variable price after 10.000 clients 20% decrease of the variable price after 10.000 clients
Add-on sales No possibility €20.000,‐ per year
€10.000,‐ per year
22 The choice sets consists of five different attributes which are:
‐ Fixed price ‐ Variable price ‐ Variable pricing unit ‐ Add‐on sales
‐ Bonus tariff
Fixed price
The first attribute is ‘fixed price’. This attribute consists of three different levels, which are €0,‐, €100.000,‐ and €200.000,‐. As mentioned in the previous section, only realistic combinations of the price attributes are created. Furthermore, the fixed price is divided by 1.000 in the output for simplifying the interpretation. This will be accounted for in the results section.
Variable price
The second attribute in the choice sets consists of a variable price component. There are four levels present; €0,‐ , €5,‐ , €10,‐ and €15,‐. The only combinations that are possible in the choice sets are those with a fixed price of €0,‐ combined with a variable price of €10,‐ or €15,‐, a fixed price of €100.000,‐ with a variable price of €5,‐ or €10,‐ and finally a fixed price of €200.000,‐ with a variable price of €0,‐ or €5,‐. This way, the non‐realistic combinations are excluded from the choice sets.
Variable unit
The variable units that will be investigated are the price per client, per contact and per solution.
Add-on sales
23
Bonus tariff
The bonus/malus tariff will be measured by four different levels of which two are focused on the fixed price and the other two focus on the variable price. The levels present in the choice sets are: 20% increase of fixed price, 20% decrease of fixed price, 20% increase of variable price and 20% decrease of variable price.
3.5 Choice design
This conjoint analysis is a fractional factorial design, which is a subset of the full factorial. Hauser and Toubia (2005) state that before a fractional factorial design can be considered as efficient, it should be balanced (all levels should be displayed an equal number of times) and orthogonal (all level combinations should appear an equal number of times). A balanced and orthogonal design is essential to avert a correlation between attributes. An orthogonal and balanced list of 180 different stimuli was created.
Ten different choice sets are presented to the respondents. Each choice set consists of four alternatives including a ‘no‐choice’ option. The three alternatives consists out of five different attributes. In the choice sets, the attributes of ‘fixed price, ‘variable price’, ‘variable pricing unit’ and ‘bonus tariff’ are brought together into one attribute. This is done since impossible combinations have to be left out of the survey. For example, a combination with a fixed price of €200.000,‐, €0,‐ per contact, a possibility of €10.000,‐ in add‐on sales and 20% increase in variable price after 10.000 units is impossible since a variable price of €0,‐ cannot increase in percentage terms. Since it is expected that the total price will be perceived as the most important attribute, some adjustments have to be made to assure that realistic alternatives are presented. For example, a €0,‐ fixed price combined with a €5,‐ price per customer and other similar alternatives are excluded from the choice sets.
Originally, the total number of possible combinations would be 432 (3 X 4 X 3 X 4 X 3 levels). After the four previously mentioned attributes have been converted into one, the total number of possible combinations is brought back to 180 (60 x 3 levels). A subset of these 180 possible combinations can be found in Table 17 in the appendix.
3.6 Sample
24 on the annual revenue in this market is created yearly by Twinkle100 on their website. Employees from multiple departments of these top 200 businesses will be approached to answer the survey.
Table 3 Elements of the method
Elements Explanation
Population Dutch E‐Commerce businesses
Sampling method Top 200 E‐Commerce shops based on www.twinkle.nl
Sample size 52 respondents from multiple businesses. Of these 52 respondents,
12 were deleted. The final sample size is therefore 40.
Representativeness 40 respondents are sufficient from a statistical point of view.
Data collection Digital surveys via My Preference Lab in the Dutch B2B E‐
Commerce market.
Analyses + tools used ‐ SPSS
‐ Latent Gold ‐ Excel
25
4. Results
4.1 Descriptive statistics
4.1.1 Exclusion of no-choice answers
Since the ‘no‐choice’ option is included to increase realism, it is possible for respondents to only pick this alternative. 12 respondents only chose the ‘no‐choice’ option. This could have several reasons. One of them is that the respondent has no direct feeling with the subject of research and just answers the easy way. If the required time of responding is analyzed, it seems that these respondents need significantly less time to answer the whole survey than needed. Second, the purpose of this research is to investigate the most valued pricing models. As this is not the intended target group, the decision was made to delete these respondents from the data. The analyses will be conducted on the 40 remaining respondents. Since there are 400 choice sets (40 respondents times 10 choice sets), this should be sufficient to obtain reliable results from a statistical point of view.
4.1.2 Sample description
Of the 40 remaining respondents, 29 were male and 11 were female with an average age of 36. Most of the respondents worked at their company for five years or more and in the marketing department or customer service (73%). A small group was active in the business intelligence field (9.6%). The rest of the respondents worked in another department. Table 4 presents the hierarchical level of the respondents in their company. This table shows that about half of the respondents work in the management (42.3%) or in the board (7.7%) of their company. 34.6% of the respondents can be considered as a specialist in their department.
Currently, 13 respondents stated that their company has outsourced or is currently (partially) outsourcing their client helpdesk. In addition, 18 respondents answered that they would probably outsource their client helpdesk in the future. In addition, figure 6 shows the most common reasons to outsource their client helpdesk. The most chosen option was cost saving, which 31 respondents stated
Table 4 Hierarchical level of respondents
Hierarchical level Percentage
Board 7.7%
Management 42.3%
Specialist 34.6%
26 as the most important reason. Innovative options (e.g. introducing a new way to maintain the CRM) follows with 22 respondents. Extra revenues seems to be less important.
Figure 6 Reasons for outsourcing CRM
4.2 Model selection and validation
4.2.1 Main effects
First, a model with the main effects is analyzed to test whether all these effects are significant predictors for the most valued pricing models. Furthermore, this model will be used for benchmarking purposes when the interaction effects are inserted. The next formula presents the ‘main effects‐only’:
= + + + + + (5)
Table 18 in the appendix shows the outcomes of the conjoint analysis for this model
together with a few validation diagnostics. This model assumed nominal (i.e., part‐worth) main effects of the pricing variables fixed price, variable price and add‐on sales. To check whether these variables could be simplified into linear effects, the Log‐likelihood of both models (linear vs. part‐worth) is calculated and compared to each other.
Table 5 Linearity of fixed price
LL(B) n-par hit rate R2 R2adj χ2 Threshold-value
Fixed price (part worth) ‐527.14 13 39.0% 0.0494 0.0259 1.63 3.84 Fixed price (linear) ‐527.95 12 38.0% 0.0479 0.0263
Table 5 presents the findings of the linearity test for the fixed price variable. The most
relevant validity check is conducted by comparing the χ2 to the outcomes from the χ2 distribution table. It can be calculated by: χ2 = −2((LL0) − (LLβ)). The (LL0) in this case is the (LLβ) of the part worth model. The value of 1.63 does not exceed the threshold‐value of 3.84 (1 degree of freedom, α=0.05), which means that there is not a significant difference
0 10 20 30 40
Cost saving Extra revenues Expertise
personell Benchmarking with own CRM Innovative options None TOTAL
27 in fit between both models and therefore no significant loss in information. The linear model has a lower R2 and a lower hit rate (which makes sense, since less parameters are present in this model). The R2adj is a more reliable indicator since it accounts for the number of
parameters (Malhotra, 2010). The higher value indicates that the linear model performs better. Therefore, the fixed price is assumed to be linear, for ease of interpretation.
Table 6 Linearity of variable price
LL(B) n-par hit rate R2 R2adj χ2 Threshold-value
Variable price (part worth) ‐527.14 13 39.0% 0.0494 0.0259 3.10 5.99 Variable price (linear) ‐528.69 11 39.0% 0.0466 0.0267
The same process was conducted for the variable price and the add‐on sales. Table 6 and
table 7 presents the findings for these linearity tests. The results for the variable price are
similar to the results of the fixed price. The R2adj scores higher in the linear model. In
addition, the calculated χ2 of 3.10 does not exceed the threshold‐value of 5.99. Therefore, the variable price is assumed to be linear as well.
Table 7 Linearity of add-on sales
LL(B) n-par hit rate R2 R2adj χ2 Threshold-value
Add-on sales (part worth) ‐527.14 13 39.0% 0.0494 0.0259 0.03 3.84 Add-on sales(linear) ‐527.15 12 39.0% 0.0494 0.0277
Next, the linear model of the add‐on sales performs almost equally well on the validation diagnostics for the add‐on sales. Again, the R2adj scores higher in the linear model.
Furthermore, the χ2 value of 0.03 does not exceed the threshold value of 3.84. This implies that there is not a significant difference in model fit between both models. Therefore, the add‐on sales is assumed to be linear as well.
Table 8 Linearity of all pricing attributes
LL(B) n-par hit rate R2 R2adj χ2 Threshold-value
All (part worth) ‐527.14 13 39.0% 0.0494 0.0259 7.19 9.48
28 Finally, table 8 shows the results when all pricing attributes mentioned above are translated into linear variables. This final validity check shows that it is safe to assume that all these variables are linear and that there is no significant difference. The χ2 value of 7,19 does not exceed the threshold value of 9,48. Since it is easier to interpret a model with fewer parameters and a continuous trend is present (Malhotra, 2010) these three variables will be translated into linear variables. This model will be used as a starting point to measure the interaction effects and for benchmark purposes. The output is shown in table 9.
Table 9 Model results of benchmark model (main effects)
Main effects
β-values Wald p-value
Fixed price ‐0.005 8.062 0.0045***
Variable price ‐0.072 6.677 0.0098***
Add‐on sales ‐0.009 1.686 0.19
Variable Unit 9.247 0.0098***
Price per contact ‐0.142 Price per solution 0.278 Price per client ‐0.135
Bonus tariff 16.882 0.00075*** Fixed 20% discount 0.279 Fixed 20% increase ‐0.282 Variable 20% discount 0.271 Variable 20% increase ‐0.267 None‐option ‐1.483 15.296 0.000092*** * p < 0.10, ** p < 0.05 and *** p < 0.01 4.2.2 Interaction effects
In this section, the moderators will be tested separately. Since multiple interaction effects might influence each other, this is necessary to check whether certain (insignificant) effects should be removed from the model. The significance and the validation diagnostics are analyzed of both models and a comparison will be made between them.
The variables and the interaction effects mentioned in the methodology chapter, will be inserted into two separate models. This leads to the following formulas which represents the rational utility for selecting a price model as most valued ( ). Formula 1 represents the effect of the add‐on sales interactions and formula 2 the effect of the total number of contacts.
29 fixed or the variable price (e.g. 20% discount after 10.000 units on the fixed price), it is likely that there could be an interaction effect between these variables. Therefore, a third formula is created in which this interaction effect is included.
1: = + + + + + + ( ∗ ) + ( ∗ ) 2: = + + + + + + ( ∗ ) + ( ∗ ) + ( ∗ ) 3: = + + + + + + ( ∗ ) + ( ∗ )
Table 19 in the Appendix represents the outcomes with the parameter estimates of the
three models with the add‐on, number of contacts and bonus tariff interactions included. Note that the interaction effects are measured as linear variables. If the interaction effects were measured as part‐worth variables, the validity diagnostics were significantly better than its linear counterparts. However, the total number of parameters would increase dramatically. For example, the total number of parameters with part‐worth interactions between the total number of contacts and fixed price (14), variable price (19) and add‐on sales (14) would increase with a total of 47. With this amount of parameters, there would be no degrees of freedom anymore (i.e., more parameters than observations present). A negative number of degrees of freedom implies that the model has no independent ways in which the dynamic system can move (Malhotra, 2010). Since these results cannot be interpreted correctly, the interactions are measured as linear effects.
30 All interaction variables of the total number of contacts have a (marginally) significant effect (p<0.10). The presence of the interactions also influences the direct main effects of fixed price, variable price and add‐on sales.
Model 3 shows a significant effect for the variable unit and the none‐option. The other main effects are insignificant. All interaction effects between the bonus tariff and the fixed‐ or variable price appear to be insignificant (p>0.10). The insignificance in all of the interaction effects implies that there is not a moderating effect between the bonus tariff and the pricing attributes. A model in which all interactions were included provided the same results.
In short, the moderating effects of the add‐on sales and bonus tariff interactions have been proved to be insignificant. No significant effect can be found in any analysis for the inclusion of add‐on sales or bonus tariff as a moderator. Therefore, the interaction effects are deleted from the model and hypothesis 6 is rejected. The interaction effects between the total number of contacts and the pricing attributes is included in the final model for their significant effect.
4.3 Model estimation
As was explained in the methodology part already, some variables had to be transformed for simplifying the interpretation process. The first one is that the fixed price and the add‐on sales are divided by 1.000 in the dataset. For estimation purposes, the same procedure should be followed. The second change is the rewritten number of contacts. The number of contacts are translated towards scale points. In the scenarios below, this means that a total number of 3‐5 contacts per year a scale point of ‘3’ is created. Imagine a scenario in which two pricing models are compared to each other:
Table 10 Scenarios for estimation
Variables Pricing model 1 Pricing model 2
Fixed price Variable price €100.000,‐ €5,‐ €0,‐ €10,‐ Add-on sales €10.000,‐ €10.000,‐
Variable unit Price per contact Price per solution Bonus tariff Variable 20% discount Variable 20% discount Number of contacts (per year)
Overall utility
Between 3 and 5 times ‐0.0856
31 For estimation purposes, the final model can be written down with the following formula:
= + + + + + + ( ∗ ) + ( ∗ ) +
( ∗ ) (6)
The results are presented in table 11. Table 11 Final model results
Variables Final model
β-values Wald p-value
Fixed price ‐0.001 0187 0,66
Variable Price ‐0.130 9.3796 0,0022***
Add-on sales 0.046 6.6072 0,01**
Variable unit 8.5305 0,014**
Price per contact ‐0.135 Price per solution 0.280
Price per client ‐0.133
Bonus tariff 18.4105 0,00036*** Fixed 20% discount 0.3026 Fixed 20% increase ‐0.3018 Variable 20% discount 0.797 Variable 20% increase ‐0.2804 None-option ‐1.513 15.4604 0.000084***
Nr. of contacts * Fixed price ‐0.0014 5.0735 0,024** Nr. of contacts * Variable price 0.0209 3.4124 0,065* Nr. of contacts * Add-on sales ‐0.02 11.0333 0,0009***
* p < 0.10, ** p < 0.05 and *** p < 0.01
With the β‐values found and ‘pricing model 1’, the formula can be rewritten to:
= (−0,001 ∗ 100) + (−0,130 ∗ 5) + (0,046 ∗ 10) + (−0,135) + (0,280) + (−0.0014 ∗ 3 ∗
100) + (0.0209 ∗ 3 ∗ 5) + (−0.02 ∗ 3 ∗ 10) (7)
The same procedure can be followed for pricing model 2. In the bottom of Table 10, the overall utility of both pricing models can be found. The utilities of both scenarios, as presented in the bottom of table 10, can be inserted in the probability formula mentioned in the methodology part to obtain the probability of choosing scenario 1 over scenario 2. This leads to:
( | ) = ( ( . )
. ) ( . ) (8)
32 4.4 Model fit
The final model should be validated to check whether it performs better than the null model and the main effects‐only model. In table 12, some validation diagnostics are presented. First, the Likelihood Ratio Test has been conducted. This resulted in a χ2 of 71.06, which was more than the threshold value of 21.03 (12 degrees of freedom, critical α‐value of 0.05). Therefore, the estimated model predicts the choice decision significantly better than the null‐model. The final model has a hit rate of 39.0%, which means that the model can predict the right choice of respondents 39.0% of the time. In the null‐model, the hit rate is 25%, an increase of = 56.0%, hence the final model performs better than the null‐model. The hit rate of the final model can predict the right choice better in .
. 4.0% of the time.
In addition, the information criteria of BIC‐, AIC‐, AIC3‐ and CAIC‐values are compared to the same values of the ‘main effects‐only’ model. These values punish a model for increased complexity (number of parameters). A general rule of thumb is that a model with lower information criteria generally performs better. The information criteria of BIC, AIC and AIC3 are significantly lower in the final model than the ‘main effects‐only’ model, which implies that there is an improvement.
Finally, the R2 and the R2adj between the final model and the main effects are compared. The
higher values in the final model compared to the ‘main effects‐only’ model imply that this final model is a better predictor of the most valued pricing model. It shows that the final model explains 6.41% of the variance and 4.24% after the adjustment for the amount of parameters (R2adj).
Table 12 Validation diagnostics
Validation diagnostics Main effects Final model
33 4.5 Interpretation
Firstly, considering the effect of the fixed price on the most valued pricing model, it seems that this effect is insignificant. Therefore, this effect cannot be interpreted.
Next, an increase in variable price seems to lower the utility for a pricing model (p<0.01). This makes sense, since an increase in price should be perceived with less utility by respondents. An increase of €1,‐ in variable price will lead to a decrease in utility of ‐0.130. The next result is that an increase in add‐on sales possibilities increases the utility (p<0.05). An increase of €1.000,‐ in possibilities of add‐on sales will lead to an increase in utility of 0.046. This is in line with what was expected and hypothesized. Therefore, H5 is accepted.
Furthermore, the part‐worth utility for the variable unit shows that H2c is confirmed as well.
As expected, the price per solution (ppc+s) is more valued over the price per client and the price per contact. This is not surprising, since the ppc+s model offers direct value for the outsourcer (a payment only after a certain agreement is met). The presence of a ppc+s in a pricing model increases the utility with 0.2701, whereas the presence of a price per contact (‐0.1374) and a price per client (‐0.1327) actually decreases the utility (p<0.05).
The last direct main effect also follows the expected trend; the presence of a bonus (instead of a malus) increases the utility of a pricing model. This confirms the expectations and therefore H2b is accepted as well. The presence of the 20% discount on the fixed price
increases the utility with 0.3026 and the utility increases with 0.2797 if a model with 20% discount on the variable price is present (p<0.001). If the interaction effects are absent, both discounts show a significant, positive utility. In contrast, the utility decreases with respectively ‐0.3018 and ‐0.2804 if a 20% increase on fixed price or 20% increase on variable price is present.
34 For the variable price, the interaction effect is the other way around. The β‐value of 0.0209 implies that with an increased total number of contacts, the utility for a variable pricing model is perceived as more positive (p<0.10). Thus, the total number of contacts of a company with its customers moderates the effect on the perceived value of different pricing models such that companies with a higher total number of contacts value variable pricing models over a mixed‐ or fixed pricing model. This is in line with what was expected and therefore H4 is accepted. Next, the interaction between the add‐on sales and the total
number of contacts is significant (p<0.001) with a β‐value of ‐0.02. This is not in line with the expectations; the higher the total number of calls, the lower the utility towards a model with add‐on sales possibilities.
Finally, for interpretation purposes it is relevant to analyze the relative importance of the variables. These are shown in table 13. Since there are multiple interaction effects which correlate with the main effects, the importance should be interpreted carefully. The main effect of the variable price seems to be more important compared to its fixed (insignificant) counterpart (18.6% vs 1.9%). As was mentioned before, a higher total number of contacts seems to influence the preference towards a variable pricing model. The interaction between those two has an importance of 14.9%, exceeding the relative importance of the interaction between the fixed price and the number of contacts (13%). Add‐on sales as a main effect is relatively unimportant (8.7%), but the importance of the interaction with number of calls is high (19%). The relative importance of the main effect variable unit (3.9%) and bonus (5.8%) is quite low. It seems that the price, as was expected, is the most important attribute.
Table 13 Relative importance of attributes
Variable Relative importance Fixed price 1.9% Variable price 18.6% Add-on sales 8.7% Variable Unit 3.9% Bonus 5.8% None option 14.4%
Nr of contacts *Fixed price 13.0%
Nr of contacts * Variable price 14.9%
35 4.6 Comparing different pricing models
To test H1, H2a and H3 different models can be compared to each other the same way as in
section 4.3. To check whether a variable pricing model is preferred more over a mixed‐ or a fixed pricing model, multiple models can be compared to each other while the other main effects and interactions are kept constant. To test the first hypothesis whether a mixed pricing model is more valued over a fixed pricing model, 4 scenarios, including the no‐choice option, have been sketched.
In the bottom of table 14, the utilities for the different scenarios are given. The fixed model has a utility of ‐1.018, whereas both mixed models have a utility of ‐1.173 and ‐0.836. The difference between the utilities is explained by the variable price. Mixed (1) has a €10,‐ variable price and Mixed (2) has a price of €5,‐. Obviously, the model with the lower variable price is perceived with a higher utility. Furthermore, it might be interesting to compare the utility values of the three scenario’s to the utility of the no‐choice option. For example, the probability that a respondent prefers ‘scenario 1’ over the no‐choice option, is
( . )
( . ) ( . ) = 62.1%. The probability that the no‐choice option is chosen over the
other three scenarios, is 16.63%.
However, there are more effects that should be taken into account when comparing the different models. Since there are three interaction variables in the final model, it is highly relevant to investigate what the effect of changes in the total number of contacts and the add‐on sales might be. To test for these effects, three new scenarios are compared to each other. The models are shown in table 15.
Table 14 Scenarios for estimation
Variable Scenario 1: Fixed Scenario 2: Mixed (1) Scenario 3: Mixed (2)
Scenario 4: No-choice option
Fixed price €200.000,‐ €100.000,‐ €100.000,‐
Variable price €0,‐ €10,‐ €5,‐
Add-on sales €10.000,‐ €10.000,‐ €10.000,‐
Variable unit Price per contact Price per contact Price per contact Bonus tariff Fixed 20% increase Fixed 20% increase Fixed 20% increase Number of contacts Between 3 and 5 times Between 3 and 5 times Between 3 and 5 times
36 Table 15 Scenarios for estimation
Variable Scenario 1: Fixed Scenario 2: Variable Scenario 3: Mixed
Fixed price €200.000,‐ €0,‐ €100.000,‐
Variable price €0,‐ €10,‐ €5,‐
Add-on sales €0,‐ €0,‐ €0,‐
Variable unit Price per solution Price per solution Price per solution Bonus tariff Variable 20% discount Variable 20% discount Variable 20% discount Number of contacts Between 3 and 5 times Between 3 and 5 times Between 3 and 5 times
Final model utility ‐0,4902 ‐0,1242 ‐0,3072
In these scenario’s, no add‐on sales possibilities are present. Figure 7 illustrates the impact of the total number of contacts on the utility. As was stated in the ‘Interpretation’ chapter already, it seems that with an increasing number of contacts, the utility towards a variable model increases. The opposite happens to a fixed model, in which the utility decreases with an increased number of contacts. The utility of a mixed model increases slightly (from ‐0.78 towards ‐0.51) with an increased number of contacts.
A variable model seems to be more appealing than a mixed and a fixed model at the point where the number of contacts per year with a customer is at least 3. With no add‐on sales present, it seems that a fixed model is highly preferred over a mixed or variable model when there is limited contact between a client helpdesk and its customers (less than three times per year). To measure the impact of the add‐on sales, two additional estimations are made.
Figure 7 Impact of no add-on sales on the utility of the pricing models ‐1,2 ‐1 ‐0,8 ‐0,6 ‐0,4 ‐0,2 0 0,2 0,4
Between 0‐1 Between 1‐3 Between 3‐5 Between 5‐10 More than 10
UTILITY
TOTAL NUMBER OF CONTACTS (PER YEAR)
Impact of no add-on sales
37 ‐2 ‐1,5 ‐1 ‐0,5 0 0,5 Between 0‐1 Between 1‐3 Between 3‐5 Between 5‐10 More than 10 UTILITY
TOTAL NUMBER OF CONTACTS (PER YEAR)
Impact of €10.000,- add-on
sales
Fixed model Variable model Mixed model
Figure 8 The impact of €10.000,- add-on sales on the utility
Figure 8 and figure 9 show the utilities for the three pricing models with respectively
€10.000,‐ and €20.000,‐ add‐on sales. The same trends can be observed as in the scenario in which there is no possibility for add‐on sales. When the average number of contacts is below three per year, both fixed and mixed models appear to be more appealing than a variable model. The impact of add‐on sales is clearly present for the variable pricing models. Figure 7 shows an increasing line for the utility with an increasing number of contacts. In figure 8, this line increases significantly less and the utility for a variable model does not even reach the value of 0. Figure 9 even shows a decreasing line from the beginning onwards.
In conclusion, the hypotheses H1, H2a and H3 can be accepted, but under strict conditions. It
seems that a mixed pricing model is more valued over a fixed pricing model at the point where a client helpdesk contacts its customer at least 3 times per year. The same conclusion can be drawn for a variable pricing model compared to the mixed pricing model. An overview of the acceptance and rejection of all hypotheses is given in table 16.
Table 16 Hypotheses table
Hypothesis Accepted/rejected (+/‐) 1 (+/‐) 2a (+/‐) 2b + 2c + 3 (+/‐) 4 + 5 + 6 ‐ ‐2,5 ‐2 ‐1,5 ‐1 ‐0,5 0 0,5 1 Between 0‐1 Between 1‐3 Between 3‐5 Between 5‐10 More than 10 UTILITY
TOTAL NUMBER OF CONTACTS (PER YEAR)
Impact of €20.000,-
add-on sales
Fixed model Variable model Mixed model
38
5. Discussion and conclusion
The shopping behavior of consumers has shifted from physical stores towards online channels, which has disruptive consequences for the way that companies sell their products and maintain their client helpdesk. The high amount of new competitors in the E‐Commerce businesses causes the market to act in a different way than it used to do. Since the E‐ Commerce market is relatively new, the CRM of these businesses is also less developed compared to more traditional businesses like the Energy‐ or Telecom market (Papaioannou et al., 2014). The purpose of this study was to find the most valued pricing model for outsourcing CRM activities. This study adds knowledge to the research field of CRM‐ outsourcing from the value based costs perspective, especially in the E‐Commerce market. The link between value‐based pricing and the optimal pricing model for the CRM outsourcing decision in the E‐Commerce market has not been researched yet. This way, a gap in the literature was addressed.
By the means of a choice based conjoint analysis, the utility towards fixed, mixed and variable pricing models was investigated. As expected, there is a significant negative effect for both the fixed price and the variable price towards the most valued pricing model (the higher the price, the lower the utility). If the main effects of the price attributes are compared, it seems that the variable component of the pricing model was perceived as most important. With the inclusion of the total number of contacts as an interaction variable, some surprising effects are found. It seems that a fixed pricing model is preferred over a mixed or variable model when there is limited contact between a company and its customers (less than three times per year). Since the expectation was that a variable model would be more valued than a fixed or a mixed model regardless of the number of contacts, hypothesis 1, 2a and 3 are not fully accepted. This also contradicts earlier findings of Aksin et al., (2008) and Ren & Zhou (2008).
One of the explanations might be that the statistical power of this research is somewhat limited, since only 40 respondents are used in this study. Another explanation could be that there are confounding variables, which also has to do with the limited sample size. The variance explained in this model is relatively small (R2adj of 4.2%), which implies that there
39 A different explanation could be that the markets which were investigated in the studies of Aksin et al., (2008) and Ren & Zhou (2008) focused on companies in more traditional market in which CRM outsourcing is very commonly used. It might be that these companies prefer other drivers in the price related attributes than the relatively young companies (who have limited experience in outsourcing) in the E‐Commerce market. Nonetheless, it remains surprising that a fixed pricing model is preferred when there is limited contact with the customers. The answer to the research question regarding the most valued pricing model for outsourcing CRM activities in the Dutch E‐Commerce market is therefore twofold. When the number of contacts with a customer on yearly basis is below a total of three, a fixed pricing model is preferred, whereas a variable pricing model is preferred when the number of contacts on a yearly basis is three or more.
Furthermore, the main‐ and moderating effects of add‐on sales were researched. The moderating effect was insignificant but the main effect of add‐on sales seems to have a positive, significant effect on the perceived value of a pricing model. However, the interaction effect between the add‐on sales and the number of contacts seems to decrease the utility for a variable pricing model. The higher the add‐on sales, the lower the utility towards a variable pricing model. In addition, an increase in add‐on sales combined with a higher total of contacts, lowers the utility of all of types of models. This is not in line with what was expected. One would expect that the utility for a pricing model would increase with a probability for a higher amount of add‐on sales in comparison to no add‐on sales possibilities.
40 The effects of the remaining variables are in line with what was hypothesized. The main effect of bonus tariff was significant, in which the discounts towards both the fixed and the variable price are perceived with a positive utility. However, there appears to be no interaction between the bonus tariff and the fixed and variable price. Furthermore, price per solution is perceived as more valuable compared to a price per customer or a price per contact.
