1
Consumer Preferences in the Telecom Industry:
Understanding the Perception of Youth
Coen Zweers Master Thesis University of Groningen Faculty of Economics and Business
MSc Marketing Intelligence Govert Flinckstraat 48-2 1072EJ Amsterdam Student number: S2484382 E-mail: c.t.zweers@student.rug.nl Tel: +31(0)630701403 Supervisors
2
MANAGEMENT SUMMARY
The telecom market is rapidly changing and the demand for understanding consumer preferences for mobile post-paid plans is increasing. In this paper the effects of smartphone brands, telecom brands, subscription fee, price promotions and additional services on consumer preferences regarding youth are investigated. Paradoxically, the smartphone penetration in de market is expanding, but the telecom companies’ revenues are declining. Approaches are changing from a product-based view into a customer-based view. This study gives valuable insights into the relative importance of certain attributes and attribute levels specifically for the youth segment (up to 24 years old). To gain a deep understanding of the youth and the differences with all other people, a comparison is made by analysing the mass segment (older than 24) simultaneously. The impact of these attributes is known, but the relations between the levels are not. Also, the differences between the segments are not known yet. This research is performed with a choice-based conjoint analysis (CBC). In this study two surveys are used, with in total 2.352 respondents. Specifically, the consumer preferences, relative importance and latent classes are measured.
3
PREFACE
In September 2014 I started the pre-master marketing with a goal, finishing with a master degree. While working in a full time board during my master’s, I managed to pass the exams. Now the time has come to successfully finish my master Marketing Intelligence. During my master I discovered my interest for marketing research. I enjoyed writing my thesis and I am very pleased with the end result. After months of hard work this paper is finished.
This could not be accomplished without the help of several persons that I would like to thank. First of all, my first supervisor: dr. ir. M.J. Gijsenberg for all his feedback, fun meetings and great guidance’s throughout the whole process. Furthermore, I would like to thank dr. J.T. Bouma for all his comments and flexibility. I would like to thank the students that supported me while writing this paper. Finally, I would like to thank all the people that participated in this research.
5
1. INTRODUCTION
The number of smartphone users is continuously increasing. In the Netherlands, there were 10.87 million users in 2015 and this is expected to reach 14.55 million in 2021 (Statista 2016). This growth potential makes the telecom sector a rapidly changing market, with new players entering fast, high competition and instantaneous technological developments. The sector is becoming more and more saturated (Ahn, Ahn, Oh and Kim 2011). In terms of revenue telecom providers in the Netherlands are getting more under pressure. Where the revenue development index was 99,4 in 2007, it is 84,2 at the end of 2015 (Statistics Netherlands 2016). This is a revenue loss of 16 per cent since 2007. The main reason for mobile is because it became cheaper over the last years. In quarter two of 2016 the market dropped 5.6 per cent compared to quarter one in the same year. There is increase of price pressure and tenacious competition. Out-of-bundle revenues become lower and regulation in roaming results in negative influences on mobile service revenue (van Roer 2016). This is a considerable reason why telecom services and technologies are rapidly changing. The mobile telecom competition becomes harsh and it is essential that businesses differentiate continuously. Operators are changing from voice communication services towards value-added services (VAS). These services can create higher average revenue per user (ARPU) for the business (Ahn et al. 2011). Keep in mind that added services are important but is not the only driver that results in sales. Nevertheless, competition increased as a result on focussing on high-value customers (Han, Lu, and Leung 2012). This high-value can be measured by the customer lifetime value that is considered as essential for segmentation base (Donkers, Verhoef, de Jong and Quant Market Econ 2007). Nowadays, telecom companies strive to put more emphasis on the customer focus dived into segments. Segments such as youth, family, mass market and high value. After dividing the market into segments, companies can create better understanding about the differences between and within these segments.
6
segment to identify the differences between these segments. Segmentation increased importance since telecom providers are changing from a product-based view to a customer-based view. Since telecom companies changed from a product-based view towards a customer-based view, thinking in segments originated. Product-based view is based on quality and has a cost emphasis. For the customer-based view, there is an emphasis on the exploration of new goals, strategies, technologies and processes towards the demands of the customer. This is done by focussing on retaining customers and discovering innovations (Rust, Moorman and Dickson 2002). Customer retention and creating loyalty can be accomplished by developing services and products that fit the customer’s demands. This start by accepting that customer’s demands and needs are heterogeneous. The aim of the segmentation depends on the probability of the persons placed together into the segment based on their behaviours and demands that are in line with marketing strategies (Băcilă, Rădulescu, and Mărar 2012). Telecom providers offer two main types of connection types, post-paid and pre-paid services. A service plan that accommodate paying a subscription fee after the usage period is defined as a post-paid customer. A customer who pays for credit before using and is limit to the amount bought is defined as pre-paid (Shrivastava and Israel 2010). The distribution between pre-paid and post-paid is out of balance in terms of revenue. The overall mobile service revenue account for 94.7 per cent of post-paid (van Roer 2016). This research is focused on post-paid customers in the youth segment. The pre-paid revenue percentage is less relevant, since it is very low. To extend this research, it is very interesting to figure out if young people value telecom product and services characteristics differently compared to other segments. Therefore, another sample is made for the mass segment (everyone older than 24 years old). A comparison of consumer preferences between the youth segment and the mass segment is included.
7
group between 18 until 24 years old consists of 1.3 million people (CBS Statline 2016). The smartphone penetration in this group is 96 per cent and monthly spending average is 27 Euros. The mass segment account for all people that are older than 24 years, which are 11 million people (CBS Statline 2016). To figure out the consumer preferences, the optimal characteristics and their effects, the following main research question is answered:
What are the effects of different product and service characteristics for youth on consumer preferences in the consumer mobile telecom market?
From the main research question three sub questions arise:
1. What are the effects of the different attributes (levels) for the youth of mobile post-paid plans?
2. What are the attributes levels of the preferred combinations for the youth within mobile post-paid plans?
3. What are the different consumer preferences between youth and mass segment of mobile post-paid plans?
8
The paper is structured as follows: first, an extensive literature review is executed. The most relevant and important characteristics of telecom services and products are discussed. The literature review ends with a visualized conceptual model. The model explains the relationships between the attributes and the consumer preference. Second, the research design describes the methodology that consists of a choice-based conjoint analyses and utility estimation. Furthermore, it explains how data is collected and describes the plan of analysis. Third, the analyses and empirical data is presented and explained in the results chapter. Fourth, the theoretical and managerial implications, limitations and future research proposals are described. Finally, the paper ends with the references used and the appendices.
2. LITERATURE REVIEW
This chapter consists of a review of existing literature on consumer preferences. Moreover, it presents the literature that is available about relevant characteristics. The review is necessary to provide proper and adequate understanding of the attributes that influence consumer preferences. This research focuses on the relative weights of the attributes and attribute levels. In addition to this research, hypotheses are stated. The relations are visualized in a conceptual model and can be found at the end of this chapter in figure 1.
2.1 Consumer preferences
Consumer preferences are defined as ‘subjective (individual) tastes that are measured
by utility of various bundles of goods’ (Daud 2013, p. 80). A difference between
9
2.2 Determinants of consumer preferences
The elements that influence consumer preferences create a kick-off for investigating the effects and differences among customers. The relevant and relative important attributes consumer preferences in post-paid plans are brand, subscription fee, price
promotion, and additional services. These are relevant explanatory variables in
measuring the effects of consumer preferences and are explained separately using academic literature in the subparagraphs below.
2.2.1 Brand
A brand is a distinguishing name and/or symbol (such as logo, trademark, or package design) intended to identify the goods or services of either one seller or a group of sellers, and to differentiate those goods or services from those of competitors.
10
on consumer preference provides interesting thoughts to see what the effect would be within this study on smartphone and telecom brands. For this reason, the following hypotheses are tested:
H1: There is stronger positive effect with Samsung Galaxy as a brand on the consumer preferences of youth for mobile phone post-paid plans compared to Apple iPhone.
H2: There is a stronger positive effect of Company A as a telecom brand on consumer preferences of youth for mobile phone post-paid plans compared to Company B. 2.2.2 Subscription fee
Another interesting thought is the subscription fee. This is the price that is paid after each period within the subscription. Elements of price are defined as ‘the value of
items which are needed for the acquisition of a product or service (Al-Dmour, Zu'bi
and Kakeesh 2013). If a customer is able to purchase goods or can afford an subscription, this does not mean directly that it is liked (Daud 2013). The satisfaction of a customer is indirectly influenced through price fairness and directly influenced by the price perception (Malik et al. 2012). Literature describes that price is the most important preferential factor in choosing the telecom communication provider in the context of the Indian market (Paulrajan and Rajkumar 2011). Findings show similarities, but these are from a totally different market. Telecom communication providers use subscriptions to retain customers. In a subscription plan customers pay a fee after each period, and by doing so they can use their product and service until the contractual period is reached. Subscriptions help to gather and maintain customer data. Companies can set accurate pricing strategies that can result in customer renewing contracts and to stay as a customer (Penmetsa, Gal-Or and May 2015). The consumers’ perception and the actual price of products as an object deliver consumers meaning. This can be perceived as inexpensive, moderate or expensive. These perceptions are linked to the subjective image that has effect on the consumer’s evaluation of products (Sharma and Garg 2016). Consumers evaluate low-quality products with a low price less negative, than low-quality products with a high price (Gneezy et al. 2014). Through the importance of pricing from existing literature and relevance for the consumer, the following hypothesis is formed:
11 H3: Higher subscription fees have a negative effect on the consumer preferences of youth for mobile phone post-paid plans.
2.2.3 Price promotion
12
et al. 2014). For this reason, the following hypothesis is suggested:
H4: Stronger price promotions have a stronger positive effect on the consumer preferences of youth for mobile phone post-paid plans.
2.2.4 Add-ons
Besides the basic elements of telecommunication (calling, texting and data usage), other services can be offered as well as a bonus or as a bundle. Add-ons are products or services that can function complementary to the base product. Consumers can have greater need for additional products when the base product is purchased (Erat and Bhaskaran 2012). The base product and the add-ons can also be offered at the same time. In service or product bundling, different service or products are offered as one. This way the willingness to pay can be influenced (Morrison 2016). As mentioned before, the telecom market is changing from an operational perspective towards a customer-driven way of approaching the market. Reducing costs and on the same time optimizing the life cycle, is a huge challenge for operators. Converged products give an outcome for enough revenue-generating competitive services (Huang, Lee, Crespi 2012). Operators combine subscriptions with extra services like music subscription services (Spotify) and movie subscription services (Netflix). Telecom providers have post-paid plans that include these bundled services on subscriptions for music, movie or potentially an extra data package of one gigabyte. The influence of these extra services has not been studied for youth within the telecom market. Therefore, the following hypothesis is suggested:
13 Consumer preferences Smartphone brand Telecom brand Subscription price Price promotion Add-ons Attributes H1+ H2+ H3- H4+ H5+ 2.3 Conceptual model
Based on the literature review, a conceptual model is used to visualize the attributes and interrelationships of the aforementioned and hypotheses, and is presented in figure 1. The core of this study lies on the relative weights of the attributes and attribute levels. The attribute levels will be explained in the research design.
Figure 1: Conceptual model
3. RESEARCH DESIGN
After a deep dive in the existing literature, this chapter will explain the research design and methodology. Firstly, the research methodology is described. This consists of the data collection, attributes (levels) and control variables. Secondly, the study design is scrutinized. Finally, the analysis and data preparation methods are expressed.
3.1 Research methodology
14
(Eggers and Sattler 2011). The most preferred choices, based on utilities from attribute (levels) are considered the consumer preference.
Data collection
As mentioned before, the dataset is collected by creating a choice-based conjoint survey that is aimed at existing customers in a customer panel of a telecom provider. It includes several demographics and choice based questions wherein respondents chose the most preferred option. The choice based survey is created using my.preference.com, which is a website made for choice based conjoint surveys (Eggers and Sattler 2011). Using a panel of a telecom company two survey samples are created, aimed on the segments youth and mass. Both samples consist of 20.000 customers. This is done based on the demographic age (youth is 18 until 24 years old, mass is 25 years and older). Respondents were able to fill in the questionnaire from November 16th till 26th, 2016. This is an eleven-day period. All survey questions can be found in the appendix A.
Attributes and levels
In this paragraph levels of each attribute are described. These are listed and can be found in table 2.
Attribute Attribute levels
Smartphone brand Apple iPhone 7, Samsung Galaxy S7, Sony Xperia XZ Telecom brand Company A, Company B, Company C
Subscription fee (per month) € 35,- , € 40,- , € 50,- Price promotion (discount) 0%, 20%, 30%
Additional services Music (Spotify), Movies (Netflix), 30% extra data Table 2: Input conjoint analysis
15
these are big companies in the mobile market. The subscription fee is the post-paid fee per month. This varies between 35 and 50 Euros. Price promotions vary between zero per cent discount, 20 per cent discount and 30 per cent discount. The discount is limited to 30 per cent. This way it does not affect the price-quality inference for customers. The add-ons consist of free subscriptions on music or movie streaming services or 30 per cent extra data. Prices are set according to the market prices for the mentioned smartphones and providers in 2016. It accounts for all choices that it is a post-paid plan with a device that includes unlimited calling, five gigabytes and free calling in the Netherlands and EU.
Control variables
Several demographical questions are included to the questionnaire to get good understanding of the respondents. These function as control variables in the sample. This way every respondent provided their gender, educational level, age, employment, living situation and if they posses a smartphone.
3.2 Study design
Existing customers of a telecom company received an email with an invitation link to the online questionnaire. Respondents were explained to make choices based on the attributes (levels). The attributes ad attribute levels are randomized and a force response is required in this design. After the different choice sets the questionnaire ends with questions regarding demographics of the respondent. To increase the respondents, gift cards are raffled on all participated respondents. An example of the choice set looks as follows:
Choice A Choice B Choice C
Apple iPhone 7 Samsung Galaxy S7 Sony Xperia XZ
Company B Company A Company C
€ 40,- per month € 50,- per month € 35,- per month
30% discount 20% discount 0% discount
Free music subscription (Spotify)
Free movie subscription (Netflix)
Extra data package 30%
O O O
16
An alternative to the forced response model is a base model. A base model can be added to benchmark the post-paid plans. This can function as a none-choice option. This could be added as well if the respondent is not interested in the post-paid choices at all and prefers prepaid. It is not included for the fact that respondent could avoid difficult answers by clicking only on the none-choice option. Furthermore, most customers have a post-paid contract. Also price is not the primary interest in this study. For these reasons, the none-choice option is left out and will also not be included in the estimation model (Eggers and Sattler 2011). Instead, a separate question is added to check if respondents actually buy the choice they picked in real life. This is not included in the estimation. Tests of the survey ease of use are performed to figure out the correct amount of choice sets. The determination of the optimal choice design is a complex research issue (Eggers and Sattler 2011). In the context of this research five attributes, including three levels, create (3 x 3 x 3 x 3 x 3) 243 different possible mobile post-paid plan combinations. The complexity is increases by adding the different choice sets. After testing the survey with a test group, eight different choice sets are created. Each choice set consists of three options. This generates huge data, but it is still convenient for the respondents to fill in the survey.
Mass segment
As described in the introduction, another way to benchmark the findings of the choice-based conjoint and to extend the research is by comparing the youth sample with a mass sample. This segment implies all people older than 24 years old. The same choice-based conjoint analysis is performed for this segment. The differences of the effects and relative importance are written down in the results and discussion chapter.
Data preparation
17
since the coding of the values will be correct this way. Furthermore, the exported data sets still consist of irrelevant data like respondents pass, IP address, entry date, elapsed time and email addresses that need to be removed. Than using, IBM’s statistical software, SPSS®, the demographics and choice set data are merged again linking it to the respondent’s id. The demographics are also provided with values to make the analysis more comprehensible and opportunities to create different latent classes. The remaining data in the final sample is used for the estimation.
3.3 Analysis
For the analysis the utility is estimated for each attribute. To do so, an estimation of the value for each respondent and attribute is performed. The equation that meets this estimation is as follows:
The utility (U) of all respondents of the consumer preferences (j), is the sum of the utilities (β) of the different attribute levels (X; k=1,...K). The conjoint data is analysed using the software LatentGold®. This is statistical software specialized in choice based conjoint analysis.
The analysis that suits the conjoint analysis is the multinomial logit. The dependent variable can exhibit multiple states. The chosen option can be any alternative from choice set J = {1, …, i, …, m}. It is the choice of alternative i from choice set J. The multinomial logit model gives the following probability function:
18
table. The largest differences between part-worth’s utilities within an attribute are calculated.
Goodness of fit comparison
To check for model fit, multiple fit checks are performed. The goodness of fit of the part-worth model is compared with a linear model. The linear model consists of a numeric attribute, price. All other variables are nominal and hence part worth. The range of likelihood values is measured for the linear and part-worth model. This is the maximum if β’s fit observed choices. Furthermore, the likelihood ratio test explains if the linear estimated model is better than the part-worth model. The estimated model parameters need to be significantly different from zero.
Another fit test that is applied includes the The McFadden (adjusted) Pseudo-R2 . It is the sum of squared errors form the model. It show how much the data is explained by the model. The likelihood is between 0 and 1. The magnitude will be larger of the log of the likelihood, if the model contains a very low likelihood compared to the log of a more likely model. All statistics are in-sample, which means that there is an estimation of the model by using all the data that is available. The model’s fitted values are compared to actual realizations. The hit rate is calculated by comparing the choice predicted by the model, including part-worth estimations, for each individual by making use of the maximum utility. A correct prediction is a hit, total hits devided by total sample size is the hit- rate (Akaichi et al. 2013). The goodness of fit of the predictive validity can measured by the mean absolute error (MAE). This is done by dividing the predicted shares with the observed shares. Lastly, a goodness of fit comparison is performed between the null-, part-worth and linear model.
Latent Class analysis
19
mixture models. A few steps are taking in this analysis. Firstly, prior segmentation defines segments based on demographic information. This is a more explorative approach of finding similarities in preferences. Secondly, mixture models assume that utilities are distributed across consumers. Finite mixture models, also called latent classes, explain discrete classes that differ in preferences. It is a posterior segmentation. Respondents are allocated into classes based on calculated probabilities. In this study, the finite mixture models ‘Latent classes’ are estimated. The finite mixture model creates the added value by offering a model-based clustering that derives clusters by using a probabilistic model. This is described by the distribution of the data and therefore used in this study. Finally, the estimation of models for several numbers of classes is made. The best fit is chosen according to log-likelihood-based measures, information criteria BIC and CAIC since these punish the strongest for adding parameters, classification error and interpretability. To identify the effects of demographics in the prior segmentation it is decided not to include demographics as covariates for estimating latent classes. This way the latent classes are estimated focussing on the probabilistic models of the data without support of the demographics.
4. RESULTS
In this chapter the results of the conjoint analysis are described. Firstly, the demographics of the samples are presented. Secondly, the relative attribute importance is measured and this indicates the influence of each attribute on consumers’ choices. These are the main effects of the choice-based-conjoint analysis. Finally, the latent classes of both youth and mass sample are created and differences are studied.
4.1 Demographics
20
Table 4: demographics
The youth sample consists of 1039 respondents and the mass sample 1313. 99% of youth has a smartphone compared to 98% for mass. The majority of youth is female (52%), for mass there are more males (59%) than females. Most respondents under youth are 23-24 years old (32%). For the mass sample, most respondents are between 36-45 years old (28%). In the youth sample most respondents finished a MBO study (45%) and the respondents in the mass sample are slightly higher educated with an HBO diploma (39%). Important to mention, most younger people are still attending their studies, whereas older people have finished their studies already. In the youth
Youth (n=1039) Mass (n=1313)
Frequency Percentage (%) Frequency Percentage (%)
Smartphone Yes 1035 99.6 1286 97.9 No 4 0.4 27 2.1 Gender Male 496 47.7 778 59.3 Female 543 52.3 535 40.8 Age Younger than 16 0 0.0 16-18 39 3.8 19-20 187 18.0 21-22 329 31.7 23-24 419 40.3 Older than 24 65 6.3 Younger than 25 21 1.6 25-35 304 23.2 36-45 364 27.7 46-55 313 23.8 56-65 200 15.2 Older than 65 111 8.5 Education Secondary school 122 11.7 98 7.5 MBO 466 44.9 461 35.1 HBO 302 29.1 507 38.6 WO-bachelor 83 8,0 56 4.3 WO-master 49 4.7 163 12.4 None of these 17 1.6 28 2.1 Living situation Live alone 194 18.7 258 19.7 With my parents 491 47.3 45 3.4 With a family/partner 242 23.3 974 74.2 With friends/other family/unknown people 112 10.8 36 2.7 Work situation Go to school/study 474 45.6 34 2.6 Employed 456 43.9 876 66.7
Search for a job 55 5.3 66 5.0
Self-employed 54 5.2 180 13.7
21
sample most respondents live at their parents (47%) and for the mass sample most people live together with a family/partner (74%). Finally, in the youth sample most people go to school/study (46%); the mass sample consist mostly of people that are employed (67%).
4.2 Choice-based conjoint effects
In this paragraph the effects of the choice-based conjoint analysis is presented. The relative importance of all the attributes and attribute levels give different observations. All the differences are compared within and between the samples.
4.2.1 Main effects model
A comparison in log-likelihood between the null model and the main part-worth effects models are made. As can be seen for youth (Chi-square difference of 3006,39) and mass (Chi-square difference of 4009,28), the estimated part-worth models can make better predictions. To measure the effect-coded preferences all attributes need to be made nominal. This way the attributes are assumed as a part-worth model and ready for the estimation. All variables that are included can be coded in a numeric or nominal way. One of the attributes is subscription fee, which makes it interesting to test as numeric variable compared with a nominal specification. The goodness of fit measurements presented in table 5.
Table 5: Goodness of fit comparison
The log-likelihood and R-squared (R2) improves by adding parameters. The Log-likelihood (LL) of the part-worth models for youth (LL= -7628,47) and mass (LL=19070,37) are smaller than the log-likelihood of the linear models. The same counts for the R2.
A measurement that does punishes for adding parameters is the adjusted R2. Although the penalty for having an extra parameter, the adjusted R-squared (R2 adj) seem to be Models
Npar df p-value
R2 R2
adjusted
LL L2 AIC BIC Hit
rate
Youth Null model -9131,6653 33,33%
Youth part-worth 10 1029 <0.001 0,1672 0,1671 -7628,4727 15256,9454 15276,9454 15326,4056 55,03% Youth linear 9 1030 <0.001 0,1669 0,1669 -7629,9829 15259,9659 15277,9659 15322,4800 54,87%
Mass Null model -11539,8234 33,33%
22
almost the same. For youth, the part-worth model (R2 adj= 0,1671) is slightly better than the linear model (R2 adj = 0,1669). Therefore, the youth part-worth model gives a better representation of the data. For mass, both part-worth and linear model are the same (R2 adj=0,1724), so it does not matter which one to choose.
The log-likelihood ratio test can explain if the numeric vector model model is significantly different from the nominal part-worth model. This is calculated as follows: X2 = -2 (LLlinear – (LLpart-worth)). Than the difference between de degrees in freedom (part-worth – linear) is calculated. The p-value is calculated using a internet chisquared calculator. For youth, Chi-squared (X2) = -2*7629,9829 – (-7628,4727)) = 3,0204. The difference in degrees of freedom is one. Ultimately, there is no significant difference between the two models (p-value = 0,0822). For mass, X2 = -2*(-9535,7708 – (-9535,1832)) = 1,1752. Difference in degrees of freedom is also one. After calculating the probability (p-value = 0,2783), there is no significant difference between the mass models. From the log-likelihood ratio test this means the models are the same and both can be used.
Another measurement that penalties for adding parameters are the Akaike information criterion (AIC) and Bayesian information criterion (BIC). BIC punishes even stronger than AIC. The lower the scores, the better the model. A comparison between the youth part-worth and linear specifications explain a better fit for the part-worth model (15276,9454). For the BIC the scores the linear specification is a better fit (15322,4800). Under mass, the linear model show a better score under AIC (19090,3665) and BIC (19142,1672) compared to the mass part-worth.
23 4.2.1 Main effects model results
Table 6 shows an overview of all the estimations regarding the attributes, attribute levels and a comparison between the youth and mass samples. As can be see in the table, all attributes are significant (p<0,001) and no attributes need to be dropped.
Table 6: part-worth estimations
A visualized version of the part-worth results can be found in figure 3. The largest differences between youth and mass can be found for subscription fee and price promotions. All effects are described at paragraph 4.3 hypotheses testing. A clustered line diagram of the part-worth effects, are visualized in appendix B. A clear distorted image of Company B customers can be found on telecom brands.
Youth Mass
Attributes Utility S.E. Z-value P-value Wald Utility S.E. Z-value P-value Wald
24
Figure 3: Part-worth attribute estimation
4.2.2 Relative importance
25
Figure 4: Relative importance
4.3 Hypotheses testing
All attribute levels are estimated by performing the choice-based conjoint analysis. The path-worth models for youth and mass are estimated and can be found in table 6. All attributes are highly significant (p<0,01). For the attribute smartphone brand, it turns out to that the Samsung Galaxy S7 has a weaker positive effect on the consumer preferences for youth (β = 0,1882) and mass (β =0,2525) compared to the Apple iPhone 7. Therefore, it can be stated that hypothesis one is not supported. It seems that Apple iPhone 7 has a much stronger significant positive effect on the consumer preference for youth (β =0,5599) and mass (β =0,5560). Sony Xperia XZ seems to have a strong negative effect for both youth (β =-0,7481) and mass (β =-0,8085). For telecom brand, Company A has a small negative effect on the consumer preferences for youth (β = -0,0827) and mass (β = -0,0863). Company B seem to have a positive effect on youth (β =0,3199) and mass (β = 0,3611). It is an opposing outcome. For this reason, hypothesis two is not supported. Company C has a negative effect on the consumer preferences for youth (β = 0,2372) and even stronger for mass (β = -0,2748). Since a sample is used with Company B customers, there might be a problem of endogeneity. For this reason the outcome is slightly biased. Furthermore, there is a significant negative effect of subscription fee. The highest fee of 50 euro has a negative effect for youth (β = -0,2372) and even stronger negative effect for mass (β = -0,4199). Hypothesis three is there for supported. For price promotions, there is a significant positive effect for youth (β =0,1228) and mass (β =0,0950). Also hypothesis four is supported. Finally, It turns out that additional services in terms of
40% 17% 27% 9% 8% 42% 20% 26% 6% 6%
Smartphone brand Telecom brand Subscription fee Price promotion Additional services Relative importance
26
music subscriptions give a negative effect for youth (β = 0,0172) and mass (β = -0,0389) on the consumer preferences. For this reason, the last hypothesis is not supported.
4.4 Latent class analysis
This paragraph consists of the prior segmentation, whereas the demographics are estimated among the data. Followed by the latent class results that are found performing a probabilistic latent class analysis.
4.4.1 Prior segmentation
All data of both samples, youth and mass are divided in prior segments. Consumers belong to discrete segments that differ in preferences. For each demographic variable a prior segment is estimated for youth and mass. The main results within and between these classes are described. Only the classes that have a significant differences (p<0,05) are considered. All detailed scores can be found in the tables and figures in appendix C.
Gender
There is a significant difference in preferences between males and females for youth. There is a stronger positive effect on females (β = 0,6401) for the Apple iPhone 7 compared to males (β = 0,4758). Furthermore, females (β = 0,3563) prefer the telecom brand Company B rather than males (β = 0,2824). Lastly, females (β = 0,4553) prefer 35 Euro subscription fee slightly more compared to males (β = 0,3873). These results explain that females are more critical.
Education
27
Low educated people prefer Samsung rather than the Apple iPhone 7. WO-master educated people are slightly more price sensitive under youth (β =0,4163) and mass (β = 0,5448) compared to people educated with a secondary school degree under youth (β = 0,6271) and mass (β = 0,3934).
Living situation
Overall, the Apple iPhone 7 is most preferred for each living situation under youth. Nevertheless, it is interesting to see that the Samsung Galaxy S7 is preferred by people that live together with their partner or family (β = 0,5526) and least preferred by people that live on their own (β = 0,6083).
Age
For both segments, the preference for an Apple iPhone 7 decreases over age. For youth, people that are older than 24 years old (β = 0,3736) results in a lower positive effect than of people that are between 19-20 years old (β = 0,6904). For mass, people older than 65 years (β = 0,2999) show a much lower positive effect than people between 25-35 years (β = 0,7098). For mass, the preference for a Samsung Galaxy S7 increases over age. People older than 65 years show a higher positive effect (β = 0,3474) than people between 25-35 years (β = 0,1808). For youth, people aged between 19-20 years (β = -0,8711) have a higher negative effect on Sony Xperia XZ compared to people older than 24 years (β = -0,5674). For mass, people between 56-65 years (β = 0,466) have a slightly higher positive effect on the telecom brand Company B, compared to relatively young people between 25-35 years (β = 0,3092). The effect increases over age. For youth, there is a relatively strong positive effect for people between 21-22 years for 30 per cent extra data on the consumer preference (β = 0,2167). Additional services are preferred most for people between 21-22 years old. Young people are relatively sensitive for discounts on consumer preferences. These effects range for 30 per cent discount and consist of a small difference in effect (β =
0,169).
Work situation
28
people that are searching for work highly prefer Samsung Galaxy S7 (β = 0,6923) instead of an Apple iPhone 7 (β = -0,2748). Surprisingly, young people searching for work show even a negative result for the Apple iPhone 7. For mass, the effect is the opposite around. There is a higher positive effect on the Apple iPhone 7 (β = 0,4773) compared to the Samsung Galaxy S7 (β = 0,2784). People that are searching for work, in the mass sample, have a positive effect on the telecom brand Company B (β = 0,6233) and a most negative effect on Company A (β = -0,4179). For youth, people that go to school or study are most sensitive for prices. There is a strong positive effect for 35 Euro (β = 0,508) and a strong negative effect on 50 Euro (β = -0,532). The additional services attribute level, 30 per cent extra data, has the strongest positive effect on retired (β = 0,2766) and self-employed (β = 0,2239) people for the mass.
4.4.2 Latent classes
In LatentGold®, the statistic software used for finding the latent classes, eight different classes are estimated for the youth and mass sample.
Goodness of fit
29
Youth LL BIC(LL) CAIC(LL) Npar p-value Class. Err. R² adj. R² 1-Class Choice -7628,4727 15326,4056 15336,4056 10 <0,001 0 0,1671 0,1672 2-Class Choice -6420,0158 12985,8979 13006,8979 21 <0,001 0,0273 0,4052 0,4053 3-Class Choice -6010,4245 12243,1215 12275,1215 32 <0,001 0,0321 0,4832 0,4831 4-Class Choice -5794,9908 11888,6601 11931,6601 43 <0,001 0,0828 0,5495 0,5495 5-Class Choice -5648,3335 11671,7517 11725,7517 54 <0,001 0,1004 0,5938 0,5938 6-Class Choice -5534,9469 11521,3847 11586,3847 65 <0,001 0,1068 0,618 0,618 7-Class Choice -5472,6801 11473,2572 11549,2572 76 <0,001 0,1093 0,636 0,636 8-Class Choice -5419,0301 11442,3635 11529,3635 87 <0,001 0,1052 0,6544 0,6544 Table 7: Youth goodness of fit latent classes
Mass LL BIC(LL) CAIC(LL) Npar p-value Class. Err. R² adj. R² 1-Class Choice -9535,1832 19142,1672 19152,1672 10 <0,001 0 0,1724 0,1724 2-Class Choice -7609,8177 15370,4169 15391,4169 21 <0,001 0,0205 0,4511 0,4511 3-Class Choice -6984,1232 14198,0086 14230,0086 32 <0,001 0,0425 0,5555 0,5555 4-Class Choice -6616,5768 13541,8967 13584,8967 43 <0,001 0,0664 0,6337 0,6337 5-Class Choice -6431,4783 13250,6805 13304,6805 54 <0,001 0,0718 0,6647 0,6647 6-Class Choice -6252,5259 12971,7564 13036,7564 65 <0,001 0,085 0,7001 0,7001 7-Class Choice -6150,8501 12847,3855 12923,3855 76 <0,001 0,0854 0,7179 0,7179 8-Class Choice -6088,4103 12801,4866 12888,4866 87 <0,001 0,0978 0,7301 0,7301 Table 8: Mass goodness of fit latent classes
These measurements show at slight gap difference between class 5 and 6 compared to other classes, but the differences are not really big. The strongest punishers for adding parameters are the BIC and CAIC. Also here both scores steady decline after adding more classes for youth and mass. As can be found in figure 5 and 6. In class 1 all the different have almost the same values. To get a good understanding of the differences in the other classes, the figures are scaled in a way that it presents the last classes the best. 10800 11000 11200 11400 11600 11800 12000 1 2 3 4 5 6 7 8 LL Classes
Youth sample fit - Information criteria
30
Figure 5: Information criteria youth sample
Figure 6: Information criteria mass sample
Classification error and class sizes
By setting up the estimation, the classification posterior is included. It explains the probability of each class. The classification error is the average minimum probability across all cases. Another indicator of the relevant segments is the class sizes. Through the latent class estimation all respondents are allocated to classes. For identifying relevant findings, the minimum percentage of respondents that are in a class, should be around ten per cent. Five classes is the optimal amount of classes for both youth and mass. The respondent and class sizes can be found in table 9.
Class 1 Class 2 Class 3 Class 4 Class 5 Class 6 Class 7 Class 8 Total Youth Class respondents 289 246 126 130 114 62 47 25 1039 Youth Class Size (%) 27.11 23.39 12.77 12.48 11.83 5.21 4.79 2.42 100 Mass Class respondents 345 276 214 137 135 123 46 37 1313 Mass Class Size (%) 26.00 21.42 15.83 11.06 9.83 9.38 3.5 2.98 100
Table 9: class sizes
Class estimation
The five classes that remain are the selected classes of the studied samples. What stands out in the analysis on the attribute effects is allocated to a class label. The estimations of the youth sample are presented followed by the mass estimations. All effects can be found in table 10 and 11.
12200 12400 12600 12800 13000 13200 13400 13600 13800 14000 1 2 3 4 5 6 7 8 LL Classes
Mass sample fit - Information criteria
31
Youth Class 1 Class 2 Class 3 Class 4 Class 5 Attributes Average
customer Apple fans Loyal telecom customers Samsung Fans Low budget Wald(=) p-value Smartphone brand Apple iphone 7 0,9173 4,0414 0,2683 -1,4786 -0,7472 586,575 <0,001 Samsung Galaxy S7 0,2224 -1,5024 0,29 3,0114 0,6349 <0,001 Sony Xperia XZ -1,1398 -2,539 -0,5583 -1,5328 0,1123 <0,001 Telecom brand Company A -0,0003 -0,622 -0,4069 -0,4021 0,0601 210,7715 <0,001 Company B 0,1383 1,019 1,5804 0,5917 0,3716 <0,001 Company C -0,138 -0,397 -1,1735 -0,1896 -0,4317 <0,001 Subscription fee € 35,- 0,721 0,2267 0,2952 0,2012 1,2312 181,4161 <0,001 € 40,- 0,3163 -0,2253 0,0492 0,2379 0,4843 <0,001 € 50,- -1,0373 -0,0014 -0,3444 -0,4391 -1,7155 <0,001 Price promotion 0% -0,2766 -0,9492 -0,3173 -1,1311 -0,5041 28,385 <0,001 20% 0,1039 0,2187 0,1123 0,4566 0,1308 <0,001 30% 0,1727 0,7306 0,205 0,6745 0,3733 <0,001 Additional services Music (spotify) 0,1273 0,1949 0,0556 0,1113 -0,3403 128,0061 <0,001 Movie (Netflix) -0,1897 -0,3052 -0,2494 -0,2936 0,0597 <0,001 30% extra data 0,0624 0,1103 0,1938 0,1823 0,2805 <0,001
Table 10: Youth latent classes
Youth
32
expected negative effect on the Sony Xperia XZ (β =-1,5328). Class five is labelled as “Low budget”. Customers in this class show a strong negative effect on 50 Euro subscription fee (β =-1,7155) and high positive effect for 35 Euro (β =1,2312). Overall, all other effects are moderate. Nevertheless, there is a rather strong negative effect on the Apple iPhone 7 (β =0,7472). Low budget customers might assume that the Apple iPhone 7 is more expensive than other smartphone brands.
Mass Class1 Class2 Class3 Class4 Class5 Attributes Apple
fans Average customer Samsung fans Moderate budget Low budget Wald(=) p-value Smartphone brand Apple iPhone 7 3,5774 0,9746 -0,9656 -0,821 0,3183 1033,9276 <0,001 Samsung Galaxy S7 -1,7661 0,3077 2,5294 1,0377 -0,1538 <0,001 Sony Xperia XZ -1,8113 -1,2822 -1,5637 -0,2167 -0,1645 <0,001 Telecom brand Company A -0,2171 0,0999 -0,2245 -0,1921 -0,006 186,2449 <0,001 Company B 0,6455 0,434 0,6094 0,5907 0,2367 <0,001 Company C -0,4284 -0,5339 -0,3849 -0,3985 -0,2307 <0,001 Subscription fee € 35,- 0,4941 0,7086 0,3075 1,1673 3,2989 176,5761 <0,001 € 40,- -0,2672 0,2581 0,1917 0,2822 0,8547 <0,001 € 50,- -0,2269 -0,9667 -0,4992 -1,4494 -4,1536 <0,001 Price promotion 0% 0,1076 -0,3369 -0,1818 -0,3359 -0,4223 30,225 <0,001 20% -0,1515 0,125 -0,0702 -0,0256 0,256 <0,001 30% 0,0439 0,2119 0,252 0,3615 0,1663 <0,001 Additional services Music (spotify) 0,2581 0,0952 -0,034 -0,2969 0,1833 106,3336 <0,001 Movie (Netflix) -0,3436 -0,1384 -0,0126 -0,1342 -0,1509 <0,001 30% extra data 0,0856 0,0432 0,0465 0,4311 -0,0324 <0,001 Table 11: Mass latent classes
Mass
33
positive effect on the Samsung Galaxy S7 (β =2,592) and negative effects towards the Apple iPhone 7 (β =-0,9656) as well as the Sony Xperia XZ (β =1,5637). Class four is labelled as the “Moderate budget” class. Subscription fee seems to be most important since the 35 Euro fee has a strong positive effect (β =1,1673) and an even stronger negative effect towards 50 Euro (B=-1,4494) on the consumer preference in this class. There is a positive effect for the Samsung Galaxy S7 (β =1,0377). Customers might assume this telecom brand is cheaper. The fifth class represents the “Low budget” class. It shows a very strong positive effect to the 35 Euro subscription fee (β =3,2989) and an even stronger negative effect on the 50 Euro subscription fee (β =4,1536). These effects are much stronger compared to the Low budget class for youth. Furthermore, this class shows small positive and negative effects on the telecom brands.
Overall, for both samples five classes are estimated. For youth, the average customer has the biggest class size and there is a loyal group of customers towards Company B. For mass, the Apple fans represent the biggest class size. Instead of a loyal class, a “ moderate budget class” is found for mass. The preference of a low price mobile post-paid plan for Mass is even stronger for the low budget class compared to youth.
5. DISCUSSION
The goal of this study is to examine the consumer preferences for the youth segment in telecom post-paid plans in the telecom industry. The main and sub research questions are answered. These findings are linked with the theoretical and managerial implications. Firstly, the main effects of the post-paid attributes and the relative importance of these attributes are discussed. Secondly, a prior segmentation is discussed. Lastly, the latent classes are evaluated.
5.1 Theoretical implications
34 Smartphone brand
Smartphone brand has a significant influence on the consumer preferences. Although the Samsung Galaxy S7 is stated in the hypotheses and global market leader followed by Apple, this study shows a clear preference for the Apple iPhone 7. As described by Keller (2014), brands are intangible assets and important to manage. This is most definitely approved by the results of this research. It shows that Apple brands its products much better than expected and described in literature. According to the relative importance, this smartphone brand is most important varying between 40-42 % for youth and mass. One of the main differences between youth and mass lies in the preference for the Samsung Galaxy S7 and the Sony Xperia. Moreover, youth prefer the Samsung Galaxy S7 slightly less than mass. For the Sony Xperia XZ it seems youth and mass do not prefer this brand. Sony’s brand image gives clearly less added value in the thoughts of consumers (Leone et al. 2006), compared by the other two giants. Overall, the hypothesis (1) for the Samsung Galaxy is supported and links to the literature. Nevertheless, Apple iPhone 7 is the key element through the eyes of the consumers and in line with the literature of Zahid and Dastane (2006).
Telecom brand
35
supported. This outcome is biased because of the fact that a Company B customer base is used. The main difference shows that people from the mass segment prefer Company B even more and do not prefer Company A in a more intense way compared to the preference of youth.
Subscription fee
The price is the value of items that is needed for acquisition for products and services (Al Dmour et al. 2013). This is true for mobile post-paid plans. Price turns out to be an important driver for consumer’s decision making aimed at mobile post-paid plans. For both youth and mass, the higher the subscription fee, the less likely a consumer prefers the mobile post-paid plan. The hypothesis (3) explain a negative effect through subscription fees, is supported. Mobile post-paid plans of 50 euro are too expensive proven by the negative effect on the consumer preferences. However, 40-euro subscription fee positively affects the consumer preferences for youth. Assuming younger people have less money than older people, it is in line with the existing literature of Daud (2013). If people are able to afford a subscription fee, it does not necessarily implicit that it is preferred. Young people prefer a subscription fee of 40 Euros instead of not preferring it like the mass group. This is a clear difference. Nevertheless, the relative importance of subscription fees between youth and mass are very similar. Besides the intense preferences for a smartphone brand, subscription fee is determined second most important. Subscription fee, in the three price levels can be perceived as relatively inexpensive, moderate or expensive. This is related to the subjective image of products that are being evaluated by consumers (Sharma & Garg, 2016). This study reveals clear difference in price perceptions between youth and mass. It can also be stated that through these results youth perceive certain products that are of high quality less negative than others with lower quality (Gneezy et al. 2014).
Price promotion
36
effect of price promotions on the consumer preferences for youth and mass. For this reason, hypothesis 4 is accepted. It can have impact with a smaller elasticity on the short-term. The impact of on the long term would become zero (Nijs et al. 2001). This can also be stated on effects in this study. Buying decisions are actually influenced by these price promotions (Lowe et al. 2014). It seems that youth is more influenced than others. Young people show greater positive effects on the consumer preferences for mobile post-paid plans. Lee & Tsai (2014), describe that these discounts can influence emotions and decrease attention. Nevertheless, the impact is lower than expected as described in literature. It contributes to a pervasive information processing and quicker and easier preferences (Aydinli et al 2014), but not as effective as in other industries like fast-moving consumer goods.
Add-ons
In this study the influence of additional services on the consumer preferences are examined. There can be greater need for additional services when a base product is purchased (Erat & Bhaskaran, 2012). The results show negative effects for both youth and mass of additional services for music and movies subscriptions. 30 per cent extra data is seen as a benefit for the overall mobile post-paid plan. As data is part of the mobile post-paid already, for this study additional services are less relevant. This contradicts Morrison’s (2016) statements that product bundling can influence consumer preferences. All other attributes in this study show higher relative importance percentages towards the consumer preference. However, youth seem to be more interested in product bundling including these additional services compared to mass. Music add-ons are negatively related to the consumer preference of youth for mobile post-paid plans. Therefore, the literature and hypothesis 5 did not hold in this research.
5.2 Managerial implications
37
First of all, the ideal attribute levels that are preferred in the proposition for youth can be found in table 12.
Attribute Attribute level
Smartphone brand Apple iPhone 7
Telecom brand Company B
Subscription fee 35 Euro
Price promotions 30% discount
Add-ons 30% extra data
Table 12: Most preferred attribute levels
A young consumer prefers an Apple iPhone 7, Company B telecom provider, for 35 Euro, 30 per cent discount and 30 per cent extra data. These outcomes are similar for mass. This deal is most likely not profitable for companies. For this reason, the preferred attribute levels are not proposed all together. Secondly, the relative importance shows clear impact of the smartphone brand. The strongest positive effects are reached for the smartphone brand Apple iPhone 7, followed by the Samsung Galaxy S7. Subscription fee is most important after smartphone brands. Explaining that consumers are sensitive for price in their overall preferences towards mobile post-paid plans. There are differences among the strengths of telecom brands. Besides smartphone and subscription fees, the telecom brand seems to have moderate impact on the consumer’s decisions. Nevertheless, there is a bias for the fact that a Company B customer base is used in this research. The effects on young people are a bit smaller than the effect on others. As brands have strong impacts, price promotions are less effective. An interesting finding is the fact that youth are more attracted to price promotions than others. Overall, the impact of additional services is not high. Additional services are slightly more effective for youth compared to older people. Creating propositions that are aimed at each segment, with their own specifications that is preferred by the audience, could improve the customer-based view.
38
males. Also telecom providers are preferred as important in a more intense way compared to males. Telecom companies can focus on targeted audience based on gender to increase their brand consideration. Second, education does influence the preferences as well. The higher someone’s background in education, the higher the chance of preferring an Apple iPhone 7 instead of a Samsung Galaxy S7, and vice versa for the Samsung Galaxy S7. Based on these insights, managers could focus their digital marketing strategy of the Apple iPhone 7 towards highly educated people. For lower educated people both Apple iPhone 7 and Samsung Galaxy S7 are important and should be targeted simultaneously, which can be done for both youth and mass, since these groups show similarities here. Third, living situation of young people influence the preference towards smartphone brands. People that live alone have stronger preferences to the Apple iPhone 7 compared with their partner or in a family. The Samsung Galaxy S7 should be targeted to people that live together with their partner or family. Fourth, age turns out to be a very interesting demographic towards the preferences of mobile post-paid plans. Customer’s interests for the Apple iPhone 7 declines over age for both youth and mass. Customer’s interests for the Samsung Galaxy S7 increases over age. Young people should be focused on the Apple iPhone 7 and older people on both the Apple as the Samsung smartphone. Most customer audience does not prefer the Sony Xperia XZ, especially not young people. People that are older prefer Company B as a telecom brand to Company A and Company C. These insights give managers directions to strengthen their brands.
39
to Company A or Company C. This group is also most sensitive towards prices and need to be targeted with the lowest priced mobile post-paid plans.
The latent classes describe probabilistic classes found by analysing finite models. The outcome represents five classes for both youth and mass. Based on class sizes the youth sample can be divided and ordered in the following classes: average customer, Apple fans, loyal telecom customers, Samsung fans, and low budget customers. For mass these are the following: Apple fans, average customer, Samsung fans, moderate budget and lastly the low budget class. The classes that are mentioned first present higher class sizes and therefore need more attention than others. Instead of focussing on demographics, customers are now allocated to classes that are discovered by significant data. It gives marketing managers quantitative evidence for launching new propositions.
Overall, managers from telecom companies can benefit from this study for having new valuable insights. This can create optimized mobile post-paid plans that suit each and every target group described in this study. Smartphone brand, telecom brand, subscription fee, price promotions and additional services vary in importance towards consumer preferences. Most important, not only do we know the differences between the perception for youth and mass for mobile post-paid plans, but estimated prior demographic segments and latent classes that can be applied in marketing strategies.
6. LIMITATIONS AND FUTURE RESEARCH
40
not from the same telecom company. Thirdly, no none-choice option is included in the research design, which means all respondents were forced to choose for a mobile post-paid plan. This could affect preference when exposed in such a decision making process in real life. In future research the none-choice option could be applied to identify if customer truly buy the option in real life. Fourthly, this study is only performed on consumer market and not for the enterprise market. The enterprise market can have different attributes and therefore should use a different research design. Future research could be performed for small, medium or large companies. Fifthly, this investigation is only executed on mobile post-paid plans that include a smartphone. There are possibilities to wider this research on to pre-paid and SIM-only customers. Taking away the most relative important attributes, smartphone brand, can give very interesting findings for pre-paid and SIM-only propositions. Lastly, the outcome is only representative for the Dutch mobile market. All preferences on data usage, brand perceptions, and influence of demographics can all be different for other countries in Europe and could be even more different in other continents.
REFERENCES
Ahn, H., Ahn, J. J., Oh, K. J., & Kim, D. H. (2011). Facilitating cross-selling in a mobile telecom market to develop customer classification model based on hybrid data mining techniques. Expert Systems with Applications, 38(5), 5005-5012.
Akaichi, F., Nayga, R. M., & Gil, J. M. (2013). Are results from non-hypothetical choice-based conjoint analyses and non-hypothetical recoded-ranking conjoint analyses similar?. American Journal of Agricultural Economics, aat013.
Chicago
Akın, M. (2011). Predicting preferences of university students for pre-paid vs post-paid cell phone service plans. Expert Systems with Applications, 38(8), 9207-9210.
Al-Dmour, H., Zu'bi, M. F., & Kakeesh, D. (2013). The effect of services marketing mix elements on customer-based brand equity: An empirical study on mobile telecom service recipients in Jordan. International Journal of Business and Management,
8(11), 13.
41
Aydinli, A., Bertini, M., & Lambrecht, A. (2014). Price promotion for emotional impact. Journal of Marketing, 78(4), 80-96.
Băcilă, M., Rădulescu, A., & Mărar, I. L. (2012). Pre-paid telecom customer’s segmentation using the K-mean algorithm. Annals of the University of Oradea,
Economic Science Series, 21(1), 1112-1118.
Bijmolt, T. H., Heerde, H. J. V., & Pieters, R. G. (2005). New empirical generalizations on the determinants of price elasticity. Journal of marketing research, 42(2), 141-156.
Bossert, W., Pattanaik, P. K., & Xu, Y. (2000). Choice under complete uncertainty: Axiomatic characterizations of some decision rules. Economic Theory, 16(2), 295. CBS Statline. Bevoling; geslacht; leeftijd en burgerlijke staat, 1 januari. Retrieved 11-10, 2016, from http://statline.cbs.nl/Statweb/publication/?DM=SLNL&PA=7461 BEV&D1=0&D2=0&D3=1925&D4=0,10,20,30,40,50,l&HDR=T,G3&STB=G1,G2 &VW=T
CBS Statline. Bevoling; geslacht; leeftijd en burgerlijke staat, 1 januari. Retrieved 11-10,2016, http://statline.cbs.nl/Statweb/publication/?DM=SLNL&PA=7461BEV&D1 0&D2=0&D3=106-120&D4=0,10,20,30,40,50,l&HDR=T,G3&STB=G1,G2&VW=T Chen, Y. H., Hsu, I. C., & Lin, C. C. (2010). Website attributes that increase consumer purchase intention: A conjoint analysis. Journal of business research,
63(9), 1007-1014.
Daud, M. A. (2013). Awareness of Brand and Increasing Customer’s Demand for FMCG Products in Rural Market: The Rural Market of Uttar Pradesh. International
Journal of Engineering and Management Science, 4(1).
42
Donkers, B., Verhoef, P.C. & de Jong, M.G. Quant Market Econ (2007) 5: 163. doi:10.1007/s11129-006-9016-y
Eggers, F., & Sattler, H. (2011). Preference Measurement with Conjoint Analysis. Overview of State-of-the-Art Approaches and Recent Developments. GfK Marketing Intelligence Review, 3(1), 36-47.
Erat, S., & Bhaskaran, S. R. (2012). Consumer mental accounts and implications to selling base products and add-ons. Marketing Science, 31(5), 801-818.
Garrido, E., & Whalley, J. (2013). Competition in wholesale markets: Do MNOs compete to host MVNOs? Telecommunications Policy, 37(11), 1124-1141.
Ghodeswar, B. M. (2008). Building brand identity in competitive markets: a conceptual model. Journal of Product & Brand Management, 17(1), 4-12.
Gneezy, A., Gneezy, U., & Lauga, D. O. (2014). A reference-dependent model of the price-quality heuristic. Journal of Marketing Research, 51(2), 153-164.
Han, S. H., Lu, S. X., & Leung, S. C. H. (2012). Segmentation of telecom customers based on customer value by decision tree model. Expert Systems with Applications, 39(4), 3964-3973.
van Heerde, H. J. (2005). The proper interpretation of sales promotion effects: supplement elasticities with absolute sales effects. Applied Stochastic Models in Business and Industry, 21(4-5), 397-402.
Huang, C., Lee, G. M., & Crespi, N. (2012). A semantic enhanced service exposure model for a converged service environment. IEEE Communications Magazine, 50(3), 32-40.
Keller, K. L., Lehmann, D. R., & Farley, J. U. (2008). The structure of survey-based brand metrics. Journal of International Marketing, 16(4), 29-56.
43
knowledge structures: Elaboration and interference effects on the processing of sequentially advertised brand benefit claims. Journal of Marketing Communications, 20(3), 176-196.
Keller, K. L. (2014). Designing and implementing brand architecture strategies. Journal of Brand Management, 21(9), 702-715
Kiiver, P., & Kodym, J. (2015). PRICE-QUALITY RATIOS IN VALUE-FOR-MONEY AWARDS. Journal of Public Procurement, 15(3).
Lalwani, A. K., & Forcum, L. (2016). Does a Dollar Get You a Dollar's Worth of Merchandise? The Impact of Power Distance Belief on Price-Quality Judgments. Journal of Consumer Research, ucw019.
Lee, L., & Tsai, C. I. (2014). How price promotions influence postpurchase consumption experience over time. Journal of Consumer Research, 40(5), 943-959.
Leeflang, P.S.H., Wieringa, J.E., Bijmolt, T.H.A. & Pauwels, K.H. (2015) Modeling Markets Analyzing Marketing Phenomena and Improving Marketing Decision Making, 1st edition, New York: Springer.
Leone, R. P., Rao, V. R., Keller, K. L., Luo, A. M., McAlister, L., & Srivastava, R. (2006). Linking brand equity to customer equity. Journal of Service Research, 9(2), 125-138.
Lowe, B., Chan Fong Yee, F., & Yeow, P. (2014). Price promotions and their effect upon reference prices. Journal of Product & Brand Management, 23(4/5), 349-361.
Mahajan, V., Green, P.E., Goldberg, S.M.: A conjoint model for measuring self- and cross-price demand relationships. J. Mark. Res. 19, 334–342 (1982)
44
Malik, M. E., Ghafoor, M. M., & Hafiz, K. I. (2012). Impact of Brand Image, Service Quality and price on customer satisfaction in Pakistan Telecommunication sector.
International journal of business and social science,3(23).
Morrison, W. G. (2016). Product bundling and shared information goods: A pricing exercise. The Journal of Economic Education, 47(1), 49-63.
Nijs, V. R., Dekimpe, M. G., Steenkamps, J. B. E., & Hanssens, D. M. (2001). The category-demand effects of price promotions. Marketing science, 20(1), 1-22.
Paulrajan, R., & Rajkumar, H. (2011). Service quality and customer’s preference of cellular mobile service providers. Journal of technology management & innovation,
6(1), 38-45.
Penmetsa, N., Gal‐Or, E., & May, J. (2015). Dynamic Pricing of New Services in Subscription Markets. Production and Operations Management, 24(6), 896-916. Orme, B. K. (2010). Getting started with conjoint analysis: strategies for product design and pricing research. Research Publishers. 64.
Rahman, S., Haque, A., & Ahmad, M. I. S. (2011). Choice criteria for mobile telecom operator: Empirical investigation among Malaysian customers.International
Management Review, 7(1), 50.
van Roer, A. (2016). Dutch mobile operators 2016 Q2 (Research report) 5, Telecompaper
van de Roer, A (2016). Dutch mobile market slows revenue decline in 2015 to 4.9
billion euro. Retrieved November/2016, from http://www.telecompaper.com/
pressrelease/dutch-mobile-market-slows-revenue decline-in-2015-to-49-billion-euro--1133340
