by Alexander R.S. Oude Elferink
University of GroningenFaculty of Economics and Business Department of Marketing MSc Marketing thesis February – June 2015 Gelkingestraat 3-11 9711 NA Groningen (06)16284261 a.r.s.oude.elferink@student.rug.nl student number 2608405
Supervisors: Prof. Dr. P.C. Verhoef and E. De Haan MSc
F i n a l v e r s i o n t h e s i s
C o m p l e t i o n d a t e : 1 9 - 0 6 - 2 0 1 5
MANAGEMENT SUMMARY
By understanding customers’ experiences across multiple touchpoints and managing their accumulations, this may very likely culminate in avenues for building customer commitment, improved retention rates and more financial success. To facilitate managerial insight into touchpoint experiences and the overall customer experience (CE), this study did dive into researching the impact of customer touchpoints while accounting for CE differences among buying phases. Specifically CE is measured on the emotional level as it outweighs cognition in customer decision-making and thus is more valuable managerially. Consequently this study answered the following problem statement: ‘How do multiple customer touchpoints impact customer experience on the emotional level?’ To do so, multiple regression models were estimated, moderation analyses were conducted and latent class analysis (LCA) was executed to provide more thorough insight on the data (N =246); which was collected through an online survey. In the survey individuals were asked to indicate a retailer that is important to them, whereupon they evaluated one or multiple touchpoint experiences.
Findings show that physical stores are of crucial importance when one wants to deliver extraordinary experiences. However, therefore online stores are not necessarily inferior as these touchpoints can complement each other; i.e. when one can meet but preferably exceed customers’ expectations across touchpoints, CE will have positive valence. Moreover, LCA showed clear partitioning among single and multi touchpoint customers. Hence retailers should not focus on single touchpoint strategies, but should integrate touchpoints, to able to provide customers consistent experiences. Furthermore retailers are suggested to not focus on delivering a positive CE on one concrete buying phase; seeing that purchase is not by
PREFACE
It all started nearly 6 years ago: when I was occupied as Sales Assistant at a multi-brand fashion retailer, while I started to study for my BBA Marketing and Sales. Holding this position fostered my continuously growing interest in the Customer Experience (CE) construct: having had the privilege to facilitate positive shopping experiences is what delighted me and inspired me to extend my knowledge on this marvellous topic.
Still, the completion of this dissertation would not have been possible without guidance and support of several people who in some way contributed with their assistance during the fulfilment of my master thesis journey. Therefore I would like to genuinely disseminate my gratitude to them.
First above all, I would like to specially thank my supervisors Prof. Dr. Peter C. Verhoef and Evert de Haan for their critical, helpful and constructive comments as well as suggestions during my research. Thank you for sharing your knowledge, motivating me and showing me that the power of quality research lies within simplicity. Additionally I would like to thank Frank Beke for sharing his research thoughts for research improvement.
Nextly, I would like to thank my parents, sister and friends from the bottom of my heart for their motivation, interest, unconditional support and imperative distraction, during the realization of my thesis. Last but not least, I would like to thank Dr. Marcel E.A. Weber for preparing, guiding and learning me how to conduct academic research. Moreover, I would like to thank him for making me even more enthusiastic about the customer experience subject, and making me a so to say, ‘customer experience evangelist.’
For the Love of Customer Experience: Diving Into the
Impact of Customer Touchpoints And the Moderating
Effects of Buying Phases
TABLE OF CONTENTS
MANAGEMENT SUMMARY ………....………... 2
PREFACE ………..………....………... 3
1. INTRODUCTION ………....………... 5
1.1 PROBLEM STATEMENT AND RESEARCH QUESTIONS .………..… 6
1.2 RELEVANCE: LITERATURE CONTRIBUTION ………..…….. 7
2. THEORETICAL FRAMEWORK ……….……….……...……… 9
2.1 CUSTOMER TOUCHPOINTS: A CLASSIFICATION ……….... 9
2.1.1 DEFINING AND SELECTING TOUCHPOINTS …………...……… 12
2.2 CUSTOMER EXPERIENCE: WHAT IS IT? ……….…….……… 13
2.2.1 DIMENSIONS ………...……….. 14
2.2.2 FLOW AND PEAK EXPERIENCES ……...…….………...…... 15
2.2.3 CUSTOMER EMOTIONS: A SUCCINCT REVIEW …..………….…... 15
2.2.4 RESEARCH ON EMOTIONS ……...…….……….…... 16 2.3 HYPOTHESES ……….…….……….. 17 2.4 CONTROL VARIABLES ……….……….. 21 3. RESEARCH DESIGN ……….…….……...……….………... 23 3.1 DATA COLLECTION ……….……… 23 3.2 MEASUREMENTS OF VARIABLES ….………... 23
3.3 METHODS AND MODEL OF ANALYSIS ….………... 26
4. RESULTS ……….…….……...………..………...….... 29
4.1 PREPARATION FOR ANALYSIS ……….…… 29
4.2 DESCRIPTIVE STATISTICS AND CORRELATIONS ……….…...… 30
4.3 MODEL VALIDATION AND TESTING …...………..…….….… 33
4.3.1 HETEROSCEDASTICITY ………...……... 33
4.3.2 NONNORMAL ERRORS ………...….……... 34
4.3.3 MULTICOLLINEARITY ………...….……... 35
4.4 MAIN EFFECTS: THE IMPACT OF CUSTOMER TOUCHPOINTS ……….… 36
4.5 TOUCHPOINT INTERACTION ……….….… 39
4.6 CE DIFFERENCES AMONG BUYING PHASES …...……….….… 39
4.7 LATENT CLASS ANALYSIS …...………..………..….… 40
5. CONCLUSIONS & RECOMMENDATIONS …...……….……... 46
5.1 DISCUSSION ……….……….…… 46
5.2 MANAGERIAL IMPLICATIONS ……….…… 49
5.3 LIMITATIONS AND FUTURE RESEARCH ………...……….…… 51
REFERENCES ……….…….……...……….………... 54
APPENDICES ……….…….……...……….……….... 62
I: ENGLISH QUESTIONNAIRE ….………... 62
II: DUTCH QUESTIONNAIRE ….………... 66
III: DESCRIPTIVE STATISTICS AND CORRELATIONS …….………... 70
1. INTRODUCTION
For a long time companies accentuated the importance of touchpoints – multiple critical moments of customer interactions with the company and its offerings from stages before till after moments of purchase. Nevertheless, focussing on satisfaction maximization on specific touchpoints is to narrow: this creates distorted views of what customers truly feel. This focus might suggest that customers are actually not as happy as they state they are. Besides, it tarnishes the bigger – more important – view: the customers’ end-to-end journey (Rawson, Duncan and Jones, 2013). Therefore one should distinguish broadly between paid, owned and earned touchpoints, which vary in extent of firm controllability and the extent to which they demand resource allocation (Stephen and Galak, 2012; Cuthill, 2013; Baxendale, Macdonald and Wilson, 2015).
Unfortunately, only few companies manage the sum of experiences across multiple
touchpoints and in multiple channels over time (Rawson, Duncan and Jones, 2013). For many companies this is a missed opportunity, since the accumulation of experiences across
touchpoints over time, creates the representation of a brand in customers’ minds (Koll, Von Wallpach and Kreuzer, 2010). Furthermore effectively managing all touchpoints, would provide a great avenue to build customer commitment, retention and sustainable financial success (Lemke, Clark and Wilson, 2011). However, to do so, requires companies to
understand the customer outcomes, processes and the way customer experience develops over time due to the interactions at various touchpoints (Bitner, Ostrom and Morgan, 2008; Jüttner, Schaffner, Windler and Maklan, 2013).
may vary across these touchpoints (Garrett, 2006; Penz and Hogg, 2011). As soon as customers’ relative levels of affect become increasingly positive, in turn levels of
participation (e.g. information sharing among customers, provision of suggestions and shared decision making), i.e. customer engagement behavior, (Van Doorn et al. 2010), spending and repeat purchase intentions (Arnold and Reynolds, 2009) increase either (Gallan, Jarvis, Brown and Bitner, 2013). Specifically, customer value management (CVM) principally considered cognitive aspects of customer experience, but did pay less attention to emotional and affective attitudes (Verhoef and Lemon, 2013) in order to foster e.g. customer loyalty. Additionally, research primarily focussed on studying touchpoints in isolation (Baxendale, Macdonald, Wilson, 2015).
1.1 Problem statement and research questions
As appears, not many companies have a profound view of customer experience, since they did not look into their customers’ experiences across multiple touchpoints in the whole journey they make, i.e. from search to moments after purchase. This is very likely caused by the fact that there is no substantial amount of customer experience research that investigated
emotional and affective attitudes. Specifically not over time, across multiple touchpoints and the consequence(s) corresponding developments may have. Therefore the problem statement regards: How do multiple customer touchpoints impact customer experience on the emotional level? Corresponding research questions are:
- Why do(es) (a) certain touchpoint(s) elicit specific customer emotions?
1.2 Relevance: literature contribution
What can be derived from the aforementioned literature, problem statement and research questions, is that former customer experience research did not cover the emotional and affective components of customer experience (Verhoef and Lemon, 2013) and other research did not execute research on multiple touchpoints simultaneously (Baxendale, Macdonald, Wilson, 2015). Consequently this study intends to make a significant contribution to customer experience literature, by diving into the impact of multiple customer touchpoints. More specifically, this study will extend the study of Baxendale and colleagues (2015) by not only investigating multiple touchpoint experiences - by measuring affective responses - but by showing how multiple touchpoints impact customer experience on the emotional level. At first it might seem to confined to focus on the emotional level solely, as customer experience comprises cognition as well (Jüttner, Schaffner, Windler and Maklan, 2013). However, an experience comprises for more then 50% emotions (Shaw, 2007). Adjacent to this, since several studies showed that emotions outweigh cognition in customer decision making (Shiv and Fedorikhin, 1999; Lerner, Small and Loewenstein, 2004), across offline as well as online touchpoints (Rose, Clark, Samouel and Hair, 2012), this study focuses on emotional responses that are evoked during a customers’ experience. These findings are verified by a meta-analysis containing 35 years of knowledge on emotions and decision making that point out that emotions serve as potent, pervasive, predictable, coordinating mechanisms that can be either beneficial or harmful for decision making (Lerner, Yi, Valdesolo and Kassam, 2015).
Customer touchpoints Physical stores Online stores Catalogues Mobile phones Touchpoint experience - Disconformation - Positivity Customer experience Customer delight Control variables Customer characteristics - Gender - Age - Culture Brand characteristics - Price image - Service quality H1 - H5 Buying phase Search / Purchase
2. THEORETICAL FRAMEWORK
As can been seen from the conceptual model in figure 1, this study investigates the impact of multiple customer touchpoints on customer experience’ emotional level by means of customer delight. In particular this study examines the experience of touchpoints independently in terms of whether expectations of the experience, are or are not, either met or exceeded (i.e. disconfirmation) and in terms of valence. Furthermore this study examines how these
accumulated touchpoint experiences affect customer experience, if touchpoints interact (i.e. if synergies exist) and how they affect customer experience and whether customer experience differs in the two distinguished buying phases; search and purchase. As customer experience may vary across individual customers, controlled is for gender, age, culture and their
perceived price image and service quality of the retail brand that is important to them.
Figure 1 – Conceptual model
2.1 Customer touchpoints: a classification
Specific touchpoints that customers might engage in nowadays are: stores, internet,
Table 1 – Touchpoint classification and research study examples
Touchpoint types
Paid touchpoints Owned touchpoints
Earned touchpoints Br an d ad ver ti si ng Re ta il er ad ver ti si ng Te le vi si on Ou td oo r Pr in t Ra di o (In ) S to re s (d is pl ay s) On li ne Ca ta lo gu es Mo bi le phone s E -ma il s Gi ft -pa cks / Sa m pl in g Ev en ts / Co m pe ti ti on WO M Pe er o bs . So ci al m ed ia Pu bl ic R el . (Burke, 2002) x x x x x (Ansari, Mela and Neslin, 2008) x x (Trusov, Bucklin and Pauwels, 2009) x x (Romaniuk, Beal and Uncles, 2013) x x x x x x x x x x x x (Lim, Al-Aali and Heinrichs, 2015) x x (Baxendale, Macdonald and Wilson, 2015) x x x x x x Touchpoint Charac-teristics
1.) Control (firm / agent owned) Controlled (firm / agent owned) Controlled Uncontrolled (customer
2.1.1 Defining and selecting touchpoints
Thus, when considering table 1, but also accounting for the definition of touchpoints from Meyer and Swager (2007) and some others (Payne and Frow 2004; Logman, 2004; Gentile, Spiller and Noci, 2007), customer touchpoints are defined as: customer interactions with (a) product(s) and/or service(s) or representations of it, by a brand (direct contact) or a third party (indirect contact), in certain channels at specific moments in time; from pre-purchase to purchase to post-purchase. Seeing the three factor touchpoint classification (table 1),
attention needs to be focussed on owned and earned touchpoints, as consumers are mostly influenced during ‘initial consideration’ were they seek information from retailers, peers or reviewers and as allocated marketing budgets from 70 to 90% to (advertising) paid
touchpoints seem ineffective (Edelman, 2010). Moreover, given that owned touchpoints - principally service interfaces and its elements - are under company control, in contrast to influences of others (Verhoef et al. 2009) and the fact that they demand longer-term thinking on scales of resource allocation (i.e. they are more future oriented) (Baxendale, Macdonald and Wilson, 2015), this study focuses on sales channels. Consequently, the first touchpoint distinguished regards: physical stores and is extended to a set of touchpoints by respectively: online stores, catalogues and mobile phones, as channel assembly affects consumers’ retailing decision-making, based on uniquely enriching experiences they can provide (Nicolau, 2013).
experience, that stimulates thinking and/or senses by certain stimuli and thus provides an individual cognitive as well as emotional benefits, that may satisfy corresponding needs (Hirschman, 1984; Tynan and McKechnie, 2009).
2.2 Customer experience: what is it?
Also Schmitt (1999) emphasizes that experiences are personal and manifest as a reaction evoked by certain stimuli (e.g. marketing communications), thus are principally induced and not created by individuals themselves. Meyer and Schwager (2007) argue that customer experience refers to a customers’ internal, subjective response, which occurs through direct or indirect contact with a company. This indicates that customer experience may nearly always occur, through elements that are controlled or uncontrolled by a company (Verhoef et al. 2009). To properly define customer experience, this study fundamentally builds on the definition of Gentile et al. (2007) for three reasons. Firstly, given the many relevant scientific contributions it takes into account. Secondly, due its acknowledgement of customer
By considering that customer experience occurs at different touchpoints, one should view customer experience as a blend of all touchpoints and its aforementioned levels. Also the fact that customer experience is strictly personally unique and changeable (Mascarenhas, Kesavan and Bernacchi, 2006; Gentile, Spiller and Noci, 2007; Palmer, 2010), indicates that even the same individual may experience the same touchpoint differently, at an other moment in time. Consequently, customer experience as a total construct of touchpoints, may differ individually as well.
2.2.1 Dimensions
2.2.2 Flow and peak experiences
To define what is a perfect or extraordinary customer experience (Frow and Payne, 2007), referred is to flow and peak experiences, respectively proposed by Mihaly Csikszentmihalyi (Csikszentmihalyi and LeFevre, 1989) and Daniel Kahneman (Fredrickson and Kahneman, 1993). Flow can be characterized as total absorption in an activity, which is achieved through mastery of a certain activity by intense focused engagement. So, flow can be described as a positive, often playful, highly fulfilling package created by performance and experience being coalesced. Indication of having a flow experience relates to being in a state of transcendence, a feeling of temporally losing the sense of reality and feeling of being united with some higher level of experience. Similar to flow experiences, peak experiences result in a state of transcendence. However, peak experiences regard more to an epiphany instead of a deliberate process, arise frequently externally from individuals and elevate them to unexpected
emotional heights (Schouten, McAlexander and Koenig, 2007), which will foster emotional bonding or not (Johnston and Kong, 2011). Given this, plus the fact that emotions are a substantial denominator of customer experience (Ismail, Melewar, Lim and Woodside, 2011; Shaw, 2007), academic literature on customer emotions will be outlined subsequently.
2.2.3 Customer emotions: a succinct review
differ from affect in sense that affect represent mental feeling processes that can be considered general (since they may include emotions, moods and attitudes) instead of a particular psychological process. Secondly, emotions differ from moods since emotions are characterized by higher intensity, are generally shorter lasting, are intentional (i.e. have a object or referent) and are more connected to behavioral intentions or even explicit behavior. Thirdly, emotions differ from attitudes as attitudes can be considered as merely evaluative judgements (e.g. good versus bad reactions). Thus very briefly, emotions are internal feelings marked by mental, bodily or behavioral symptoms that are directed at an individual or
something (Frijda, 2007).
In this study customer emotions are viewed as ‘mental sates of readiness that arise from cognitive appraisals of events or thoughts, that have a phenomenological tone, are
accompanied by physiological processes, are often expressed physically and may result in specific actions to affirm or cope with the emotion(s), depending on its nature and meaning for the person having it’ (Bagozzi, Gopinath and Nyer, 1999). This definition is considered appropriate as its elaborate composition corresponds to findings in research on the dimensions of emotions (Fontaine, Scherer, Roesch and Ellsworth, 2007).
2.2.4 Research on emotions
Every experience that a touchpoint causes, may induce a broad spectrum of emotions (Hirschman and Holbrook, 1982) instead of a basic set (Ekman, 1992). Despite of this, two basic emotion measure constructs that got really popular, are those of Izard and Plutchik. However, still there is no widespread agreement on how many basic emotions exist and several authors did note unbalanced constructs (Richins, 1997; Westbrook and Oliver, 1991). Unbalance is principally created since the number of positive emotions are a minority
received more attention since they are associated with problems and should be solved and emotion theorists currently focussed on explaining emotions in general (Fredrickson, 1998). Next to the basic emotion constructs of Izard and Plutchik, up till today, various manners to measure emotions are used. Dimensional theories can be seen as one of the more influential approaches to investigate emotions. They simply distinguish specific emotions based on common dimensions of affect. Examples of dimensions theories are the PAD model (Russell and Mehrabian,1977) and the circumplex model (Watson and Tellegen, 1985). Furthermore the attribution theory – although developed for an other purpose – has been utilized for prediction of differentiated emotional responses that arise from various distinctions individuals make about the causes of an event (Weiner, 1985).
Within the marketing discipline in general, emotions tend to be measured empirically by relying on self-reported unipolar or bipolar items in questionnaires. Afterwards, factor analysis, multidimensional scaling or cluster analysis are frequently utilized to indicate underlying emotional dimensions (Bagozzi, Gopinath and Nyer, 1999). Batra, Ahuvia and Bagozzi (2012) for example, focussed on studying one specific emotion; they created a construct to measure brand love. In contrast, Shaw (2007), derived four clusters of emotions based on value and loyalty, respectively (from low value and loyalty); the destroying cluster, attention cluster, recommendation cluster and advocacy cluster, which distinguish 20 discrete emotions in total. How emotions are measured in this study, is outlined in the research design chapter.
2.3 Hypotheses
Tynan and McKechnie, 2009), which are studied here in the light of touchpoints as sales channels. What is known, is that the impact of individual touchpoints vary: specifically Baxendale, Macdonald and Wilson (2015) find that in-store communications, brand
advertising and the observation of other customers have the highest impact. So with regard to the considered touchpoints in this study, it can be assumed that physical stores have the most substantial impact on customer experience. This is empirically justified since varying levels of emotions can be evoked by ambient store conditions such as store layout, music and its tempo, light, design, employee and customer appearance (Shukla and Babin, 2013; Mohan, Sivakumaran and Sharma, 2013; Knoferle, Spangenberg, Herrmann and Landwehr, 2012). From the aforementioned appears that not only controllable factors (for retailers) in a physical store influence customer emotions, the presence of other consumers does so too (Verhoef et al. 2009). Next, the behavior of salespersons play a significant role (Mallalieu and Nakamoto, 2008; Harris, 2007; Mosley, 2007) as a result of expertise that help customers in effective decision-making and the possible enhancement of emotional bonding. In contrast to the assumed impact of physical stores, the impact of printed media – in this case catalogues – is expected to have the lowest impact, given that it lacks (human) interactivity. This is also what mainly discerns printed media from digital media; whereas information consumption in printed media occurs linearly, consuming information digitally is interactive and consumer driven (Kirk, Chiagouris and Gopalakrishna, 2012). So more explicitly, given that printed media only deliver on visual stimuli (i.e. text, images and graphics), makes them static (Hoffman and Novak, 1996). Consequently hypothesized is:
H1: Touchpoint experiences of physical stores have the strongest impact on customer experience compared to
online stores, catalogues and mobile phones
H2: Touchpoint experiences of catalogues have the less strongest impact on customer experience compared to
However, touchpoints not only impact customer experience solely; as pointed out in section 2.2, customer experience should be viewed as a blend of all touchpoints. When customers can engage in multiple touchpoints, it is more likely that different needs will be met, consequently resulting in positive touchpoint experiences (Wallace, Giese and Johnson, 2004) and in the end a positive overall customer experience. This can be asserted given the vivid and intentional nature of touchpoint experiences, that leave multiple traces in memory, which jointly form a brands representation (Biedenbach and Marell, 2010; Koll, Von Wallpach and Kreuzer, 2010) and is contingent on the consistency of all touchpoint experiences (Homburg, Koschate and Hoyer, 2006). In other words, it can be hypothesized that:
H3: More positive touchpoint experiences accumulate in a better customer experience
From H3 it can be derived that touchpoint experiences work additive; i.e. the more positive touchpoint experiences one has, the better the overall customer experience will be.
Nevertheless, all touchpoint experiences are rather synergistic than additive. Because when all touchpoints are properly melded, the total customer experience will exceed all touchpoint experiences separately (Berry, Wall and Carbone, 2006).
To elucidate when a combination of touchpoint experiences, results in a certain customer experience, one mechanism is distinguished, that is at work during the so called research-shopper phenomenon (i.e. the extent to which a consumer is inclined to search a product or service on one touchpoint and purchase on another touchpoint) (Verhoef, Neslin and
induce synergy effects. That simply implies: 1+1=3 (Romaniuk, Beal and Uncles, 2013). So channel synergy refers to higher attitudes toward search or purchase on e.g. online stores, which transposes into higher attitudes toward search or purchase on e.g. physical stores. What should be noted is that the latter effect can be inversed when higher attitudes toward search on e.g. mobile phones, transposes lower attitudes toward purchase on e.g. catalogues. In this case one may speak of a negative cross-channel synergy. In other words, touchpoints can be considered as substitutes. If there is a matter of cross-channel synergy, searching on
touchpoint X will enhance the experience of purchasing on touchpoint Z (Verhoef, Neslin and Vroomen, 2007). Customers can for example, easily engage in pre-purchase shopping
activities such as information search and product comparison online to get more knowledge about e.g. prices and other store attributes in order to maximize shopping value (Forsythe, Liu, Shannon and Gardner, 2006; Gupta and Kim, 2010). Accordingly customers’ internal locus of getting a good deal, may evoke a smart-shopper feeling (Scarpi, 2012), but not always (Pechpeyrou, 2013). Hence hypothesized is:
H4: Touchpoint synergies result in a more positive customer experience
As consumers seek sensory stimulation (Hirschman and Holbrook, 1982) for either learning, but moreover, for their emotional well-being (Krishna, 2012), The last hypothesis of this study concerns:
H5: Customers will have a better customer experience during the buying phase of purchase in contrast to search
2.4 Control variables
3. RESEARCH DESIGN
Respectively this section describes the: method of data collection, measurements of variables and the methods and model of analysis.
3.1 Data collection
Reflecting on the proposed conceptual model, ideal data for this study would stem from: i.) Individuals that are customer of a brand with multiple channels to engage in (brand may be known for its omni-channel strategy); ii.) And/or may serve as status symbol due to the brand’ equity.
An online questionnaire will be spread within the scope of customers who ‘shop’ at a retail brand that is important to them. Consequently, this study can quantitatively measure customers’ customer experience on the emotional level and touchpoint experiences across multiple touchpoints and buying phases. The study will utilize a questionnaire that is
developed in English (Appendix I) and will be distributed online (via Facebook, LinkedIn and Twitter by Qualtrics) in Dutch (Appendix II), to engage a larger group of respondents. In order to overcome any form of discomfort of confusion among respondents regarding the questionnaire, it will made clear that they will be kept anonymous, that results will
exclusively be used for this study, the questionnaire will be kept as succinct as possible and the purpose of the study will be disclosed after completion on demand.
3.2 Measurements of variables
latter, might be a lot to investigate across multiple touchpoints and given individuals’ structured knowledge on emotions (Laros and Steenkamp, 2005; Shaver, Schwartz, Kirson and O’Connor, 1987) and possible inaccurate recalls of emotions after a period of time (which may be reconstructed due to biases such as self-justification) (Cowley, 2008; Aaker, Drolet and Griffin, 2008), due to not measuring customer responses directly like in the study of Baxendale et al. (2015), this study does not measure customer experience on scales of emotions directly. Instead, customer experience is operationalized by measuring emotions more indirectly through customer delight, that can be considered as an emotional response metric (Finn, 2012; Oliver, Rust and Varki, 1997; Wang, 2011) and is different from mere satisfaction as this is a more cognitive metric (De Haan, Verhoef and Wiesel, 2015). Concretely the three item 5 point scales, proposed by Finn (2012) and used by others (e.g. Kim and Mattila, 2013) are adopted and measure how frequently one felt gleeful, elated and delighted (1 = never; 5 = always) based on experience(s) with his / her important retail brand, to operationalize customer experience.
Touchpoint experience, is gauged by four items. The first three items relate to
(1 = much worse than expected; 7 = much better than expected) and to what degree the overall touchpoint experience was better or worse than expected (1 = much worse than expected; 7 = much better than expected). The last item was operationalized by measuring how a certain touchpoint made customers feel in terms of positivity on a 7-point Likert scale (1 = very negative; 7 = very positive) (Baxendale, Macdonald and Wilson, 2015; White and McFarland, 2009).
Further, buying phases (i.e. search and/or purchase) are operationalized by dummy coding them (Gensler, Verhoef and Böhm, 2012); 0 = search, 1 = purchase and if differently, there is a matter of both phases in a specific touchpoint for a customer. Controlled is for gender, age, culture, perceived price image and service quality. For the last two control variables, build is on concrete constructs of previous studies. Price image is measured through the extent that one thinks brand x has very good prices (1 = totally disagree; 7 = totally agree) (Hamilton and Chernev, 2013).
Table 2 – Overview measurement variables
Measurement construct Literature Scale
Customer experience (customer delight): gleeful,
elated, delighted
Finn (2012); Kim and Mattila (2013)
5-point Likert
Touchpoint experience:
disconfirmation + positivity
Wallace, Giese and Johnson (2004); Baxendale, Macdonald
and Wilson (2015); White and McFarland (2009)
7-point Likert
Buying phases: search,
purchase
Gensler, Verhoef and Böhm (2012)
Dummy coded (0 = search, 1 = purchase and differently = both)
Price image (control) Hamilton and Chernev (2013) 7-point Likert
Service quality (control) Estelami, Grewal and
Roggeveen (2007)
Finally, service quality is operationalized by (Estelami, Grewal and Roggeveen, 2007) measuring to what degree people perceive that brand x takes good care for its customers (1 = totally disagree; 7 = totally agree).
Additionally, although not formally included in the conceptual model (figure 1) nor in the measurement overview (table 2), this study will include three extra variables. The first extra variable concerns the Net Promoter Score (NPS) developed by Reichheld (2003). NPS measures how likely one would recommend brand x (in this context the retail brand that one indicated to be important) to a friend or colleague on a 0 to 10 point scale (0 = extremely unlikely to recommend; 10 extremely likely to recommend). NPS allows calculating the ratio of promotors (the ones that are very likely to recommend) to detractors (the ones that are very unlikely to recommend). The second extra variable relates to loyalty intentions, that is
operationalized by a self reported probability of engagement in repurchase behavior by picking a number between 1 and 100 (Rust, Lemon and Zeithaml, 2004) for the retailer mentioned to be important. Lastly, the third extra variable regards brand liking, which is closely related to brand preference (Anselmsson, Johansson and Persson, 2008) and is
included since concrete brand rankings cannot be made due not studying a specific brand and its competitors. Brand liking will capture the consumers’ overall perspective of how positive and strong perceived brand assets are. Specifically, this study will measure what individuals’ general attitude is (1 = very negative; 9 very positive) towards the retailer they find important (Anselmsson, Johansson and Persson, 2008).
3.3 Methods and model of analysis
touchpoints (Verhoef, Neslin and Vroomen, 2007). Still, customer experience can be viewed as a blend of all touchpoints, i.e. more additive (Grewal, Levy and Kumar, 2009). Therefore data can be analyzed by building a model that is linear in parameters and variables (Leeflang, Wieringa, Bijmolt and Pauwels, 2015), but also includes interaction effects, given that if touchpoints are melded, the effects of all touchpoint experiences separately will be surpassed by the total customer experience (Berry, Wall and Carbone, 2006).
The specified mathematical formula listed below, represents the conceptual model:
where:
CEj = customer experience of retail brand j
α0j = the unknown constant (intercept) of retail brand j
α1 .. α12 = unknown slope (effect) parameters
TPchoicePhyi = dummy variable touchpoint choice physical store(0 = no, 1 = yes) at retail brand j
TPchoiceOnli = dummy variable touchpoint choice online store(0 = no, 1 = yes) at retail brand j
TPchoiceCati = dummy variable touchpoint choice catalogue(0 = no, 1 = yes) at retail brand j
TPchoiceMobi = dummy variable touchpoint choice mobile phone(0 = no, 1 = yes) at retail brand j
TPchoiceij = dummy variable touchpoint choice i (0 = no, 1 = yes) for retail brand j
ExpTpij = experience touchpoint i (1 = physical store, 2 = online store, 3 = catalogue,
4 = mobile phone) by respondent x at retail brand j
means a touchpoint involves both buying phases) of respondent x at retail brand j
genj = (control variable) gender of respondent x at retail brand j
agej = (control variable) age of respondent x at retail brand j
cultj = (control variable) nationality of respondent x at retail brand j
Pimgj = (control variable) perceived price image of retail brand j by respondent x
Servqj = (control variable) perceived service quality of retail brand j by respondent x
εj = disturbance term for retail brand j
4. RESULTS
This chapter outlines what was done before estimating the proposed model and formally testing hypotheses; it describes various analyses, reports descriptive statistics and correlations and tests and validates the initiated model. Thereafter it expounds results regarding the impact of customer touchpoints, whether touchpoint interaction exists, differences on scales of CE across buying phases and lastly, it covers findings of LCA.
4.1 Preparation for analysis
In total 340 respondents were collected, from which after deletion of incomplete and invalid cases, 284 were valid and used for at least subsequent analyses in this section.
As preparation for analysis Cronbach’s alpha’s were computed individualistically for Customer Experience (CE) and for the touchpoint experiences of the physical store, online store, catalogue and mobile phone, to check whether satisfactory levels of internal consistency (i.e. their reliability) were exhibited. Cronbach’s alpha’s were only computed for all
aforementioned variables, since these were the only multiple item constructs measured in this study. Cronbach’s alpha was for CE (.813) and for respective touchpoint experiences; the physical store (.714), online store (.865) catalogue (.878) and mobile phone (.933). Following Nunnally’s criterion for acceptable internal consistency (Bloch, 1981), these levels are
considered to be satisfactory, since they are above .6. Consequently, for all four variables, all individual items were summed up and averaged so these could be used in later analyses.
variable to be displayed, was set at .01 percent.
When examining results from MVA, it was seen that from 4 out of 27 variables data was missing (14,8%) and in terms of values, that from 738 out of 7668 (9,6%) values that could be recorded, were missing. In a descending order data was missing for touchpoint experiences on mobile phones (267/284 respondents; 94%), for catalogues (257/284 respondents; 90,5%), for online stores (191/284 respondents; 67,3%) and for physical stores (23/284 respondents; 8,1%). In other words, most respondents had an experience in the physical store(s) of their important retailer but did not on their mobile phones.
Viewing missing value patterns and corresponding descriptive statistics, it can be stated a great amount of data is missing for catalogue and mobile phone touchpoint experiences. Therefore, unfortunately no valid conclusions can be made about present data for these touchpoint experiences. Accordingly all cases with mobile and catalogue experiences were omitted for subsequent analyses. Although some of these cases have had a physical and/or online store experience, full case omittance is justified; since from a theoretical viewpoint, cases with originally e.g. 4 touchpoint experiences, can not have an identical overall CE, when only accounting for 2 touchpoint experiences in the data (i.e. CE would lose measurement accuracy). In total this leaves 246 valid cases for computing descriptive statistics and analyses to follow. What should be noted is that only H2 cannot be tested anymore, given data exclusion of catalogue touchpoint experiences.
4.2 Descriptive statistics and correlations
(35%) and did belong principally to the age category of 18-29 years. Furthermore it can be noted that only 5 out of 246 respondents reported to have a non-Dutch culture. Lastly it can be noted that the majority of all customers perceived the price image and service quality of their important retailer to be positive, as the corresponding mean was approximately 5. Details about how many respondents had a certain touchpoint experience and how frequencies are distributed across buying phases, can be seen in table 3.
Table 3: Frequency distribution touchpoint experiences and buying phases.
What can be seen from Appendix III (summarizing table incl. descriptive statistics and correlations), is that touchpoint experiences for physical stores are moderate to strong positively correlated to customer experience (r = .349; p < .01) This implies that if these touchpoint experiences would get more positive, customers will have a better overall customer experience. Concretely the physical store touchpoint experiences are greatly determined by perceived service quality (r = .394; p < .05). Besides, if the overall customer experience gets less positive, the chance that people are inclined to buy at their important retailer decreases (r = -.181; p < .01). Noteworthy is that for individuals who searched for products/services in physical stores, are more inclined to buy in online stores and the other way around (r = .339; p < .01). When looking at gender, it can be stated that females are more
Variable Frequency
(missing)
Percentage
(missing)
TP exp. Physical store 230 (16) 93.5% (6.5%)
TP exp. Online store 66 (180) 26.8% (73.2%)
Buying phase Physical store 230 (16) 93.5% (6.5%)
Search 4 1.6%
Purchase 59 24.0%
Search and Purchase 167 67.9%
Buying phase Online store 66 (180) 26.8% (73.2%)
Search 11 4.5%
Purchase 8 3.3%
likely to have a better overall customer experience (r = .206; p < .01), which may be due a better perceived price image (r = .173; p < .01). Still what holds in general, is that a better perceived price image (r = .186; p < .01) and service quality (r = .229; p < .05) influence customer experience positively. If this perceived price image is less positive, people will be less inclined to buy in physical stores (r = -.286; p < .01). More specifically, females buy less likely in physical stores in contrast to men (r = -.147; p < .05). Moreover, the older
individuals get, the more they seem to have a negative overall customer experience (r = -.309; p < .01). A last remarkable thing is that when people have a better perceived price image, they tend to have a better perceived service quality and visa versa (r = .220; < p .05).
Finally, there seems to be no indication for multicollinearity, since all correlations are well below .8 (Mason, William and Perreault, 1991). However it seems there are no severe problems of multicollinearity present in the data, checked will be for multicollinearity in section 4.3, were the initiated model will be validated.
4.3 Model validation and testing
Before the initiated model from section 3.3 will be further estimated and all hypotheses will be tested, the model needs to be validated. In order to validate the proposed model, 6
assumptions are considered (if relevant) regarding various model elements, as is intended to estimate this model and test hypotheses through an ordinary least squares (OLS) procedure (Leeflang, Wieringa, Bijmolt and Pauwels, 2015). There are four assumptions for the disturbance term:
1.) E(εt) = 0 for all t;
2.) Var(εt) = σ2 for all t;
3.) Cov(εt εt’) = 0 for t ≠ t’;
4.) εt is normally distributed.
The other two assumptions concern:
5.) There is no relation between predictors and εt i.e. Cov(xt εt) = 0 (one-variable case).
6.) For the K-variable case, the matrix of observations X has full rank.
4.3.1 Heteroscedasticity
H0 = Variance in the residuals are assumed equal
H1 = Variance in the residuals are assumed unequal
In order to conduct the test, data is split by creating a dummy variable, whereupon
unstandardized residuals are compared. Levene’s test of homogeneity of variances, turns out to be insignificant for a p-value of (.290). In other words, variances in the residuals are
assumed equal and H0 cannot be rejected. Therefore there is no violation of heteroscedasticity
and thus no remedy is needed by e.g. a generalized least squares (GLS) procedure to re-estimate the parameter re-estimates i.e. the beta values.
Before continuing to test the fourth assumption, it should be noted that the third assumption will not be tested, since the data is pure cross-sectional and does not include a time factor. Hence testing for autocorrelation does not make sense, as data is not naturally ordered (Leeflang, Wieringa, Bijmolt and Pauwels, 2015).
4.3.2 Nonnormal errors
4.3.3 Multicollinearity
Finally is tested if variables in the model are correlated, i.e. if there is a relation between different predictor variables. To detect multicollinearity VIF values can be checked, which should be smaller than 5, following Leeflang et al. (2015). What can be seen from the estimated model is that for all touchpoint experience variables (incl. the touchpoint choice interaction dummies), touchpoint choice dummies, moderator dummies and interactions with touchpoint experiences, plus for the touchpoint experience interaction between the physical and online store, VIF values are far above 5, except for all control variables (gender, age, culture, perceived price image and service quality).
These discerned VIF values are not surprising, as products of variables (i.e. interactions) are often highly correlated with constituents (Smith and Sasaki, 1979; Leeflang and Wieringa, 2010), given their power to capture effects simultaneously due to overlap.
By stepwise omittance of all moderator dummies, the touchpoint experience interaction of the physical and online store and touchpoint choice dummies, all VIF values drop well below 5 for the (moderator) interaction terms and stay below 5 for all other variables as well. Hence is proven that the high VIFs were purely present due to products of variables and thus are no severe harm for further analyses. However, to mitigate any potential threat of
multicollinearity, decided was to apply mean-centering in order to improve interpretation of lower-order terms, while still accounting for that it may remain part of the inherent variance structure (Shieh, 2011; Echambadi and Hess, 2007).
4.4 Main effects: the impact of customer touchpoints
To test the main effects in this study, three regression models are build and estimated stepwise. For all analyses, p-values below .1 are considered significant, given the relatively small sample size of N = 246. The first regression analysis for the basic model (the model without the buying phase moderator):
was found to be significant as F = 9.953 for p = .000. Hence, data showed good model fit as it sufficiently explained variations in customer experience. Specifically, the model explained 27,5% of the variation in customer’ overall customer experience at their important retailer. Looking at both touchpoints, only touchpoint experiences at physical stores had a significant impact on customer experience (γ = .336; t = 4.682, p < .1; .000). Touchpoint experiences for the online stores (γ = -.112; t = -1.174, p < .1; .241) had an insignificant impact on customer experience (see table 4).
Table 4: Basic regression model (excl. bphase moderator) - N =246 R2 = .275, Adj. R2 = .247, F-stat = 9.953, P-value = .000
* Results are significant at p < .1 Independent variables Parameter estimate Standard error P-value Constant 1.996 0.495 0.000*
ExpTPphy x TPchoicephy 0.336 0.072 0.000*
ExpTPonl x TPchoiceonl -0.112 0.095 0.241
For the second regression analysis, the basic model was extended by adding the dummy variables of the buying phase moderator:
Also this model was overall significant corresponding to F = 7.625 for p = .000. This model slightly explained more variation in customer’s overall customer experience: 29,9%. Still, also for this model only touchpoint experiences at physical stores impact customer experience significantly (γ = .359; t = 4.999, p < .1; .000), while the online stores do not (see table 5).
Table 5: Basic regression model (incl. dummies bphase moderator) - N =246
R2 = .299, R2 change = .024, Adj. R2 = .260, F-stat = 7.625, P-value = .000
* Results are significant at p < .1
Lastly, the third regression model was extended by adding interaction effects between touchpoint experiences and the buying phase moderator:
Independent variables Parameter estimate Standard error P-value Constant 2.214 0.502 0.000*
ExpTPphy x TPchoicephy 0.359 0.072 0.000*
ExpTPonl x TPchoiceonl -0.142 0.098 0.149
This model showed that F = 5.739 for p = .000, matching an R2 of .300. Hence this model
explains 30% of the data; thus marginally more as the second regression model. Neither in this model the impact of touchpoint experiences on customer experience did not change, rather its effects remained relatively stable (see table 6).
Table 6: Basic regression model (incl. dummies + interactions bphase moderator) - N =246 R2 = .300, R2 change = .000, Adj. R2 = .247, F-stat = 5.739, P-value = .000
* Results are significant at p < .1
Based on all three regression models and its parameter estimates, it can be stated that there is evidence for H1, as touchpoint experiences at physical stores have the largest impact on customer experience. Unfortunately, there is no evidence for H3, seeing that the signs for touchpoint experiences at online stores are not positive; i.e. more (positive) touchpoint experiences do not necessarily result in a better overall customer experience.
Independent variables Parameter estimate Standard error P-value Constant 2.202 0.514 0.000*
ExpTPphy x TPchoicephy 0.361 0.086 0.000*
ExpTPonl x TPchoiceonl -0.145 0.138 0.295
TPchoicephy 0.164 0.210 0.436 TPchoiceonl 0.042 0.131 0.752 BphasePhysearch 0.422 0.459 0.359 BphasePhypurchase -0.261 0.111 0.020* BphaseOnlsearch -0.175 0.247 0.481 BphaseOnlpurchase -0.059 0.303 0.846
ExpTPphy x BphasePhysearch 0.190 0.686 0.782
ExpTPphy x BphasePhypurchase -0.012 0.143 0.934
ExpTPonl x BphaseOnlsearch 0.020 0.219 0.927
ExpTPonl x BphaseOnlpurchase -0.025 0.321 0.938
Gender 0.216 0.094 0.023*
Age -0.197 0.039 0.000*
Culture 0.461 0.311 0.140
Price image 0.041 0.040 0.305
4.5 Touchpoint interaction
To test whether a touchpoint interaction exists among physical and online store experiences, the interaction is entered in the latter (third) regression model. By adding this interaction, the proposed model from section 3.3 was built. For this procedure, the interaction was already centered. Notable is that by entering the interaction term, the touchpoint choice dummies and the moderator dummy for online store purchases dropped.
From table 7 it can be derived that it is unlikely that an interaction exists between physical and online store experiences based on changes in R2 and the model not being overall significant anymore. To formally test whether the interaction is indeed nonfactual, the uncentered interaction term is mean-centered via Hayes’ (2013) PROCESS macro, while not including all other predictors in the model to be estimated. As expected, seeing the very small changes in R2 and the model not being overall significant anymore (table 7), no interaction is found. Hence, there is no evidence to support the notion (H4) that touchpoint synergies result in a more positive customer experience.
Table 7: Possible touchpoint interaction: phy x onl (centered) - N =246.
4.6 CE differences among buying phases
For the last hypothesis (H5) will be tested whether customers have a better overall customer experience during their buying phase of ‘purchase’ in contrast to their buying phase of
‘search’. Differences are found by conducting 2 One-way ANOVA’s; i.e. for each considered touchpoint one test. Respectively is found that there exist significant differences in CE among
Regression model R2 Adj. R2 R2 change F-stat. P-value
buying phases at physical stores (F = 4.651; p < .1; .010) but not for online stores (F = 0.533; p < 1; .589). Plotting the differences graphically (figure 2 and 3) seen is that 1.) respondents had a better CE during search than purchase in physical stores and 2.) a better CE during purchase than search in online stores. Accordingly it can be stated that there is only partial evidence to support H5, as respondents only had a better CE during their buying phase of ‘purchase’ in contrast to ‘search’ at online stores.
Figure 2: CE among buying phases Figure 3: CE among buying phases for physical stores for online stores
4.7 Latent class analysis
age, gender perceived price image, perceived service quality and two touchpoint choice dummies; which indicate whether one had a certain touchpoint experience or not. Culture is excluded as covariate, given that only 5 of the respondents in the data are non-Dutch.
To ascertain that a sufficient clustering solution will be attained, the LCA model is estimated with one to fifteen clusters. The Bayesian information criterion (BIC) was used when looking for a optimal cluster solution, given its proven accuracy for detecting optimal cluster solutions for LCA in contrast to other information criteria (Nylund, Asparouhov and Muthén, 2007; Konuş, Verhoef and Neslin, 2008). Nextly, classification errors (i.e. the expected proportion of respondents being misclassified) were are also considered for detecting the appropriate amount of clusters.
Considering all clusters, a minimum is obtained for BIC (4960) at a 4 cluster solution, corresponding to a relative low classification error of .0660 (see table 8). In addition to these selection indices, this cluster solution is regarded sensible from a managerial perspective, as larger cluster solutions would create really small clusters. Descriptive statistics of the 4 cluster solution can be found in table 9.
Table 8: Log-likelihood statistics for model selection - N =246
Specifically, cluster 2 consists of customers that generally did not have an online store experience (0.054), but did have a physical store experience. In special this segment shows the most positive overall customer experience (3.65) and the highest values for NPS (8.713) and LI (87.096). Further, cluster 3 is the segment were most customers had experiences for both touchpoints. But despite multiple experiences, customers from cluster 3 have a less positive overall customer experience and postulate themselves less positive by means of CFMs. Cluster 1 is similar to cluster 2 in terms that most customers from this cluster only had a physical store experience. Nevertheless, this cluster scores lowest on all metrics. Finally cluster 4 mostly consists of individuals that only had an online store experience (0.007). Remarkably enough this cluster has a better overall customer experience compared to cluster 1. Furthermore, this cluster is more likely to recommend its important retailer to friends or colleagues but will be slightly less likely to go shopping at its important retailer again.
LL BIC (LL) Class. Err.
Table 9: Profiling final segments (LCA) - N =246 Standard R2 = .85
* Results are significant at p < .1
To be able to describe the 4 clusters more thoroughly and to subsequently label them,
covariate effects across segments are considered, which are displayed in table 10. Covariates that have a large impact on a segment are highlighted bold; i.e. strong positive values imply that customers who score high on the focal covariate, are very likely to be a member of the concerning segment. Strong negative values imply the opposite.
Table 10: Covariates of CE and CFMs (active covariates) - N =246 Standard R2 = .63
* Results are significant at p < .1
Overall, 3 out of 6 covariates are found to be significant: gender, service quality and the touchpoint choice dummy for online store experiences. Firstly looking at gender (m), it can be seen that men are most likely to represent cluster 1 and 4 given their positive signs.
Cluster 1
(41%) Cluster 2 (32%) Cluster 3 (20%) Cluster 4 (7%) P-value (χ2-test)
CE 2.786 3.656 3.411 3.105 0.00*
ExpTPphy 4.554 5.332 4.984 0.007 0.00*
ExpTPonl 0.003 0.054 4.756 4.742 0.00*
NPS 6.875 8.713 7.918 7.380 0.00*
LI 79.767 87.096 84.860 76.539 0.07*
Cluster 1 Cluster 2 Cluster 3 Cluster 4 Wald P-value
The negative signs infer that cluster 2 and 3, are most likely represented by women. Secondly, service quality strongly determines customers’ membership of cluster 2. This seems logical, as this cluster is most positive on all experience / CF metrics. Finally, individuals that had an online store experience (1) are most likely members of cluster 3 and 4 and not member of cluster 1 and 2.
By melding all findings from segment profiling and the impact of covariates, four clusters can described succinctly and are subsequently labelled. Customers in cluster 2 - the second largest cluster (32%) - that are mostly females with a physical store experience, are in contrast to customers from other clusters, very likely to be emotionally satisfied with their important retailer. It is likely attributable to a high extent of positive perceived service quality. This can be concluded based on the most positive overall customer experience, high intentions to recommend their important retailer to colleagues or fiends and a high likelihood of shopping at the focal retailer again. Hence cluster 2 is labelled as ‘Physical Retail Idolizers’ (PRI).
Cluster 3 is either principally resembled by females, but generally had physical and online store experiences. Although they had both experiences they are less positive about their important retailer on scales of experience and CF metrics in contrast to cluster 2. A less positive level of perceived service quality may elucidate this. Accounting for 20% of the respondents, this segment is labelled as ‘Critical Multi Touchpoint Shoppers’ (CMTS).
5. CONCLUSIONS & RECOMMENDATIONS
This final chapter is organized as follows: firstly, research questions (formulated in section 1.1) will be answered and discussed. Secondly, corresponding managerial implications will be outlined and thirdly, limitations of the study and directions for future research are suggested.
5.1 Discussion
That only few companies manage the accumulation of experiences across multiple touchpoints over time, is very unfortunate. Especially when considering that effective
touchpoint management may result in opportunities for building customer commitment, better retention rates and more financial success. But to reach possibilities as such, companies need to understand customers’ experiences across various touchpoints. Emotions may vary across these touchpoints (Garrett, 2006; Penz and Hogg, 2011), given that touchpoints can be viewed as a ‘referents’ that induce certain emotions. Specifically, depending on whether customers are absorbed in an activity during a touchpoint experience, through e.g. mastery of an activity or by intense focused engagement, emotional valence is determined. For example, when a customer is in a physical store from a local fashion retailer, and finds that present sales associates do not comply deliberately with complaints, negative emotions such as anger and sadness may be induced.
touchpoints interact and by how customer experience varies across buying phases.
Consequently managers and marketers can be more effective / efficient in terms of resource allocation among touchpoints (Baxendale, Macdonald and Wilson, 2015).
Results for all three regression models (section 4.4), showed that physical stores have the largest (significant) impact on customer experience in contrast to online stores. This finding is as expected, seeing that Baxendale, Macdonald and Wilson (2015) find thatin-store
communications, brand advertising and the observation of other customers (i.e. elements associated with physical stores), have the highest impact in comparison to e.g. traditional earned media. Other research also supports the impact of physical stores seeing that varying levels of emotions can be elicited through ambient store conditions e.g. music, light and layout (Shukla and Babin, 2013; Mohan, Sivakumaran and Sharma, 2013; Knoferle, Spangenberg, Herrmann and Landwehr, 2012) and seeing customers’ concerns of sensory trial prior purchase (Dholakia et al. 2010; Spence and Gallace, 2011).
But, regarding other findings in the light of this study; why was there no (full) evidence for: more positive touchpoint experiences accumulating in a better CE (H3), touchpoint synergies resulting in a more positive CE (H4) and customers having a better CE during the buying phase of purchase (H5)?
In total seven reasons are proposed: one for H3 and H4 concerted, three for H3, two for H4 and one for H5 separately. The covering argument for H3 and H4 not being supported follows logically from the data omittance of touchpoint experiences for catalogues and mobile
For H3 specifically, no evidence was found on account of online store experiences negatively impacting customer experience. Theoretical reasoning for this, firstly stems from the
disability to touch products. Not being able to touch products in online stores, means getting less sensory feedback and lowers chances of positive experience evaluation (Peck and Shu, 2009). Secondly, online stores are known for lower levels of interpersonal interactivity compared to physical stores (Florenthal and Shoham, 2010). Therefore online stores may undermine the fundamental human motivation to belong and subsequently do not have the ability to increase positive emotions (Argo, Dahl and Manchanda, 2005). Thirdly, what can be seen from the data is that online store experiences (M = 4.70) have a lower mean than
physical store experiences (M = 4.92), which may indicate inconsistencies on how important retailers delivered its products and services across channels. If present, these inconsistencies can dilute the overall customer experience (Kwon and Lennon, 2009; Rangaswamy and Van Bruggen, 2005; Berry et al. 2010).
Furthermore no touchpoint synergy effect, i.e. an interaction effect between physical and online store experiences was found. Considering the data, it is very likely that no synergy effect is found since respondents chiefly evaluated important retailers e.g. Albert Heijn, Hema and Starbucks that sell particularly convenience goods (food and beverages) (Holton, 1958), for which gain of making price and quality comparisons do not outweigh required costs in time, money and effort. Hence consumers are less inclined to shop across different
Finally for H5, partial evidence was found for a better CE during the buying phase of purchase. Only at physical stores respondents had a better CE for search in contrast to
purchase. This finding is no oddity since younger consumers (mainly present in the data) tend to enjoy explorative behavior more than older ones (Cox, Cox and Anderson, 2005). In special this holds for younger females, as they associate it with excitement (Baker and Wakefield, 2012).
5.2 Managerial implications
Considering results of the focal study, this section compiles research findings into
implications that are worthwhile for retail managers / marketers to comply with nowadays. Findings indicate that physical stores are from great importance when retailers want to deliver an extraordinary CE. These touchpoints are the places were customers get their senses fully stimulated and thus are more likely to get emotionally engaged with the retail brand. This may happen as a result of friendly sales associates, unique aesthetical environments and
demonstrations of how products are made and/or how services work. Important to note is that recent research (Pauwels and Neslin, 2015) shows that spending more budget to physical stores by opening more, not necessarily results in cannibalization of other touchpoints. Opening more physical stores may even increase net revenues approximately with 20%. Despite the experiential importance of physical stores, the importance of online stores should not be underestimated. These can be namely complementary touchpoints in the sense that they provide convenience that cannot be provided by physical stores. For example, online stores provide opportunities to: find products / services quickly, price comparison and
However, to truly know whether touchpoints are complementary (i.e. touchpoint synergies) or rather substitutes (i.e. touchpoint dissynergies) for customers, retailers should dive into the world of customers’ desires and expectations. Herewith especially expectation management is important, as former experiences build expectations that are implicitly set. Capitalization on understanding expectations as such, will make retailers capable to more accurately guide and meet customers’ expectations (Spence, 2012). No capitalization may result in expectations that are not congruent with the occurred experience(s), which very likely lead to the overall CE to have negative valence.
To overcome touchpoint incongruence, retailers should not treat touchpoints as they were silos. This is proven by LCA, as there is a clear segment that tends to use physical and online stores. Hence there is evidence retailers should not emphasize on single touchpoint strategies, but should integrate touchpoints, so consistent experiences can be provided across all
touchpoints. To specifically support reduction of possible inconsistencies, retailers may integrate customer data across the company (Rangaswamy and Van Bruggen, 2005).
Furthermore, in line with touchpoint integration and given partial support for better customer experiences during buying phases of purchase, retailers should not solely focus on delivering the best customer experience on this point. Customers may have experiences during all
5.3 Limitations and future research
Viewing this full study, certain limitations should be exemplified for thorough interpretation, which provide avenues for further research.
Firstly, this study focussed on one particular industry: the B2C retail sector. As a result, generalizability (Malhotra, 2010) of findings cannot be directly assumed across industries. Moreover, also customers’ characteristics in the study sample limit generalizability:
respondents were primarily Dutch females (65%) within the age category of 18-29 years old. Given industry and customer differences: in B2B settings the web is expected to be superior (Merrilees and Fenech, 2007) due fast access, efficient search and ordering capability. Intuitively, these are necessary ingredients that emphasize its potential importance, when compared to B2C contexts, relatively quicker and more costly decisions have to be made. Demographically, men are expected to experience lower levels of emotional intensity
compared to women (e.g. Fujita, Diener and Sandvik, 1991). In other words, when measuring CE by customer delight, men are expected to have a less positive CE in contrast to women. Accordingly future research is impelled to conduct cross-sectional research among B2B, B2C, masculine and feminine settings.
Secondly data are pure cross-sectional, which might have limited power to draw conclusions about relationships between variables. Therefore, to increase power of corollaries upon relationships between variables and given that CE varies over time (Frow and Payne, 2007; Rose, Clark, Samouel and Hair, 2012; Heinonen et al. 2010), one should measure CE on multiple intervals over time by means of longitudinal research (Verhoef et al. 2009).
include after sales touchpoints. Hence future studies alike should include touchpoints such as e.g. social media and customer service.
Fourthly it may be argued whether customer delight was the appropriate construct to measure CE on the emotional level (Dixon et al. 2010; Oliver, Rust and Varki, 1997),since its only positive in valence and measures emotions indirectly. Future research should measure emotions more explicitly while incorporating both negative and positive valence.
Nevertheless, when one is planning to do so, emotions should be measured directly when an experiential event occurs, as accuracy for emotional recall may decrease over time (Cowley, 2008). Recommend is to apply a similar technique like real-time experience tracking (RET) (Baxendale et al. 2015).
Finally, three methodical remarks can be made in terms of survey design: 1.) For the
important retailer identification question, nearly all respondents selected given options to be their important retailer, which might have biased results; the selected retailer may not be truly important them. Hence an open question would have been better as in this case respondents had to really think about which retailer is important to them. 2.) Not many respondents did search or purchase products/services across multiple touchpoints. This caused that e.g. H2 was not testable anymore and that conclusions about H5 should be approached cautiously. More respondents with multiple touchpoint experiences could have been collected when the important retailer identification question also stated; ‘were one shops in multiple channels’. Another way to overcome the fact that not many respondents did utilize multiple touchpoints for search and purchase, is that future research could study high involvement
