• No results found

Can the influence of NPS on future purchase behaviour be optimized by considering the role of the underlying motivation of NPS and relationship characteristics?

N/A
N/A
Protected

Academic year: 2021

Share "Can the influence of NPS on future purchase behaviour be optimized by considering the role of the underlying motivation of NPS and relationship characteristics?"

Copied!
43
0
0

Bezig met laden.... (Bekijk nu de volledige tekst)

Hele tekst

(1)

Can the influence of NPS on future

purchase behaviour be optimized by

considering the role of the underlying

motivation of NPS and relationship

characteristics?

by Amber Zijlstra

University of Groningen Faculty of Economics and Business

MSc Marketing Management PO Box 800, 9700 AV Groningen (NL)

Supervised by: Hans Risselada 18 June 2018

Trompsingel 7-11 9724 CX Groningen a.e.zijlstra.1@student.rug.nl

(2)

Management Summary

Since the introduction of the Net Promoter Score (NPS) by Reichheld in 2003, the NPS became one of the most frequently used corporate metrics predicting customer loyalty and purchase behaviour (Kristensen & Eskildsen, 2014). The main reason for this is that NPS is extremely simple and easy to use (Raassens & Haans, 2017). However, the predictive value of NPS on future purchase behaviour is not beyond criticism from other researchers (De Haan, Verhoef & Wiesel, 2014; Kristensen & Eskildsen, 2014). To maintain competitive advantage, companies need to improve the predictive value of NPS on customer purchases behaviour (Kristensen & Eskildsen, 2014). Therefore, this research explores two main improvements on the predictive value of NPS based on current literature. First, literature stresses that NPS is sensitive to changes in the underlying distribution of customer motives because NPS represents recommendation scores, whereby customer underlying motivations are overlooked. This results in losing a lot of relevant information which might turn out to be important determinants for customer future purchase behaviour (Kristensen & Eskildsen, 2014; Wiesel, Verhoef & De Haan, 2012; Kumar, Petersen & Leone, 2007; De Haan, Verhoef & Wiesel, 2014). Second, the NPS measurement does not take the influence of customers’ relationships into account while CRM literature demonstrates that relationships certainly shape the influence of NPS on customer purchase behaviour apparently (Rodriquez, Peterson & Krishnan, 2018). Moreover, Srinivasan, Anderson & Ponnavolu (2002) suggests that a more engaged relationship between the customer and the company, results in more committed behaviour of the customer. This study provides insights in which combinations of NPS and company components in the underlying motivation of NPS are most effective and which combinations are most valuable for the company. Moreover, this research provides new insights in challenging and improving the predictive value of NPS of future purchase behaviour by combing literature on NPS and current CRM literature.

(3)

This study illustrates some interesting findings. First, it seems that the customers NPS score indeed influences customers’ future purchase behaviour. Second, we found that the NPS score of customers addressing Returns in their underlying motivation of NPS influence the customer purchase behaviour more than the NPS score of customers who addresses any other company component in their underlying motivation of NPS. This means that considering customers’ underlying motivation of NPS strengthens the influence of NPS of future purchase behaviour and therefore must be taken into account by companies and academics. In addition, results show that in general customers being in a long relationship with the company purchasing more than one product from more than one category have a higher likelihood for making a purchase at the retailer in the future. Therefore, we suggest that companies must invest in building sustainable customer relationships. However, the relationship characteristics do not strengthen the influence of NPS on future purchase behaviour. Finally, older customers seem to be less likely in making a purchase in the future than younger customers and women seem to be more likely in making a purchase in the future compared to men.

Overall, we can conclude that more research needs to be done about which is the best metric in combination with NPS to predict customers’ future purchase behaviour. Furthermore, more research needs to provide insight what exactly drives the NPS score of customers, since we did not find a perfect solution for that.

Keywords: Net Promoter Score (NPS), Future Purchase Behaviour, Purchase

(4)

Preface

Last February I got the opportunity to combine writing my master thesis with doing an internship at one of the biggest online retailers of the Netherlands. This gave me the opportunity to gain work experience at a professional company while finishing up my master Marketing Management. I learned a lot about the business and one of their core drivers, as being challenged and investigated in my thesis. During this period, I learned how to balance between work, writing my thesis and having spare time. Moreover, at my internship I got the opportunity to developed myself on both social and professional level. When looking back at this period, I can say that I am very satisfied with the result.

Furthermore, I would like to thank the people who helped me during this process. First of all, I would like to thank my first supervisor Hans Risselada, for all his useful input, feedback, effort and support while writing my master thesis. More especially, thank you for guiding me how to keep focus on the main subject. Furthermore, I would like to thank my second supervisor Jelle Bouma, who, next to being my second supervisor, spent a lot of time and effort in order to make the combination of writing a master thesis and gaining working experience at professional companies possible. Finally, I am really thankful for the input, corporation, feedback and dedication my supervisor at the online retailer gave me in order to succeed writing this thesis. Also, he helped me with my personal development and learned me how to challenge myself in order to keep growing and in the end become a professional myself.

(5)

Table of content

Introduction 6

Literature review 9

Future Purchase Behaviour 9

NPS 10

Underlying motivation of NPS 12

Relationship Characteristics 13

Relationship Length, Width and Depth 14

Conceptual model 16

Methodology 17

Participants and design 17

Measures 17

Control variables 18

Outliners and missing values 19

Descriptive statics 20 Logic regression 21 Plan of analysis 23 Results 24 Preliminary checks 24 Hypotheses testing 26 Overview of hypotheses 27 Discussion 28

Hypotheses and findings 28

General discussion 28

Academic and managerial implications 30

Limitations and future research 31

Literature 34

(6)

1. Introduction

Since the introduction of the Net Promoter Score (NPS) by Reichheld in 2003, the NPS has become increasingly popular to predict customer loyalty (Raassens & Haans, 2017). The NPS determines to what extend the customer is willing to recommend the company to friends or colleagues (Reichheld, 2003). Due to its simplicity and ease in use, the NPS turned out to be one of the most frequently used corporate metrics predicting customer loyalty and corresponding consumer (purchase) behaviour (Kristensen & Eskildsen, 2014).

Although, the coherence of NPS and customer satisfaction is substantiated in literature, the predictive value of NPS on future purchase behaviour is not beyond criticism from other researchers (De Haan, Verhoef & Wiesel, 2014; Kristensen & Eskildsen, 2014). First of all, the results of Morgan & Rego (2006) shows that focussing on recommendation intentions and behaviours is misguided and does not lead to growth of the firm as well as effective predicting consumer purchase behaviour. In addition, Keiningham, Cooil, Andreassen & Aksoy (2007) also found no support for the claims of Reichheld illustrating that NPS is the “number one you need to grow”. Moreover, since most companies do not focus on implications of NPS, companies provide a suboptimal implementation of NPS in predicting future purchase behaviour of customers correctly (Bulik, 2013). To maintain competitive advantage, companies need to improve the predictive value of NPS on customers purchases behaviour (Kristensen & Eskildsen, 2014). This research explores two main improvements on the predictive value of NPS.

(7)

components in the underlying motivation of NPS exploratively. To do so, the customers’ underlying motivation of NPS is separated into 7 company components, based on current customer journey literature and better known as: image, assortment, website, order, returns, delivery, and customer care. Overall, this research (1) proposes that the underlying motivation of NPS will influence the predictive value of NPS on purchase behaviour and (2) explores if the effect of the underlying motivation of NPS differs in the presence of various company components.

The second potential problem of the predictive value of NPS is that the NPS measurement does not take customers relationships into account. Research on customer relationship management (CRM) demonstrates that the customers’ relationship characteristics can shape the influence of NPS on customer purchase behaviour apparently (Rodriquez, Peterson & Krishnan, 2018). According to Srinivasan, Anderson & Ponnavolu (2002), the stronger the relationship between customer and company, the more committed the customer behaves. This assumes that, even though a low NPS is given by a customer with a strong relationship to the company, the likelihood of the customer to make purchases in the future is above standard (Morgan & Hunt, 1994; Sheth & Parvatiyar, 1995). However, current literature inadequacy distributes the relationship characteristics length, width, and depth when talking about customer relationships, while other literature assumes them to have different influences on customer purchase behaviour (Verhoef, 2003). Therefore, this research considers the relationship characteristics separately and propose relationship length, width, and depth to have influences on the predictive value of NPS on customers’ future purchase behaviour.

All in all, this research proposes that by addressing two drawbacks of NPS, the predictive value of NPS on future purchase behaviour can be improved and an explanation on the differences in customer purchase behaviour per NPS can be given. Therefore, the research question is as followed:

Can the influence of NPS on future purchase behaviour be optimized by considering the role of customers’ underlying motivation of NPS and customer relationship characteristics?

(8)

complexity and number of variables. Next, the model uses a logistic regression to test the hypotheses, since we observed the probability the purchase took place or not. Moreover, we found that the NPS score of customers addressing returns in their underlying motivation of NPS influences customer purchase behaviour more than the NPS score of customers who addresses any other company component in their underlying motivation of NPS. Also, results show that customers being in a long-term relationship and purchasing several products from different categories, in general, have a higher likelihood to make future purchases at the retailer. This means that this research provides insights in which combinations of NPS and company components in the underlying motivation of NPS are most effective and which combinations are most valuable for the company. Moreover, the findings can be used a benchmark for online retailers who use NPS in predicting future purchase behaviour. Besides practical contribution, the academic contribution is that this research provides new insights in challenging and improving the predictive value of NPS of future purchase behaviour where literature desirably ask for. Connections between NPS and company component distributions and NPS, and CRM are combined which are not highlighted before in existing literature.

(9)

2. Literature review

This chapter provides a conceptual model based on arguments from current literature. The aim of the conceptual model is to tests whether the influence of NPS on customers’ future purchase behaviour can be improved by considering the role of customers’ underlying motivation of NPS and customer relationship characteristics on this effect. Future purchase behaviour is set as dependent variable, whereas NPS is set as independent variable. The underlying motivation of NPS and relationship characteristics are set as moderators. Any effect in the presence of the company components in the underlying motivation of NPS will be tested exploratively.

The chapter is structured as follows. In the first paragraph, the concepts are explained using arguments from existing literature. Next, the concept of customer future purchase behaviour is described. After that, the influence of NPS on future purchase behaviour is explained. After that, the influences of the underlying motivation of NPS and relationship characteristics on the relationship between NPS and future purchase behaviour are discussed. Finally, the variables are summarized and showed in a conceptual model.

2.1 Future purchase behaviour

Purchase behaviour is described as the decision of actual buying and using products (Morrison, 1979). According to Guo & Barnes (2011) a purchase is only made when (1) the customer recognizes the need and motivation to buy the product and (2) the intentions and attitudes of the customer toward the product and brand are certain.

(10)

Next to that, researches in the marketing and retailing field have expended significant effort to predict purchase behaviour more correctly, because of its impact on business processes. Successfully predicting purchase behaviour helps companies to create efficient long-term strategies which results in lowering costs (Qui, Lin & Li, 2015). Moreover, successfully predicting purchase behaviour helps companies to better prepare for peaks in sales and service which leads to a more effective approach towards customers, which is exceptionally preferred by customers nowadays (Guo & Barnes, 2011). As a result, many companies currently invest a lot of time and effort to predict customer purchase behaviour (Moe & Fader, 2004).

2.2 Net Promotor Score (NPS)

The Net Promotor Score (NPS) is developed in 2003 by Reichheld as a customer feedback metric, answering the question: “How likely is it that you would recommend the company to friends and colleagues”. The NPS is a 10-scale variable, where scores between 0 and 6 are labelled as detractors, 7 and 8 as passives, and scores of 9 and 10 as promoters. According to Reichheld (2003), the NPS gives more insight in the level of loyalty of the company’s customer base and the growth of the business. This logic is simple, because it is unquestionably better to have more promotors than detractors and passives. Promotors therefore recommend the company to friends and colleagues, whereas that is not the case by passives and detractors. Due to its simplicity and intuitively in practice, NPS is one of the most widely used metric by companies to predict future purchase behaviour (De Haan, Verhoef & Wiesel, 2014; McGregor, 2006; Morgan et al., 2005). According to Reichheld (2003), NPS is “the one number you need to grow”.

(11)

performances. In order to make a suitable statement, this research discusses two interesting impactful drawbacks of NPS.

As stated in chapter 1, the first potential problem with the inadequate influence of NPS is related to the sensitiveness of NPS to changes in the underlying distributions of customer motives (Kristensen & Eskildsen, 2014). Research assumes that the influence of NPS is limited because much information is lost by only focusing on a fixed scores and not the underlying motivation including feelings, emotions, and motivations of customers (Van Doorn, Leeflang & Thijs, 2013). Furthermore, customers’ diverse interpretation of NPS increases the risk of drawing wrong conclusions by not taking underlying motivations into account (Saris & Revilla, 2015). Whereas some customers interpret a NPS score of 8 as an extremely high score, others interpret a NPS score of 8 as a low to average score. However, this difference in scores does not automatically mean that the second consumer is less likely to make purchases in the future (Morgan & Rego, 2006). Moreover, this interpretation bias seems to occur at the cultural level as well. The research of Van Doorn, Leeflang & Tijs (2013) emphasize that Dutch customers, on average, rate companies lower on NPS than American customers. However, by comparing the likelihood of future purchases between the two consumer groups, the Dutch consumers do not have a lower likelihood to make future purchases than the American consumers, which was previously indicated based on a lower NPS (Van Doorn, Leeflang & Tijs, 2013).

A second potential problem with the influence of NPS is that NPS overlooks the value of customer relationships (Rodriquez, Peterson & Krishnan, 2018). A strong relationship between the customer and the company guarantees for very committed behaviour of the customer, such as frequent purchase behaviour and a positive worth of mouth. However, important to stress is that loyalty and willingness to recommend are not the same and can result into different purchase behaviours (Srinivasan, Anderson & Ponnavolu, 2002). For that reason, recommendation scores can be misguiding when not taking relationship characteristics into account (Rodriquez, Peterson & Krishnan, 2018). Therefore, research assumes that, even though a low NPS score is given by a customer with a strong relationship to the company, the possibility of making purchases in the future is above standard (Morgan & Hunt, 1994; Sheth & Parvatiyar, 1995). Therefore, the relationship characteristics must be taken into consideration when talking about the influence of NPS on future purchase behaviour.

(12)

However, besides the two discussed drawbacks of NPS, literature seems to outweigh in promoting the positive influence of NPS on Future Purchase Behaviour. Therefore, this research assumes that:

Hypothesis 1: The NPS positively influence customers’ future purchase behaviour.

2.3 The underlying motivation of NPS

It is confirmed that the growing reliance on one simple customer metric like NPS is not the best way to predict purchase behaviour (Saris & Revilla, 2015). To predict future purchase behaviour more correctly, Van Doorn, Leeflang & Tijs (2013) strongly urge companies to use more nuanced multi-dimensional approaches to measure purchase behaviour. A benefit of using more than one metrics is that every metric measures its own dimension of customer behaviour, adding unique information into the system (De Haan, Verhoef & Wiesel, 2014). De Haan, Verhoef & Wiesel (2014) found evidence that the NPS combined with a sentiment analysis on the underlying motivation of NPS could be the best predictor for customer purchase behaviour in most sectors. A sentiment analysis is a subjective analysis (Wiebe, 1994) that identifies positive and negative opinions, emotions, and evaluations expressed in natural language (Wilson et al., 2009). Customers with a positive underlying motivation are more satisfied about the company and show a more loyal purchase behaviour compared to customers with a negative underlying motivation (East, Romaniuk & Lomax, 2011). Where NPS focuses mainly on willingness to recommend, customers’ underlying motivations focus more on attitudes, thoughts, and feelings (Wiebe, 1994). This means that the underlying motivation of NPS is able to decreases the likelihood for fluctuations in outcomes by including customers “explanation” to the NPS score (Wilson et al., 2009). These explanations are valuable information, which would not be achieved otherwise. The possibility for interpretation biases between customers will decrease, since the underlying motives will explain the interpretation of the customers. As a result, the visibility of the personal explanations of customers will reduce the risk of drawing wrong conclusions which can have an enormous impact on business operations (Saris & Revilla, 2015). For that reason, we propose the following:

Hypothesis 2: The underlying motivation of NPS strengthens the positive relationship between NPS and future purchase behaviour.

(13)

based on one satisfied experience with one company components does not automatically mean that the customer is satisfied about every company component. A next experience with another company component can therefore change the customer NPS score, whereas the customers’ purchase behaviour will not change necessarily (Kristensen & Eskildsen, 2014). For that reason, the effect of the presence of the company components in the underlying motivation of NPS will be tested exploratively. The explored company components are chosen based on current customer journey literature and compared to each other. The used company components are: image, assortment, website, order, returns, delivery, and customer care.

2.4 Relationship Characteristics

(14)

According to Lehmann (2004) NPS focuses on the customers’ willingness to recommend and is not able to measure the influence of customers’ relationships characteristics on future purchase behaviour. For that reason, this research explores whether future purchase behaviour can be predicted more correctly and fluctuations in outcome will be decreased by considering the effect of customer relationship characteristics on the influence of NPS on future purchase behaviour.

Nevertheless, common literature does not discuss any distinctions in relationship characteristics at all when talking about the influence of customer relationship characteristics on NPS and future purchase behaviour (Bolton & Tasari, 2007). However, this is a remarkable conclusion, since literature on CRM demonstrates different outcomes in effect of different relationship characteristics for other company related researches (Day, 2000).

We measure the effect of customer relationship using the three dimensions: relationship, length, width and depth to see if we can support this literature for NPS and future purchase behaviour as well.

2.4.1 Relationship Length, Width and Depth

First, customer relationship length is described as the number of years in which the customer is registered in the data register of the company starting from the day of subscription (Bolton, Lemon & Verhoef, 2004). Customers that have a long relationship with the retailer have more experience with the company and therefore form more stable expectations, opinions, and attitudes towards the company compared to customers with a short relationship (Wang & Wu, 2012; Coulter & Coulter, 2002). Furthermore, customers with a long relationship do not have to consider about purchasing products from the company anymore, it is just a manner of routine (Wang & Wu, 2012). Once the routine is made, the likelihood of purchasing products from another company decreases because customers like consistency and routines (Coulter & Coulter, 2002). Therefore, we argue that the NPS score of customers with a long relationship with the company, will be influencing their future purchase behaviour less than the NPS score of customers in shorter relationships. In other words, we suggest that the relationship length weakens the influence of NPS on future purchase behaviour, since these customers show a more loyal purchase behaviour anyway (Wang & Wu, 2012). The longer the relationship, the more customers form stable opinions and routines, the less influence the NPS has on the future purchase behaviour of these customers.

(15)
(16)

feeling of familiarity and trust, the less influence the NPS has on the future purchase behaviour of these customers.

In summary, we propose the following hypotheses:

Hypothesis 3.1: The relationships length weakens the positive relationship between NPS and future purchase behaviour.

Hypothesis 3.2: The relationships width weakens the positive relationship between NPS and future purchase behaviour.

Hypothesis 3.3: The relationships depth weakens the positive relationship between NPS and future purchase behaviour.

2.5 Conceptual model

The conceptual model is graphically depicted in figure 1; its components and relations are based on the previous sections.

(17)

3. Methodology

In this chapter, the method and design of the research are shown. First, the participants’ selection and design are explained. Second, the measures of the variables are presented and the control variables are described. Afterwards, the data cleaning process is explained and the descriptive results are given. Finally, the statistical models are presented and a plan of analysis is drawn.

3.1 Participants’ selection and design

The data used in this research was retrieved from an existing dataset of a Dutch online retailer. This dataset was gathered from a survey which was automatically send by the online retailer three weeks after the customer had made a purchase at the retailer from the period 01-2017 till 10-01-2017. The survey was sent to approximately 130,000 customers with a netto response of 9,152 (7,0%), contacted via e-mail. The respondents were a random selection of the customer base, since they followed the standard pattern of the segments of the retailer with a maximum deviation of 5%. From high to low: High Engaged (74%), Newbie (11%), Disengaged (10%) and Low Engaged (5%). This survey provides more insights in the NPS scores of customers, their underlying motives of the NPS scores and their overall satisfaction with the online retailer. The survey took approximately 5-10 minutes to complete.

3.2 Measures

(18)

motive. However, we know that making this choice comes with some restrictions which will be further explained in the limitation section. Lastly, the labelling was checked and 23 motives were switched to another better fitting company component in the opinion of the researcher. Motives as “overall satisfaction” and “general ease in use” were classified as image. In the end, we were able to divide all motives into one of the company components. The customers’ relationship length, width and depth could be detected from the CRM system of the online retailer. Future purchase behaviour was indicated whether the customer made a purchase within 5 months after filling in the survey and so five months after indicating their NPS score and underlying motivation. A period of 5 months was used, since the retailer changed the measurement of NPS one year ago. To excluded biases of different measurements and still obtain enough respondents, we agreed to set the Future Purchase Behaviour at 5 months, which is 1/3 of the standard used Future Purchase Behaviour by the online retailer. All in all, the survey needed to be 5 months ahead, which means that the data collection of the survey took place at least 5 months before starting this research. Table 1 exhibits an overview of the measures.

Variable Measure Description

Future Purchase Behaviour Dummy variable 0: no purchase within next 5 months 1: purchase within next 5 months Net Promoter Score Interval scale 1-10,

mean-centred

Continuous

Underlying motivation of NPS (distributed over Image, Assortment, Website, Order, Delivery, Returns, Customer Care)

Dummy variable per company component

0: not present in underlying motive 1: present in underlying motive

Relationship Length Length in years Continuous

Relationship Width Dummy variable 0: purchases in only one product category 1: purchases across product categories Relationship Depth Dummy variable 0: only one purchase per category

1: more than one purchases across categories

Gender Dummy variable 0: male

1: female

Age Age in years Continuous

Table 1: Summary of the measures 3.3 Control variables

(19)

predicting future purchase behaviour (Padmanabhan, Zheng & Kimbrough, 2001; Van den Poel & Buckinx, 2005).

According to Rodgers & Harris (2003) male customers are assumed to be more satisfied and show a more positive attitude towards online purchases compared to female customers. Female customers are suggested to have less trust, are more sceptical and are less satisfied towards making online purchases (Rodgers & Harris, 2003).

Furthermore, research claims that age has an inverted U-shape relationship with future purchase behaviour on the internet (Venkatesh & Agarwal, 2006). This can be explained by highlighting the two extremes. Young children are not able to make a purchase, because they simply do not manage their own bank account. This can be confirmed with the variable age, which starts from 16 years and older. However, suggested is that when they do, however, purchases will grow exponentially because young people are familiar with making purchases. On the other hand, older people are less willing to make a purchase because they get more thoughtful and deliberate in their purchase process (Venkatesh & Agarwal, 2006).

Finally, the customers of the online retailer are mostly female and middle-aged, and therefore we include gender and age in the analysis in order to test whether age and gender influences our model and to exclude these possible additional explanations for purchase behaviour.

3.4 Outliners and missing values

The data needs to be cleaned for outliners and missing values to create meaningful insights. Therefore, the obtained data has been carefully examined beforehand. Histograms and descriptive statics were used to make this insightful. See table 2 for the summarized descriptive statics.

(20)

observation which were removed from the mass. However, no strange patterns were seen since the pattern of age and NPS showed a smooth distribution from the mass to the large or low numbers. For that reason, the large numbers are kept in the data. In the end, we reached a sample size of 8187 respondents.

3.5 Descriptive statics

From the respondents 73,9% were women and the age varied between 16 and 105 years old with a mean of 49 years old, as showed in table 2. The average relationship length of the respondents was 11.4 years, which is relatively high compared to the overall customer relationship length of the retailer which is 8.6 years. Of these respondents, 33,3% had a wide and deep relationship with the online retailer. More specifically, on average 33.3% of the respondents made purchases in more than one product category, whereas 67.7% of the respondents purchased more than one product across categories. Furthermore, 71.1% of the 8187 respondent did a purchases within 5 months after the survey was conducted. The average NPS of the respondents was 8.6, which is extremely high. Table 2 exhibits an overview of the descriptive statics.

Variable Min Max Mean Std. deviation

Variable Min Max Mean Std. deviation NPS 0 10 8.56 1.58 Returns 0 1 (6.7%) 0.07 0.25 Purchase Behaviour 0 1 0.71 0.45 Customer Care 0 1 (14.8%) 0.15 0.36 Image 0 1 (11.7%) 0.12 0.32 Relationship Length 0 53 11.44 10.98 Assortment 0 1 (10.3%) 0.10 0.30 Relationship Width 0 1 0.33 0.47 Website 0 1 (10.2%) 0.10 0.30 Relationship Depth 0 1 0.68 0.47 Order 0 1 (17.6%) 0.18 0.38 Age 16 105 48.87 13.30 Delivery 0 1 (28.7%) 0.29 0.45 Gender 0 1 0.74 0.44

Table 2: Descriptive statics

(21)

Delivery, it was mostly about the speed and correctness of the Delivery process. Other positive company components were Returns and Customer Care. In general, respondents were positive in their underlying motivation of NPS about the easy of returning products and the kindness and problem solving ability of the Customer Care. Order was experienced as one of the least positive experience. Topics related to order were correctness of product, package, ordering process and the paying process. Around 8% of the responds were “just satisfied” without mentioning a specific reason. These underlying motives were classified as image, since image is an overarching company component. Overall, the most negative experiences in the underlying motives were about a mixture of missed promises (eg. missed delivery time, obtaining wrong product, broken package, product out of stock) and were mostly a one-time occasional. Furthermore, respondents addressing Delivery or Website in their underlying motivation of NPS were most likely to purchase within the next 5 months. On the other hand, respondents addressing image in the underlying motivation of NPS were less likely to make a purchase in the next 5 months. Table 3 exhibits an overview of the average NPS and future purchase behaviour per addressed company component.

Company Component

Average NPS Percentage of presence Company Component making a future purchase

Company Component

Average NPS Percentage of presence Company Component making a future purchase

Image 8.8 66% Delivery 8.9 75% Assortment 8.1 72% Returns 8.6 72% Website 8.5 75% Customer Care 8.6 68% Order 8.1 68% Frame of reference 8.6 74%

Table 3: Average NPS and Future Purchase Behaviour by presence Company Component

3.6 Logistic regression

In order to test the predictive value of NPS on Future Purchase Behaviour a binary choice model is used, since our dependent variable (Future Purchase Behaviour) is a dummy variable showing whether a purchase took place in the next 5 months or not. More specifically, a logistic regression model is applied in this research.

(22)

applying a logistic CDF (Leeflang & Bijmolt, 2013). The probability that the customer will make a purchase is shown by use of an observed variable (0-1). When this probability exceeds .5 the customer will most likely make the purchase. A probability beneath .5 means that the customer is not likely to make a purchase. By interpreting the coefficients of the logistic regression model, the odds ratio is used. The odds ratio is the likelihood of making a purchase vs. not making the purchase, and is calculated by the probability of purchasing (yes) divided by the probability of not purchasing (no). An odds ratio of five means that the probability of purchasing is five times larger than the probability of not purchasing. However, the odds ratio is mostly interpreting as the log odds ratio which indicates the change in odds if the independent variable change one unit (Leeflang & Bijmolt, 2013).

Therefore, the hypotheses are tested using a logistic regression analysis. However, to give a clear and reliable overview of the influence of the variables, the variables are included stepwise because of the complexity and amount of hypotheses. Model 1 is the simplest model with only the direct influence of NPS on Future Purchase Behaviour. Model 2 includes the direct influence of NPS and the moderating effects of the company components in the underlying motivation on Future Purchase Behaviour. Model 3 & 4 are respectively model 1 & 2 including the influence of the control variables Age and Gender. Model 5 includes the direct effect of NPS and the moderating effect of Relationship Length, Width and Depth on Future Purchase Behaviour. Next, model 6 add the control variables Age and Gender to model 5. Model 7 includes the direct influence of NPS, the moderating effects of the company components in the underlying motivation on Future Purchase Behaviour, the direct influence of the relationship characteristics and the influence of the control variables Age and Gender. Lastly, model 8 includes all variables. In summary, the following model is tested:

Logit(FuturePurchaseBehaviour) = Log(Probability of FuturePurchaseBehaviour/1-FuturePurchaseBehaviour) = β0 + β1NPS + β2Image + β3Assortment + β4Website + β5Order + β6Delivery + β7Returns + β8Customer Care + β9RelationshipLength + β10RelationshipWidth + β11RelationshipDepth + β12NPS x Image + β13NPS x Assortment + β14NPS x Website + β15NPS x Order + β16NPS x Delivery + β17NPS x Returns + β18NPS x

(23)

3.7 Plan of analysis

The plan of analysis consists of two parts. First, the most reliable model is chosen for the logistic regression. As state before, currently we developed 8 different models, of which model 1 is the least complex model including only the direct influence of NPS on Future Purchase Behaviour and model 8 is the most complex model including all variables. Between model 1 and 8, the variables are included stepwise. However, the most reliable and valid model will be selected using different metrics. The most popular metrics to validate the models are the Pseudo R2: Cox & Snell and Nagelkerke. These R-squares are applied differently than in a linear model, since Pseudo R2 compare the likelihood of the null model with the log-likelihood of the estimated model (Leeflang & Bijmolt, 2013). In this case, the higher the Pseudo R2, the better the estimated model. Next, the omnibus test is applied to see if the models are significant, only significant models are reliable. Also in this case, the higher the significance the better the estimated model. Next to the omnibus test, the Hosmer and Lemeshow Goodness of Fit is calculated. Different from the omnibus, the Hosmer and Lemeshow Goodness of Fit must not be significant (Hosmer & Lemeshow, 1989). Only models with an insignificant Hosmer and Lemeshow Goodness of Fit will be considered for future analysis. Than the likelihood ratio will be estimated by comparing the estimated models to the null model. In this research, the likelihood ratio shows how likely the respondent is to make a purchase in the next 5 months. The higher the ratio, the more likely the participants will make the purchase. Therefore, the lower the number, the better the estimated model. Next, the hit rate will be calculated of the estimated models, which are the overall percentage of correctly classified predictions of de estimated models (Leeflang & Bijmolt, 2013). Therefore, the higher the hit rate, the better the estimated model. After we succeed all steps, the final choice will be based on the Bayesian Information Criterion (BIC). The BIC resolves the problem of overfitting the likelihood of the models by introducing a penalty term for the number of parameters in the model (Schwarz, 1978). We use the BIC, since the BIC is the default for model selection for complex models like we use in this research. The model with the lowest BIC will be preferred and will be used to test the hypotheses.

(24)

4. Results

In this chapter the results are showed. First, the preliminary checks are applied and the most reliable model is chosen for further analysis. Second, the results of the logistic regression are shown. Finally, a table representing an overview of the supported hypotheses is given. 4.1 Preliminary checks

To give a clear and reliable overview of the influence of the variables, the variables are included stepwise. In summary, as described in the methodology section, model 1 is the simplest model with only the direct influence of NPS on Future Purchase Behaviour. Model 2 includes the direct influence of NPS and the moderating effects of the company components in the underlying motivation on Future Purchase Behaviour. Model 3 & 4 are respectively model 1 & 2 including the influence of the control variables Age and Gender. Model 5 includes the direct effect of NPS and the moderating effect of Relationship Length, Width and Depth on Future Purchase Behaviour. Next, model 6 add the control variables Age and Gender on model 5. Model 7 includes the direct influence of NPS, the moderating effects of the company components in the underlying motivation on Future Purchase Behaviour, the direct influence of the relationship characteristics and the influence of the control variables Age and Gender. Lastly, model 8 includes all variables. Table 4 represents the results of the preliminary checks for all 8 estimated models.

(25)

a hitrate above the 71.6%. However, the highest percentage of correctly classified predictions is for model 7 & 8 which both have a hitrate of 76.8%. Since the hitrate of the null model is 71.1%, all models outperform the null model. Therefore, we can state that including the Company Components in the underlying motivation of NPS and Relationship Characteristics variables do indeed more successfully predict Future Purchase Behaviour compared to random selection. Especially including the Relationship Characteristics variables increase the predictive value of the estimated models, since model 5, 6, 7 and 8 have the highest hitrates.

(26)

effects of the Relationship Characteristics are not included in the analysis and model 7 will be used to test the hypotheses. Model 7 includes the variables: NPS, Future Purchase Behaviour and the direct and interaction effect of the company components Image, Assortment, Website, Order, Delivery, Returns and Customer Care and the direct effect of the Relationship Characteristics. Table 4 exhibits the output of the logistic regression.

4.2 Hypotheses testing

For each variable the significance level, coefficients signs, exp(β) and marginal effect will be discussed. The exp(β) of the variables shows by which factor the odds will increase or decrease. The marginal effect displays how much the purchase probability changes if the explanatory variable changes by one unit.

First, the influence of NPS on future purchase behaviour will be presented. NPS (β = 0.059, p = 0.067) has a significant positive effect on Future Purchase Behaviour at a level of 10%. This means that one number change in NPS make a future purchase more likely to occur, all other factors held constant. More explicitly, we can state that a one-point increase in NPS score yield a change in log odds of 0.059. In terms of odds ratios, we can state that the odds ratio is exp(0.059) for a one-point increase in NPS score.

Next, for the presence of Company Components in the underlying motivation of NPS, only the presence of Website (β = 0.234, p = 0.041) shows a significant effect on Future Purchase Behaviour on the 5% level with Customer Care as reference category. This means that when Website is present in the underlying motivation of NPS of the customer, Future Purchase Behaviour is more likely to occur compared to when Customer Care is present in the underlying motivation of NPS. Particularly, we can state that the change in Website from not present to present yield a change in log odds of 0.234. In terms of odds ratios, we can state that the odds ratio is exp(0.234) for a change in Website from not present to present. Moreover, there is a positive significant interaction effect of Returns at the 1% level (β = 0.210, p = 0.005), which means that when the Company Component Returns is present in the underlying motivation of the customer, the positive influence of NPS on Future Purchase Behaviour will be stronger compared to when Customer Care is present in the underlying motivation of NPS. In other words, the ratio of the odd ratios of NPS and presence of Returns turn out to be the exponentiated coefficient for the interaction term of NPS by Returns in exp(0.210). All other company components are not significant.

(27)

p = 0.000) and Depth (β = 1.635, p = 0.000) indeed increase the likelihood of purchase in the future. The longer, wider, deeper the relationship, the more likely the customer will make purchase. More explicitly, we can state that a one year/point increase in Relationship Length, Width and Depth yield a change in log odds of respectively 0.016, 0.721 and 1.635. In terms of odds ratios, we can state that the odds ratio is respectively exp(0.016), exp(0.721) and exp(1.635) for a one year/point increase in Relationship Length, Width and Depth.

Lastly, the control variables age (β = -0.015, p = 0.000) and gender (β = 0.304, p = 0.000) are both significant at a level of 1%. Results shows that age has a negative effect on future purchase behaviour, which means that one year change in age make a future purchase less likely to occur. On the other hand, gender positively effects Future Purchase. Since women is 1 and men were 0, this means that women are more likely to make a purchase in the future. 4.3 Overview of hypotheses

Table 5 exhibits an overview of the supported hypotheses. As assumed, the positive influence of NPS on Future Purchase Behaviour, as state in hypothesis 1, is (marginally) supported. Furthermore, since some presence of Company Components in the underlying motivation of NPS show significant results, we can state that the underlying motivation of NPS indeed strengthens the influence of NPS on Future Purchase Behaviour. Therefore, hypothesis 2 is marginally supported. Lastly, results show that all three Relationships Characteristics positively influence Future Purchase Behaviour. However, the interaction effects of the Relationship Characteristics are not significant and therefore hypotheses 3.1, 3.2 and 3.3 are not confirmed.

Hypothesis Effect Hypothesis Supported

H1: Direct influence NPS + Yes*

H2: Moderating influence underlying motivation of NPS

+ Yes*

H3.1: Moderating influence Relationship Length + No, but direct influence

H3.2: Moderating influence Relationship Width + No, but direct influence

H3.3: Moderating influence Relationship Depth + No, but direct influence

Note: *marginally supported

(28)

5. Discussion

In this chapter the results of this research are discussed and challenged. First, the findings are shortly described and compared to existing literature and previous expectations. Next, the academic and managerial implications are explained. Finally, the limitations are described and recommendations for future research are given.

5.1 Hypotheses and findings

The main goal of this research is to investigate whether the influence of NPS on Future Purchase Behaviour can be optimized by considering the role of` customers’ underlying motivation of NPS and customer Relationship Characteristics using real observations of an online retailer. Results shows that the higher the NPS of the customer, the more likely the customer is to make a purchase in the future from the online retailer. This result is in line with the findings of Reichheld (2003), who established the statement that an increase in NPS leads to a more profitable company. Moreover, by looking at the customer motivations of NPS, we can state that indeed customers’ underlying motivation of NPS strengthens the positive influence of NPS on Future Purchase Behaviour, since some Company Components show significant effects in our results. More specifically, results show that when a customer addresses Returns in his/her underlying motivation of NPS, his/her future purchase behaviour will be more driven by his/her NPS score compared to the Purchase Behaviour of customers addressing other Company Components in his/her underlying motivation of NPS. However, in the case of this online retailer, we did not find that the presence of Image, Assortment, Website, Delivery, Order, or Customer Care in the underlying motivation of NPS is able to increase the predictive value of NPS on Future Purchase Behaviour. Additionally, once the customer addresses Website in his/her underlying motivation of NPS, the likelihood to make future purchase from the online retailer in general increases. Moreover, the likelihood to make future purchase will increase when the customer has a long, wide and/or deep relationship with the retailer. Our research collaborates on existing research of the predictive value of NPS on Future Purchase Behaviour, which indicates to be influenced by the customers’ underlying motivation of NPS. 5.2 General discussion

(29)

with a lower NPS score. This means that indeed the customers purchase behaviour can be predicted by knowing the NPS score of the customer. However, this contradict the results of Morgan & Rego (2006) who undermine the claim of Reichheld (2003) of “NPS being the number one you need to grow” by showing no significant effect of NPS on purchase behaviour in their study. We agree with Reichheld (2003) and the results of Van Doorn, Leeflang & Thijs (2013) that NPS is able to give an insight in the loyalty of the customer base and therefore, by attracting more loyal customers, the growth of the business. Moreover, this outcome is valuable for managers because monitoring NPS is a simple and costless methods to predict Future Purchase Behaviour. Using this method will save managers much time and money which can be used for other projects. According to Qui, Lin & Li (2015) by using NPS an increase or decrease in sales can be effectively indicated, whereby NPS give companies the opportunity to react fast on changes in sales and activity. Therefore, we can clearly state that managers can use NPS to drive strategy and operations (Reichheld, 2003).

(30)

confirms the study of Wilson et al. (2009), who states that customers “explanation” of their NPS scores adds unique information and decreases the likelihood for fluctuations in outcomes in general. Therefore, managers must take the underlying motivation of NPS into account when predicting Future Purchase Behaviour upon customers NPS scores. We will explain this phenomenon more in the implication paragraph.

Third, results show that customers who are in a long relationship, purchasing more than one product in different product categories, are more likely to make repeat purchases. This is in line with Verhoef (2013), who suggest that, to increase sales, managers need to build sustainable relationships with its customers. A way to do so is by building loyalty programs for customers who behave more loyal than others. Chmura & Thompson (2015) define a loyalty program as a program or system which rewards customers for their loyal behaviour. Examples of loyalty programs are programs where customers can save for discounts or programs where customers can save for (free) products. In addition to that, managers and academics needs to invest what exactly attracts loyal customers to the company. Important questions here are “what makes them loyal” and “what keeps them loyal” (Wang & Wu, 2012). However, Bolton & Tarasi (2007) stress that some customers do not wish to be in a valuable relationship with online retailers. Perhaps these customers have had a bad experience with the retailer, and therefore prefer to shop in an offline store or do not appreciate the value of relationships at all. It is very expensive to invest in this type of customers and therefore managers need to find the right balance between investing in customer relationships and attracting new customers to maintain the most valuable and profitable customer base (Bolton & Tarasi, 2007).

The last point of discussion relates to the difference between age and gender. Our results show that older customers are less likely to make a purchase in the future than younger customers. This is not in line with the research of Venkatesh & Agarwal (2006) who found a U-shape relationship between age and Future Purchase Behaviour. Furthermore, our results contradict the assumption of Rodgers & Harris (2003) by showing a higher purchase likelihood for female customers than for male customer. Thus, we disagree that male customers show a more positive attitude towards online purchases than female customers. Therefore, to boost future purchase, we advise managers to consider targeting younger female customers instead of older female customers, even though it is contradicting current literature.

5.3 Academic and managerial implications

(31)

literature by providing more information about the interaction effect of the underlying motivation of NPS and the NPS score itself. As a result, this research strongly urges academics to consider the underlying motivation of NPS while examine the effects of NPS. We state that investigating the underlying motivation of NPS is a big benefit with respect to the NPS score, since fluctuation in outcome will be reduced. Furthermore, especially addressing Returns turned out to be strengthening the influence of NPS of Future Purchase Behaviour. Minnema, Bijmolt, Petersen & Shulman (2016) demonstrates that the highest profit for the retailer is generated at a return rate of 13%, at which the costs of product returns outweighed the benefits of increases in future purchases. Moreover, they state that an improved refund speed, helps to improve total relationship value and can increase spending at the retailer after experiencing free return. For that reason, the retailer must keep improving their returning policies to stimulate future purchases. To keep track of the customers’ experience about returns, the retailer can monitor the customers’ NPS scores. Despite the effect of Returns for the investigated retailer, other companies might have other Company Components which effect the predictive value of their NPS scores more.

As main managerial implication we can state that managers need to put more time and effort in analysing customers’ underlying motivation of NPS, which is proved to benefit company’s operations and strategy. Therefore, companies must have the opportunity to obtain this type of information. If the company is able to analyse the underlying motivation of NPS in a correct manner, the company has the opportunity to gain competitive advantage and maintaining customer trust and loyalty.

5.4 Limitations and future research

(32)

Second, the goal of this research was to investigate the influence of NPS on Future Purchase Behaviour and to explore the effect of the Company Components in the underlying motivation of NPS on this relationship. However, since the average NPS does not really differ between customers addressing one Company Components in their underlying motivation or another (see table 3), the answer to what exactly drives a positive or negative NPS was not found. Future research needs to investigate why customers give their high or low NPS scores and how this score can be influenced by the online retailer. To do so, a linear regression model can be used to estimate which Company Components in the underlying motivation are most addressed by which scores. Moreover, another question that arises is; what came first: the NPS score or the experience of the customer with the online retailer? For most customers, the NPS of this study was not based on a one-time purchase but could be influenced by earlier experiences with the retailer. A possible solution could be to exclude customers who made purchases before and thus only considering customers who make a purchase from the online retailer for the very first time. However, future research should give more clarity on this.

Third, a problem arises due to the marginally significant results of the influence NPS on Future Purchase Behaviour. Marginally significance means that the results lay into the direction of supporting, however this is no definite proof. As can be seen in the methodology chapter, most estimated models do show significant results for the influence of NPS on Future Purchase Behaviour, only not the estimated model which was best to use in this research. Therefore, the study needs to be repeated with a different database to see if the marginally significant results of this model change into significant results or no significant results at all.

Fourth, results show a high percentage of customers who make a purchase at the retailer in the next five months after filling in the survey. An explanation for this can be given by the fact that the purchase likelihood of customer is highest in the month December, which is included in the data collection period of this research. In this month both holidays St. Nicolas and Christmas are celebrated in which traditionally many presents are purchased. Therefore, our results could be biased since there is a probability that customers made a purchase in December who in generally would not make the purchase at another moment. Future research needs to replicate this study over another period to see if the purchases return rate will be lower or if the time period did not affect the high purchase return rate at all.

(33)

further research to investigate the effect of combinations of Company Components in the underlying motivation of NPS to see if this choice has effect on our results.

(34)

Literature

Bolton, R., Lemon, K. & Verhoef, P. (2004). The theoretical underpinnings of customer asset management: A framework and propositions for future research. Journal of the Academy of

Marketing Science, 32(3), 271-292.

Bolton, R. & Tarasi, C. (2007). Managing customer relationships. Emerald Group Publishing

Limited, Review of Marketing Research.

Bone, S., Lemon, K., Voorhees, C., Liljenquist, K., Fombelle, P., Detienne, K. & Money, B. (2017). Mere measurement plus: How solicitation of open-ended positive feedback influences customer purchase behaviour. Journal of Marketing, 14(2), 156-170.

Bowman, D. & Navavandas, D. (2001). Managing customer-initiated contacts with manufacturers: The impact on share of category requirements and word-of-mouth behaviour.

Journal of Marketing Research, 38(8), 281-297.

Bulik, B. (2013). How jiffy lube used its net promoter score to goose sales. AsAge, 7(1). Chmura, T. & Thompson, F. (2015). Loyalty programs in emerging and developed markets: the impact of cultural values on loyalty choice. Journal of International Marketing, 23(3), 87-103. Clemente-Ciscar, M., Matias, S. & Giner-Bosch, V. (2014). A methodology based on profitable criteria for defining the partial defection of customers in non-contractual settings. European

Journal of Operational Research, 239(1), 276-285.

Coulter, K. S. & Coulter, R. A. (2002). Determinants of trust in a service provider: The moderating role of length of relationship. Journal of services marketing, 16(1), 35-50.

Day, G. (2004). Creating a superior customer-relating capability. MIT Sloan Management

Review, 44(2), 77-82.

De Haan, E., Verhoef, P. C. & Wiesel, T. (2014). The predictive ability of different customer feedback metrics for retention. Working paper.


(35)

Guo, Y. & Barnes, S. (2011). Purchase behaviour in virtual worlds: An empirical investigation in second life. Information and Management, 48(7), 303–312. 


Hair, J., Black, W., Babin, B. & Anderson, R. (2010) Multivariate data analysis: A global perspective. Seventh ed. Upper Saddle River, New Jersey: Pearson Education, Inc.

Hassan, R., Nawaz, A., Lashari, M. & Zafar, F. (2014). Effect of customer relationship management on customer satisfactions. Procedia Economics and Finance, 23(1), 563-567. Hosmer D. & Lemeshow, S. (1989). Applied logistic regression. Wiley: New York, 10(7), 1162-1163.

Kristensen, K. & Eskildsen, J. (2014). Is the NPS a trustworthy performance measure? TQM

Journal, 26(2), 202-214.

Keiningham, T., Cooil, B., Andreassen, T., Aksoy, L. & Winer, J. (2007). The value of different customer satisfaction and loyalty metrics in predicting customer retention, recommendation, and share-of-wallet. Managing Service Quality: An International Journal, 17(4), 361-384. Keiningham, T., Aksooy, L., Cooil B. & Andreassen, T. (2008). Linking customer loyalty to growth. Mit Sloan Management Review, 49(4), 51-57.

Kumar, N., Kanchan, U. & Gupta, A. (2016). Online shopping behaviour of customers in tier III cities of India. Journal of General Management Research, 4(2), 72-80.

Kumar, V., Petersen, J. & Leone, R. (2007). How valuable is word of mouth? Harvard Business

Review, 10(1), 2-9.

Leeflang, P. & Bijmolt, T. (2013). Modeling markets. Springer, 4(1), 1-10.


Lehmann, D. (2004). Metrics for making marketing matter. Journal of Marketing, 68(4), 73-75.

McGregor, J. (2006). Would You Recommend Us? BusinessWeek, 30(1), 94-96.

(36)

Moe, W. & Fader, P. (2004). Dynamic conversion behaviour at e-commerce sites. Management

Science, 50(3), 326-335.

Morgan, R. & Hunt, S. (1994). The commitment-trust theory of relationship marketing. Journal

of Marketing, 58(7), 20-38.

Morgan, N. & Rego, L. (2006). The value of different customer satisfaction and loyalty metrics in predicting business performance. Marketing Science, 25(5), 426-439.

Padmanabhan, B., Zheng, Z. & Kimbrough, S. O. (2001). Personalization from incomplete data: What you don't know can hurt. Proceedings of the Seventh ACM SIGKDD International

Conference on Knowledge Discovery and Data Mining, 1(1), 154-163.

Qiu, J., Lin, Z. & Li, Y. (2015). Predicting customer purchase behaviour in the e-commerce context. Electron Commercial Research, 91(6), 427-452.

Reichheld, F. (2003). The one number you need to grow. Harvard Business Review, 81(12), 46-55.

Rodgers, S. & Harris, M. (2003). Gender and e-commerce: An exploratory study. Journal of

Advertising Research, 43(3), 322-329.

Rodriquez, M., Peterson, R. & Krishnan, V. (2018). Impact of CRM technology on sales process behaviours: Empirical results from US, Europe, and Asia. Journal of

Business-to-Business Marketing, 25(1), 1-10.

Saris, W. & Revilla, M. (2015). Correction for measurement erors in survey research: Necessary and possible. Social Indicators Research, 127(3), 1005-1020.

Schwarz, G. (1978). Estimating the dimension of a model. Ann. Statist, 6(2), 461-464.

Sheth, J. & Parvatiyar, A. (1995). Relationship marketing in consumer markets: Antecedents and consequences. Journal of the Academy of Marketing Science, 23(3), 255-271.

Sismeiro, C. & Bucklin, R. (2004). Modelling purchase behaviour at an e-commerce website: A task-completion approach. Journal of Marketing Research, 41(3), 306-323.

(37)

Srinivasan, R., Shervani, T. & Fahey, L. (1998). Market-based assets and shareholder value: A framework for analysis. Journal of Marketing, 62(1), 2-18.

Van den Poel, D. & Buckinx, W. (2005). Predicting online-purchasing behaviour. European

Journal of Operational Research, 166(2), 557-575.

Van Doorn, J., Leeflang, P. S. H. & Tijs, M. (2013). Satisfaction as a predictor of future performance: A replication. International Journal of Research in Marketing, 30(3), 314-318. Venkatesh, V. & Agarwal, R. (2006). Turning visitors into customers: A usability-centric perspective on purchase behaviour in electronic channels. Management Science, 52(3), 367-382.

Verhoef, P. (2003). Understanding the effect of customer relationship management efforts on customer retention and customer share development. Journal of Marketing, 67(10), 30-45. Verhoef, P., Franses P. & Hoekstra, J. (2001). The impact of satisfaction and payment equity on cross buying: A dynamic model for a multi-service provider. Journal of Retailing, 77(3), 359-378.

Wang, C. Y. & Wu, L. W. (2012). Customer loyalty and the role of relationship length.

Managing Service Quality: An International Journal, 22(1), 58-74.

Wiebe, J. (1994). Tracking point of view in narrative. Computational Linguistics, 20(1), 233– 287.

Wiesel, T., Verhoef, P. C. & De Haan, E. (2012). There is no single best measure of your

customers. Accessed on 14 April, Achieved from:

http://blogs.hbr.org/cs/2012/07/there_is_no_one_best_measure_o.html.

(38)

Appendix I – Script text mining

Case when t1.Open_antwoorden CONTAINS 'prijs concurrent' THEN 'order' WHEN t1.Open_antwoorden CONTAINS 'concurrerende prijs' THEN 'order' WHEN t1.Open_antwoorden CONTAINS 'internet' THEN 'website'

WHEN t1.Open_antwoorden CONTAINS 'goedkoper concurrent' THEN 'order' WHEN t1.Open_antwoorden CONTAINS 'duurder concurrent' THEN 'order' WHEN t1.Open_antwoorden CONTAINS 'acties' THEN 'order'

WHEN t1.Open_antwoorden CONTAINS 'kortingscode' THEN 'order' WHEN t1.Open_antwoorden CONTAINS 'duur' THEN 'order' WHEN t1.Open_antwoorden CONTAINS 'goedkoop' THEN 'order' WHEN t1.Open_antwoorden CONTAINS 'afgeprijsd' THEN 'order' WHEN t1.Open_antwoorden CONTAINS 'aanbiedingen' THEN 'order' WHEN t1.Open_antwoorden CONTAINS 'zalando' THEN 'order' WHEN t1.Open_antwoorden CONTAINS 'kortingsbon' THEN 'order' WHEN t1.Open_antwoorden CONTAINS 'prijs' THEN 'order' WHEN t1.Open_antwoorden CONTAINS 'actie' THEN 'order' WHEN t1.Open_antwoorden CONTAINS 'zoom' THEN 'website' WHEN t1.Open_antwoorden CONTAINS 'filter' THEN 'website' WHEN t1.Open_antwoorden CONTAINS 'filters' THEN 'website' WHEN t1.Open_antwoorden CONTAINS 'filteren' THEN 'website' WHEN t1.Open_antwoorden CONTAINS 'mobiel' THEN 'website' WHEN t1.Open_antwoorden CONTAINS 'tablet' THEN 'website' WHEN t1.Open_antwoorden CONTAINS 'laptop' THEN 'website'

WHEN t1.Open_antwoorden CONTAINS 'niet overzichtelijk' THEN 'website' WHEN t1.Open_antwoorden CONTAINS 'langzaam website' THEN 'website' WHEN t1.Open_antwoorden CONTAINS 'snel website' THEN 'website' WHEN t1.Open_antwoorden CONTAINS 'app' THEN 'website'

WHEN t1.Open_antwoorden CONTAINS 'foto' THEN 'website' WHEN t1.Open_antwoorden CONTAINS 'pagina' THEN 'website' WHEN t1.Open_antwoorden CONTAINS 'modellen' THEN 'website' WHEN t1.Open_antwoorden CONTAINS 'presentatie' THEN 'website' WHEN t1.Open_antwoorden CONTAINS 'model' THEN 'website'

(39)

WHEN t1.Open_antwoorden CONTAINS 'site' THEN 'website'

WHEN t1.Open_antwoorden CONTAINS 'catalogus' THEN 'customer care' WHEN t1.Open_antwoorden CONTAINS 'mail' THEN 'customer care'

WHEN t1.Open_antwoorden CONTAINS 'e-mail' THEN 'customer care' WHEN t1.Open_antwoorden CONTAINS 'berichten' THEN 'customer care' WHEN t1.Open_antwoorden CONTAINS 'reclame' THEN 'customer care' WHEN t1.Open_antwoorden CONTAINS 'reklame' THEN 'customer care' WHEN t1.Open_antwoorden CONTAINS 'communicatie' THEN 'customer care' WHEN t1.Open_antwoorden CONTAINS 'Factuur bijsturen' THEN 'delivery' WHEN t1.Open_antwoorden CONTAINS 'verzendkosten' THEN 'delivery' WHEN t1.Open_antwoorden CONTAINS 'verzend kosten' THEN 'delivery' WHEN t1.Open_antwoorden CONTAINS 'bezorgen' THEN 'delivery'

WHEN t1.Open_antwoorden CONTAINS 'bezorg kosten' THEN 'delivery' WHEN t1.Open_antwoorden CONTAINS 'bezorger' THEN 'delivery'

WHEN t1.Open_antwoorden CONTAINS 'dhl' THEN 'delivery'

WHEN t1.Open_antwoorden CONTAINS 'te laat bezorgen' THEN 'delivery' WHEN t1.Open_antwoorden CONTAINS 'tijdvak' THEN 'delivery'

WHEN t1.Open_antwoorden CONTAINS 'tijdvenster' THEN 'delivery' WHEN t1.Open_antwoorden CONTAINS 'track' THEN 'delivery' WHEN t1.Open_antwoorden CONTAINS 'trace' THEN 'delivery' WHEN t1.Open_antwoorden CONTAINS 'post nl' THEN 'delivery' WHEN t1.Open_antwoorden CONTAINS 'postnl' THEN 'delivery' WHEN t1.Open_antwoorden CONTAINS 'leveren' THEN 'delivery' WHEN t1.Open_antwoorden CONTAINS 'levering' THEN 'delivery' WHEN t1.Open_antwoorden CONTAINS 'retour' THEN 'returns' WHEN t1.Open_antwoorden CONTAINS 'retourneren' THEN 'returns'

WHEN t1.Open_antwoorden CONTAINS 'terug brengen' THEN 'returns' WHEN t1.Open_antwoorden CONTAINS 'ophalen' THEN 'returns'

WHEN t1.Open_antwoorden CONTAINS 'communicatie retour' THEN 'returns' WHEN t1.Open_antwoorden CONTAINS 'communiceren retour' THEN 'returns' WHEN t1.Open_antwoorden CONTAINS 'retour post nl' THEN 'returns' WHEN t1.Open_antwoorden CONTAINS 'retour postnl' THEN 'returns' WHEN t1.Open_antwoorden CONTAINS 'ruilen' THEN 'returns'

(40)

WHEN t1.Open_antwoorden CONTAINS 'grote maat' THEN 'assortiment' WHEN t1.Open_antwoorden CONTAINS 'grote maten' THEN 'assortiment' WHEN t1.Open_antwoorden CONTAINS 'leverbaar' THEN 'assortiment' WHEN t1.Open_antwoorden CONTAINS 'outlet' THEN 'assortiment'

WHEN t1.Open_antwoorden CONTAINS 'voorraad' THEN 'assortiment' WHEN t1.Open_antwoorden CONTAINS 'kwaliteit' THEN 'assortiment' WHEN t1.Open_antwoorden CONTAINS 'schoenen' THEN 'assortiment' WHEN t1.Open_antwoorden CONTAINS 'kleding' THEN 'assortiment'

WHEN t1.Open_antwoorden CONTAINS 'aanbieden' THEN 'assortiment' WHEN t1.Open_antwoorden CONTAINS 'betalen' THEN 'order'

WHEN t1.Open_antwoorden CONTAINS 'betaling' THEN 'order' WHEN t1.Open_antwoorden CONTAINS 'ideal' THEN 'order' WHEN t1.Open_antwoorden CONTAINS 'geld terug' THEN 'order' WHEN t1.Open_antwoorden CONTAINS 'rente' THEN 'order' WHEN t1.Open_antwoorden CONTAINS 'limiet' THEN 'order' WHEN t1.Open_antwoorden CONTAINS 'krediet' THEN 'order' WHEN t1.Open_antwoorden CONTAINS 'financ' THEN 'order' WHEN t1.Open_antwoorden CONTAINS 'te groot' THEN 'order' WHEN t1.Open_antwoorden CONTAINS 'te klein' THEN 'order' WHEN t1.Open_antwoorden CONTAINS 'tassen' THEN 'order' WHEN t1.Open_antwoorden CONTAINS 'zakken' THEN 'order' WHEN t1.Open_antwoorden CONTAINS 'doos' THEN 'order' WHEN t1.Open_antwoorden CONTAINS 'dozen' THEN 'order' WHEN t1.Open_antwoorden CONTAINS 'verpakking' THEN 'order' WHEN t1.Open_antwoorden CONTAINS 'hersluiten' THEN 'order' WHEN t1.Open_antwoorden CONTAINS 'dicht doen' THEN 'order' WHEN t1.Open_antwoorden CONTAINS 'duurzaam' THEN 'order' WHEN t1.Open_antwoorden CONTAINS 'milieu' THEN 'order'

Referenties

GERELATEERDE DOCUMENTEN

The ‘Structured in-depth email interview for YiPSA volunteers’ (attachment 3) was designed to measure the insights of the current YiPSA arts program, the aim of the YiPSA’s

For covering the costs for controlling certain chemicals and residuals in animals that are going to be slaughtered and meat together with fish, milk and egg, that are done following

voor noodzakelijke individuele aanvullende functionele diagnostiek vanuit de AWBZ als het aangrijpingspunt hiervoor anders is dan waarvoor verzekerde de behandeling in

whatever move John ever made [ec] If it were possible to fully formalize the idea that the John’s every move construction involves a light verb construction embedded in a

In particular, I proposed that receiving negative gossip as well as possessing a high level of anxiety lead to lower mastery and performance approach goals, but lead to higher

Hypothesis 2 is also be proven to be correct as people with the intend to stay long in a hotel room will have a stronger impact on booking probability than users who are

However, the effect is negative indicating that customers with a higher number of PDP visits compared to total pageviews are less likely to make a purchase and spending

The independent variables are: Net Promoter Score t-1, gender, relationship length, customer value in euro’s, moments of truth and the type of product (Water, Boiler,