1
Helping in-store shoppers to make the right purchase decision with
the aid of OCR’s.
A research to different influencers of perceived usefulness of experiential products in
store.
2
Helping in-store shoppers to make the right purchase decision with
the aid of OCR’s.
A research to different influencers of perceived usefulness of experiential products in
store.
Iris Weelink
Department: Marketing
Master thesis
January 13, 2020
Schoppenweg 8a, 7136KH, Zieuwent
i.w.weelink@student.rug.nl
0625568117
S2909227
Supervisor: J.A. Voerman
2
ndsupervisor: J.C. Hoekstra
3
Marketing summary
This research tries to explain the influence from different factors on the perceived usefulness of information. The specialty beer category is a central topic in this research. In de past decade, the consumption of specialty beer has exploded, and consumers face more variety than ever before. On the one hand, this fits to consumer’s demand as experiential products, like specialty beers, demand more variety to overcome satiation or boredom resulting from repeated consumption of these products. On the other hand, more variety is not always better, as consumers face a lot of choice, they could suffer from choice overload and perceive difficulties in choice making. Some researches advice to reduce the assortment, although this does not stroke with the consumer’s demand in the experiential product category. Therefore, this research is aimed to find decision making aids to simplify decision making for consumers.
Based on literature, different types of information could influence the perceived usefulness of information. First, the type of information influences the perceived usefulness of information, where I expect experiential search attribute information to have a stronger relation. Within this variable, the type of sender as second influencer, is nested. It implies that only in the condition of experiential attribute information, there are two types of senders who could have influence on perceived
usefulness as well. The sender of information could be a peer consumer or an expert, where I perceive experts to have a stronger influence on the perceived usefulness of information. Furthermore, these variables are influenced by the level of variety and variety does have a main relation to perceived usefulness of information as more variety causes choice overload and therefore people in high variety situation will benefit more from information and thus may perceive it as more useful. This is also the reason why variety strengthens the main relation from type of information and type of sender. Lastly, based on literature, decision making strategy will influence the relation
between type of information, type of sender and the perceived usefulness of information. I expect maximizers to perceive information as more useful because they are suffering more from choice overload than satisfiers, as they search for the best option and often examine the whole product range to search for this best product.
The research has been carried out through two surveys; a pretest and the main survey. The pretest has been conducted to determine the manipulations of high and low variety and to determine the manipulation of an expert.
4 senders of information (peer consumers/experts). The survey contains questions about perceived usefulness of information, questions about the decision-making strategy of respondents and it contains manipulation check questions and control questions as well.
5
Preface
First of all, a great thanks to my supervisor Liane Voerman. Without her, I would never been able to complete this thesis. Due to her enthusiasm, criticism and support, I’ve written my thesis with a lot of pleasure. Thanks to her great sense of humor, her personal appreciation and interest in my thesis and personal life, the meetings and short consults were always great fun.
Furthermore, I want to thank my colleagues at Royal Grolsch for their support and insights. Through them, this very interesting topic has been brought to my mind and due to developments in the company I really saw the value of doing this research. Brands and companies are always looking for guiding their shoppers and above all, help their shoppers in making good choices. Especially category management is about compose the best assortment for consumers. The category managers of Grolsch are really aimed at helping consumers in making the right choice of beer, and they have inspired me in investigating how to guide shoppers in their decision making in store.
6
Chapters
1. Introduction ... 8
1.1 Booming Beers ... 8
1.2 Variety Backfiring ... 8
1.3 Different types of information ... 9
1.4 Different senders of information ... 10
1.5 Problem statement ... 11
2. Theory ... 12
2.1 Different types of information ... 12
2.2 Different senders of information on experiential attribute ... 13
2.3 Level of variety ... 14
2.4 Choice making strategy ... 15
2.5 Conceptual model ... 16
3. Research design and methodology ... 18
3.1 Type of research ... 18
3.2 Participants and design ... 18
3.3 Pretest ... 19
3.3.1 Pretest of the expert manipulation ... 20
3.3.2 Pretest of the variety manipulation ... 21
3.4 Procedure and experimental variables in the main test ... 21
3.4.1 Procedure of main test ... 21
3.4.2 Experimental variables in the main test ... 22
3.5 Operationalization of variables ... 24
3.5.1 Perceived usefulness of information ... 25
3.5.2 Decision making strategy ... 26
3.5.3 Manipulation Checks ... 27
3.5.4 Control variables... 27
3.6 Manipulation checks ... 27
3.6.1 Variety ... 28
3.6.2 Search attribute information ... 28
3.6.3 Experiential attribute information ... 28
3.7 Plan of analysis ... 29
4. Results ... 33
4.1 Insights in the data ... 33
7
4.2.1 First Conceptual Model ... 34
4.2.2 Second Conceptual Model... 35
4.2.3 Full Conceptual Model ... 36
4.3 Discussion of results ... 38
5. Conclusion & Discussion ... 40
5.1. Limitations and Further Research ... 41
7. Literature ... 43
8. Appendix ... 50
8.1 Shelf information Jan Linders – 2019 ... 50
8
1. Introduction
1.1 Booming Beers
In the Netherlands, the beer category is divided roughly into three segments: lager, non- and low-alcoholic beers and specialty beers (De Nederlandse Brouwers, 2019). The percentage of alcohol is one of the main differences between the categories: lagers are the ‘regular pilsner beers’, containing around 5% alcohol, low- and non-alcoholic beers contain 0-4% alcohol, while specialty beers contain mostly more than 5% alcohol. Furthermore, specialty beers are well known from their deviant ingredients (compared with lager), such as special types of wheat, malt and hop and special flavor additions from natural sources like citrus fruit or herbs. In 1990, 96% of the beer consumption in the Netherlands was lager (Huizing-Charlak, 2019). Due to the lack of variety in choice, most of the consumer choice was limited to lager. Van Dijk et al. (2018) call this a ‘Pilsner desert’, because of homogeneity, even professional tasters could not distinguish between the different beers in blind taste tests. In the past thirty years the consumption in the Netherlands shifted to other beer-styles, i.e. non- and low-alcoholic beers and specialty beers (De Nederlandse Brouwers, 2019). Non- and low-alcoholic beers are already 6% of the total consumption. Moreover, the consumption of specialty beers has exploded over the past years and raised to a share of 12% of the beer consumption
(Huizing-Charlak, 2019). The highest variety ever has been measured in (Van Dijk et al., 2018). In the segment of specialty beers, consumers can find a lot of different beers; only in the last three years the amount of Stock Keeping Units (SKU’s) has grown from 134 till 176 SKU’s per specialty shelf on average, an increase of 24% (IRI, 2019). So, not only the interest in specialty beers in growing, the variety in choice increased as well.
1.2 Variety Backfiring
Specialty beers are an example of experiential products. Experiential products are defined as: “Ones which consumers choose, buy and use solely to experience and enjoy, with the main benefit from these products being pleasure or hedonic value in consumption” (Cooper-Martin, 1992).
9 However, more variety is not always better, if the benefits do outweigh the cost of more effort in decision making, consumers may switch to simple decision making. Furthermore, they may even decide not to choose at all, because of struggling choice overload (Schwartz, 2002). Hence, several studies advice to avoid choice overload by reducing the assortment (e.g., Sloot et al., 2006; Chernev, 2012; Huffman & Kahn, 1998). Nevertheless, if consumers desire variety in experiential product categories, an assortment reduction may not be the best option to simplify decision making and improve customer satisfaction. Retailers could also opt to reduce the complexity and help consumers to make a good choice. After all, the key to customer satisfaction with the entire shopping
interaction, is to ensure that the customer is equipped to handle the variety (Huffman & Kahn, 1998). The goal of helping consumers with optimal decision making is to reduce the effort of decision
making, even if there are a lot of alternatives. In conclusion, there are different developments in the specialty beer category take place: consumers show more interest in this category over the past years, and the variety on the shelf is very large. On the other hand, previous research shows consumers having difficulties when confronted with a lot of choices. The combination of these aspects makes it an interesting topic to investigate, how can retailers handle this combination of factors?
1.3 Different types of information
With the rise of online retail, brick and mortar stores try to enhance their physical store environment in order to compete with the omni-channeling retail world. When consumers are in store, retailers can, in order to improve in-store decision making, trigger them with in-store stimuli, like nudging (Inman et al., 2009; Salmon et al., 2014; Houdek et al., 2018). These are simple interventions to help people overcome their cognitive limitations in decision making, i.e. physical changes in the
environment of the product, to change the choice architecture, to subtly guide one’s decision in a particular direction (Thaler & Sunstein, 2008). Examples of nudging are; placing products at eye-level, using displays in-store and placing products at the check-out desks. Information displays can be helpful for consumers to be better informed without more effort (Kleinmuntz & Schkade, 1993). The given information must be accurate and useful because, if consumers are provided with much (unnecessary) information in a situation when they already suffering from limited cognitive capacity, it could overwhelm them (Ungemach et al., 2015). Hence, helping consumers to make the right choice, can be done by giving correct information about the products (Ungemach et al., 2015).
10 attributes play an important role in the consumer decision making process, as consumers evaluate and compare products based on their attributes. Furthermore, unique product attributes are important to stand out from competitors and provide a basis for marketeers to differentiate their product (Belch & Belch, 1995). There are different product attributes, according to Nelson (1974) and Darby & Karni (1973): search attributes, experience attributes and credence attributes. Search attributes are those attributes which can be verified through direct inspection or readily available resources, prior to purchase (Nelson, 1974). Experience attributes can be verified only after use of the product (Nelson, 1974). Credence attributes are difficult to verify, even after use of the product; i.e. if consumers do not possess the technical capabilities to evaluate the product or service, like complex medical services of automobile repairs (Darby & Karni, 1973). Different product attributes can help consumers in their decision-making process. Different sources of information in the
specialty beer shelf can be matched to different types of information attributes. First, information of the different beers is conveyed by standard product attributes, for example; the taste-profile, the beer style, the color of the beer and the alcohol percentage (Jan Linders, 2019). These are examples of search attributes; consumers don’t have to consume the product to verify these product
attributes. Secondly, information of the different beers can be conveyed through experiences of other consumers, for example Untappd, a well-known app to rate beers. With this app you can check-in beers you’ve tasted and rate them on a scale from 1-5, further you can give comments or add flavor profiles. Untappd makes it possible to share your ratings with friends, moreover it saves all ratings and combine them to rate a specific beer (Untappd, 2019). Untappd is an example online consumer reviews (OCR’s). OCR’s are any positive or negative statements made by potential, actual or former customers about their experiences, evaluations and opinions about the product of service (Park & Park, 2008). When consumers search online products, OCR’s can help them as a guide in their decision process by reducing the amount of effort (Smith et al., 2005). OCR’s as source of information are an example of experience attributes, as it conveys the experience of users after consuming the product. For this research I assume these two types of attributes will provide information about products to consumers.
1.4 Different senders of information
In an environment with a lot of variety and therefore a lot of stimuli, consumers are likely to rely on opinions of others. This can be a strategy to simplify the decision making (Sasaki et al., 2011). Therefore user-generated information may be helpful in making purchase decisions in such
11 Not only the type of information can influence the perceived usefulness, also the identity of the sender matters (Nelson, 1974). When consumers are aware the information is given by an interested party (like via an advertisement of a retailer), this should lower the effect on demand, compared to a neutral party (Friberg & Grönqvist, 2012). Neutral parties are independent from manufacturers or retailers who sell products. Those neutral parties can convey their experience through online reviews of the products. Generally speaking, there are two types of online reviews: user-generated reviews based on personal experiences, and reviews written by professional editors, i.e. experts. (Zhang et al., 2010). They can convey their experience to other customers (Zhang, 2010). So, experiential information from others has a strong effect on customers however, what is the effect of OCR’s specifically in high variety environments? In addition, how does this differ between two senders of this information: peer consumers and experts?
1.5 Problem statement
Logically, a lot of research to the use of OCR’s in e-commerce has been done (Zhang et al., 2010; Smith et al., 2005; Zhu and Zhang, 2010; Park et al., 2007) However, to my knowledge not so much on the use of OCR’s in store environments. Whereas consumers who search for experiential products in in-store environments are also confronted with a lot of variety and therefore can be overwhelmed and encounter choice overload. If retailers can help these consumers in guiding their choice, it may affect their image and eventually, sales. So my research question is: “What is the effect of
information on different types of attributes given by either peers or experts on the perceived usefulness of information, when making a product choice of experiential products in an assortment of different kinds of variety in an in-store environment?”
Sub-questions following from the main question are:
1. What is the influence of the type of attribute information on perceived usefulness of information of experiential products in an in-store environment?
2. What is the influence of the sender of experiential attribute information on perceived usefulness of information of experiential products in an in-store environment?
3. What is the effect of the level of variety on the relation of attribute-type of information on the perceived usefulness of information of experiential products in an in-store environment? 4. Which other variables influence the relation of attribute type of information on the
12
2. Theory
In this chapter, I will tell more about the underlying theories and concepts that determine the relation of different variables on the dependent variable, the perceived usefulness of information. The three main variables are: type of information (information on search attribute level or on experiential attribute level), the sender of information on experiential attribute (a peer-consumer or an expert) and the level of variety (high or low). Furthermore, other variables influence this relation. In the last section of this chapter I will tell something about their influence.
2.1 Different types of information
As determined in the introduction, this research will focus on two levels of type of information. The search attributes and experiential attributes differ in their capability to convey an experience. Search attributes are those attributes of products consumers can inspect prior to purchasing the product (Nelson, 1974). Tasting alcoholic beverages in-store is not possible, because of alcohol legislation, consumers have to inspect products without tasting directly. They can read product information to find out the type of beer, the alcohol percentage, etc. According to Friberg &
Grönqvist (2012) search characteristics are not enough to provide a complete picture of experiential products. In other words; you have to experience it.
The consumer has to purchase these goods to evaluate them (Nelson, 1974). This is called the experience process (Nelson, 1974). In a high variety environment, like the specialty beer category, it is highly unlikely to purchase all beers and taste them to know which one you prefer. Hence, it seems like consumers have to deal with information based on others, like search attribute information or experiential attribute information, without experience the product yourself. Experiential information helps consumers to gather information to base their decision upon, as experiential attributes can give a sense of a consumption experience (Cooper-Martin, 1992). To share consumption experience, experiential information can be conveyed through online consumer reviews. Through these reviews, consumers might get a sense of product experience and therefore will be more informed and more able to evaluate products without buying them all.
13 be right’ (Cialdini & Goldstein, 2004; Salmon et al., 2015). A form of social conformity is social proof; if a person is unsure of the correct way to behave, they will often look at others for clues/proof concerning the correct behavior, motivated by the wish for accuracy (Cialdini, 1984; Castelli, 2001). Salmon et al. (2015) research the effect of social comparison on purchases in-store. Their research is based on nudging consumers to more healthy choices. The outcome of the research is evidently; when consumers are in a state of low self-control, they are more likely to buy low-fat cheese when this product is associated with the social proof heuristic (Salmon et al., 2015). Social proof is communicated by the slogan: ‘Most sold in this supermarket’, because providing information about the majority of a reference group is a way to manipulate social proof (Goldstein et al., 2008). Salmon et al. (2015) manipulate self-control of participants by conducting a concentration task. Since high-variety environments are already depleting due to complexity of high-variety and limited cognitive capacity (Houdek et al., 2018) it might be expected people will also rely on social proof when confronted with a high variety environment. Flavian et al. (2016) research the interaction between online and offline channels on the search experience of consumers. By using this combination, the purchase intentions in store were higher than towards a competitor (Flavian et al, 2016).
Furthermore, reading positive reviews online before visiting the store may help consumers to have a more stable preference for the product, making other alternatives less desirable. Lastly, they
investigate and prove reading an online review during the psychical interaction improved the participants’ experience and consumer satisfaction. Furthermore, OCR’s help consumers reduce uncertainty associated with the product, leading to make purchases with a higher degree of confidence (Flavian et. al, 2016; Zhu & Zhang, 2010).
So, I expect the experiential attributes to add valence to the decision-making process of consumers.
H1: The perceived usefulness of information will be higher when the information is based on experiential attributes, compared to search attributes.
2.2 Different senders of information on experiential attribute
14 by consumers (Willemsen et al., 2011). The research of Zhang et al. (2010) shows user-generated information, like online consumer reviews, to be perceived as more credible compared to expert reviews. The research take place in the hospitality sector, like searching for a hotel or restaurant. According to their research, user-generated reviews caused more traffic on the website of different restaurants than expert reviews.
Nevertheless, a problem of user-generated content could be the anonymity with which individuals can post content, as it can lead to questions about the legitimacy of such ratings (O’Connor, 2008). Experts could have a greater impact than other channels because of their authority, their perceived legitimacy and because they don’t have a financial interest in rating the product. (Friberg &
Grönqvist, 2012(. Hilger et al. (2011) demonstrate, by using expert-ratings in store, not only the sales but also the demand of high-scoring wines increases, compared to low scoring wines. Expert ratings affect consumer decision through the provision of quality information, this distinguish it from peer consumer reviews (Hilger et al., 2011). Since wine is a comparable experience product to beer, this is likely to happen in the beer category as well. People have a natural tendency to obey authorities (even if the circumstances are lurid) (Milgram, 1975). So, experts can be a valuable source of information and therefore consumers might perceive their opinion about certain products as more useful. Important to note in this case is that, when experts are perceived as relatively less credible, their review is perceived as less valuable (Friberg & Grönqvist, 2012). Finally, Willemsen et al. (2011) confirm expertise claims to have a positive effect on the perceived usefulness of a review. Therefore, I expect in my research experts as senders of experiential information to be perceived as more useful.
H2: The perceived usefulness of information will be higher when the sender of the experiential information is an expert, compared to a peer consumer.
2.3 Level of variety
In case of a lot of variety, it is harder for consumers to make a good choice. Due to limited attention and cognitive resources, choice overload appears. Hence, people are not able to use all available information and the freedom of choice effectively to their own best interests (Houdek et al., 2018). As the number of alternatives increases, decision makers tend to switch to simple, less accurate decision strategies because the process of selection, evaluation and integration of information is too extensive in case of high variety (Iyengar & Lepper, 2000; Payne, 1982). Consumers are more
15 Salmon et al. (2015) already proves; if consumers are in a state of low self-control, they are more likely to buy a product associated with a social proof heuristic. Being in a high variety environment, causing choice overload is relatable to a low state of self-control, as cognitive resources are depleted. Hence, consumers are more likely to rely on others’ opinions (Sasaki, 2011). Opinions of others can give a sense of a consumption experience (Cooper-Martin, 1992), therefore consumers might need less cognitive capacity to imagine the consumption. Therefore, high variety seems to strengthen the positive main effect of experiential attribute information on the perceived usefulness of information.
H3: The perceived usefulness of information will be higher when there is high variety in the assortment, compared to low variety.
H4: The positive effect of experiential attribute information versus search attribute information on the perceived usefulness of information will be stronger when there is more variety.
H5: The positive effect of the sender (of experiential attribute information) being an expert, compared to a peer consumer, on the perceived usefulness of information will be stronger when there is more variety.
2.4 Choice making strategy
The relation between the dependent variable and independent variables depends on other variables as well. Whether consumers have a maximizing strategy or satisfying strategy in choice making, determines how they cope with a lot of variety in choice. Furthermore, it can determine how they are influenced by an expert or peer consumers.
16 process of maximizers (Schwartz, 2002). They rely upon a few pieces of information with which they feel can guide and help them to decide how the product might perform relative to other products or brands (Akpoyomare et al., 2012). Maximizer are more likely to suffer from choice overload, as they focus more on how different option compare relative to each other than absolute performances of option (Fasolo et al., 2009). Furthermore, maximizers will have more difficulties with larger
assortments, as they want to examine all alternatives (Schwartz et al., 2002). Satisfiers’ decision-making process, on the other hand, will not be affected much by adding assortment because they probably their choice was already in the small assortment (Schwartz et al., 2002). Unfortunately, nor Fasolo (2009) nor Schwartz et al. (2002) spends attention on helping maximizers in making their decision process easier. However, maximizers could benefit even more from it, as they will search for the best option. If others already experience it and claim it to be good, it might be more comfortable for maximizers to choose for those products. Through communicating experiential attribute
information, e.g. OCR’s, they can use less pieces of information than the whole product information. Others can help them determining the best choice. Furthermore, consumers may develop a sense of online community by using OCR’s, through mutually sharing product information and exchanging shopping experiences (Rese et al., 2014). When the sender of the online review feels familiar, they are more likely to be followed. The social exchange theory advocates information sharing and trust of community to be relational bonds between exchange parties to cope with risks and opportunistic behaviors. OCR may therefore foster consumers’ trust in online intermediary (Shi & Liao, 2017). Because maximizers have the intention to evaluate all possible options, they might perceive peers as more convincing, as multiple peer consumers review the product. A review of an expert is the opinion of a single person and might be not thorough enough for a maximizer.
H6: The positive effect of experiential information versus search attribute information on the
perceived usefulness of information will be stronger when the respondent has a maximizing strategy compared to a satisfying strategy.
H7: The positive effect of the sender (of experiential attribute information) being an expert, compared to a peer consumer, on perceived usefulness will be weaker when the respondent has a maximizing strategy compared to a satisfying strategy.
2.5 Conceptual model
Figure 1 gives a graphic idea of the relations, discussed in the hypotheses above.
17 influence. Only when the * type of information is on experiential attribute level, the distinction between ** the sender of the information: a peer consumer or an expert is present, e.g. this relation is only possible if the type of information attribute is experiential, compared to a search attribute. This has also a direct effect on the perceived usefulness of information. The level of variety
intensifies the effect between the type of information and the perceived usefulness. I expect more variety will lead to a stronger relation, as variety causes choice overload and therefore people are relying more on the experiential attribute information others provide them with.
Lastly, the choice making strategy, whether someone is a satisfier or maximizer, will influences their decision-making process and thus their preferences for different types of information and senders.
Figure 1: Conceptual model
18
3. Research design and methodology
In this chapter I will explain the structure of the empirical research, the type of research, which population has been sampled and how I’ve reached them. Furthermore, the design and procedure will be explained. The pretest and main test will be discussed separately, starting with the pretest. Thereafter the description of the experimental variables and how they are manipulated will be explained as well, followed by the operationalization of the dependent and other variables. Lastly the plan of analysis will be discussed.
3.1 Type of research
Given the research questions and relations in the conceptual model, I’ve chosen for a causal research design. If the goals of the research it to determine and understand the relationship of the causal variables and effects to be predicted, causal research is appropriate (Maholtra & Birks, 2007). This requires a quantitative, because it enables statistical analysis to test the hypothesis, based on data (Maholtra & Birks, 2007).
3.2 Participants and design
The target population consist of Dutch shoppers (male/female) who are able to buy specialty beers, so they will be above 18 years, due to the legal drinking age of 18 in the Netherlands. The sampling method is convenience sampling, because it is the easiest way to get as quick as possible a lot of respondents (Maholtra & Birks, 2007). A drawback of this method is that it might not represent the entire population. Hence, external validity can be a problem and therefore the quality of the research can be suboptimal (Etikan, 2016). To protect the research for suboptimal quality, I’ve contacted people beyond my own network via other people. Through the snowball-effect I have reached not only students between 20 and 25, but also working people of 40 years and older. I have added control questions to check this distribution. Participants are randomly assigned to the different conditions of an 2x2 design (see table 1) with variety: high versus low and type of information: search attributes versus experiential attributes. Two types of sender are nested within experiential
19 Table 1: Overview of different manipulation conditions.
Initially 294 respondents filled in the survey. 67 of them did not finish the survey, they are pulled out of analysis, as they did not finish the manipulation questions and therefore their response is not usable. This brings the set of respondents back to 227. 19 respondents were kicked out the survey because they were under legal drinking age of 18 years. It left a set of 208 respondents to analyze. 73 (35.1%) men and 135 (64.9%) women filled in the survey. The youngest respondent is 18 years old and the oldest is 64 years old, the average age of the respondents is 29.13 years. The survey
contained an attention check question. Initially the question was: “Please fill in disagree for this question”. In the first two days I got some messages from people who did not understand the question. They thought it was a trick and they consciously filled in another option. Most of them filled in ‘neutral’. I have decided to change the question into: “This is a control question, please fill in disagree for this question”. After this change, no respondent filled in the question the wrong way. In total 22 respondents filled out the attention check wrong. Their responses on the dependent variable did not differ significantly (p = 0.233) from other responses. Furthermore, none of them rushed the survey. Hence, I’ve decided to still count the responses of people who filled out the question wrong.
3.3 Pretest
In order to choose the correct operationalization of the manipulation of variety and experts as sender of the experiential information, I will start with a pretest. Participants will be approached directly through a message send via Whatsapp. First, respondents will see an explanation of the survey: to research what they perceive as an expert review and how they experience the degree of variety in different shelfs.
Then they will see four different descriptions of experts and they are asked to rank the perception of an expert per description on a scale from 1-100. Additionally, they will see four different shelfs, divided over 4 different windows in random order. Per shelf, they have to rank on a Likert scale from 1-7 they perceived variety (number of choice options) in that particular shelf.
The pretest includes 21 respondents. Based on these results, the manipulations in the main test will be built.
Variety level Search attributes Experiential attributes Peer consumers Expert
Low variety 1 3 5
20 3.3.1 Pretest of the expert manipulation
The literature is contradicting about which form of an expert label people perceive as an expert. Zhang et al. (2016) have used the Chinese travel website Qunar for their investigation to the effect of expert reviews. At Qunar, the expert reviews get a clear and visible label, indicating they are an expert. The research of Zhang et al. (2016) is executed for the tourism branch, which is, like beer, an experiential product (Cooper-Martin, 1992). Also, Racherla & Friske (2012) use an expert label (a star) to indicate professionals giving a review. Other researches focus on well-known experts for that specific type of product, e.g. wine-experts or movie-experts, both examples of experiential goods as well (Reinstein & Snyder, 2005; Hilger et al., 2011). Because of the contradictions in literature about the manipulation of an expert review I’m going to pretest these stimuli. Approximately 20
respondents have to rank how they perceive different descriptions of experts as actual experts. In the pretest be 4 options will be presented, so the respondents will have more freedom of choice. Based upon the expert label (Zhang et al., 2016; Racherla & Friske, 2012) and specific experts for certain products (Reinstein & Snyder, 2005; Hilger et al., 2011) 4 labels will be presented: “Expert”, “Bierexpert”, “Beste Biersommelier” and “Pepijn van der Waa – Beste Biersommelier van
Nederland”.
Table 2: Analysis pretest on Expert label
Label Expert Bierexpert Beste Biersommelier Pepijn van der Waa – Beste Biersommelier
van Nederland
Mean 56.57 61.57 48.33 72.33
St. Dev. 24,71 27,91 26,67 24,98
Table 2 shows “Pepijn van der Waa – Beste Biersommelier van Nederland” to be the best
21 3.3.2 Pretest of the variety manipulation
Malone and Lusk (2017) research decision making in a high-variety context of buying beers. Their study took place in a wine bar and they focus at beer sales. In terms of variety they distinguish between 6 and 12 different beers (Malone and Lusk, 2017). To verify this representation of variety in the shelf, a pretest has been conducted. 4 shelfs with different number of SKU’s have been shown to respondents, in order to check whether 6 and 12 SKU’s differ enough from each other. Based on the study of Iyengar and Lepper (2000), a test to investigate if people find it harder to choose between 6 or 24 jams, the respondents in the pretest will also see a shelf with 24 SKU’s. For the lowest variety condition, I’ve chosen for 4 SKU’s. Because it was easy to divide the shelf in 4 parts, and this makes the appearance of the shelf calmer compared to the 6 SKU’s. So, the respondents in the pretest will see the 4 different shelfs with 4, 6, 12 and 24 different SKU’s. They are asked to rank perceived variety on a 7-point Likert scale.
Table 3: Analysis pretest on variety
4 SKU’s 6 SKU’s 12 SKU’s 24 SKU’s
Mean 3.14 4.24 5.24 6.62
St. Dev. 1.153 0.831 0.995 0.590
As table 3 shows 4 SKU’s were perceived as containing least variety, with an average of 3.14. 24 SKU’s were perceived as containing most variety, with an average of 6.62. To analyze if the different indications of variety differ from each other, I conducted an ANOVA analysis. The test was significant (p=0.000). Hence, I have chosen to follow the result of the pretest instead of following existing literature. The low variety condition will contain 4 different SKU’s and the high variety condition will contain 24 SKU’s.
3.4 Procedure and experimental variables in the main test
To test the different conditions of the research design, I conducted a survey. This section will explain the structure of the survey. Furthermore, it will explain the operationalization of the different manipulations.
3.4.1 Procedure of main test
22 mentioned people can win a giftpack of specialty beers if they leave their e-mail address. An
incentive does increase response rates, without decreasing the quality of the data (Janke, 2014). So, it is a smart way to get more respondents. Lastly it will be told that the research is confidential, the data will be used only for this research and respondents can step out anytime if they want.
The survey starts on the next window. Respondents will see first one of the variety manipulations (high or low variety), determined by the scenario in which the respondent is. The next screen contains one of the information manipulations. If the respondent indicates he/she has read the information, the next window appears where they will be questioned about their perceived
usefulness of information. Subsequently, manipulation check questions will be asked. Then, decision making strategy questions will follow. The last questions will be about the control variables, gender and age. Finally; participants can (optionally) leave their e-mail address to get a chance to win the giftpack filled with beers.
3.4.2 Experimental variables in the main test
The manipulated conditions are based upon three different kinds of stimuli, representing the independent variables in the conceptual model: the amount of variety, the type of information and the type of sender.
Variety
To manipulate the variety, participants in different conditions will see different levels of variety. I have chosen to follow the result of the pretest instead of following existing literature. The low variety condition will contain 4 different SKU’s and the high variety condition will contain 24 SKU’s. Figure 5 and 6 shows how respondents will see these shelfs.
23 Type of information
The type of information conveyed to the participants will be either information on search attributes, which shoppers can inspect prior to purchasing the product, or information on experiential
attributes, which shows an opinion. In the search attribute information condition, the information will be conveyed through factual information about the following characteristics: : style of specialty beer (for example: blond, triple, double, IPA etc.), color (from light to dark), alcohol percentage and type of taste. Figure 2 shows this manipulation, based on the shelf of retailer Jan Linders, whom has been chosen for seven years in a row as best retailer in the specialty beer category; consumers liked the findability and given information about the specialty beers (GFK, 2019). To control the
experiment, participants the experiential conditions will be exposed to the same characteristics as in the search attribute condition. As said before; online consumer reviews, the opinion of other
consumers or experts, can convey a sense of experience, so in the experiential attribute conditions, respondents will see the information through describing an opinion. Figure 3 shows which
information on experiential attribute the respondents will see. Figure 3 will not be added in the survey, because experiential attribute information will be conveyed through an expert or a peer consumer.
Figure 2 – information on search attribute Figure 3 – information on experiential attribute
Type of Sender
24 Nederland”, figure 4 shows the picture respondents will see. Peer consumers will be indicated with the ‘Untappd community’, i.e. all people who checked in and reviewed a specific beer (Untappd, 2019). Figure 5 shows which picture respondents in this manipulation condition will see. Because Untappd is a well-known review app for beers and has over 300.000 users in the Netherlands, which is the most in Europe (Untappd, 2019), and therefore a common way to represent peer consumers.
Figure 4 - Expert label Figure 5 – Peer consumer label
3.5 Operationalization of variables
In addition to manipulations of variables, other variables will be measured through different items. Table 4 shows the different questions to measure different constructs. Furthermore, it contains the manipulation check questions and control variable questions. The column reliability shows the Factor analysis’ and Reliability analysis’ outcomes.
Table 4: Operationalization of variables
Concept Source Items Reliability
Perceived usefulness of information Park & Gretzel (2010)
1. This would improve my shopping experience 2. This would make my shopping trip more efficient
3. This would enable me to accomplish my shopping goals more quickly
4. This would be something I find useful 5. This would help me with my decision making
EV = 3.13 α = 0.849 Decision making strategy Choice Making Difficulty Malone et al. (2019)
1. No matter how satisfied I am with my job, it’s only right for me to be on the lookout for better opportunities
25 2. When I Am in the car listening to the radio, I
often check other station to see if something better is playing, even if I am
relatively satisfied with what I’m listening to. 3. When I watch TV, I channel surf, often
scanning through the available options even while attempting to watch one program.
4. I treat relationships like clothing: I expect to try a lot on before finding the perfect fit.
5. I often find it difficult to shop for a gift for a friend.
6. Choosing is film to watch is very difficult. I’m always struggling to pick the best one.
7.When shopping, I have a hard time finding clothing that I really love.
8. I’m a big fan of lists that attempt to rank things (best movies, best singers, best athletes, best novels, etc.).
9. I find that writing is very difficult, even if it’s just writing a letter to a friend, because it’s so hard to word thig just right. I often do several drafts of even simple things.
10. I never settle for the second best.
11. Whenever I’m faced with a choice, I try to imagine what all other possibilities are, even ones that aren’t present at the moment.
12. I often fantasize about living in ways that are quite different from my actual life.
13. No matter what I do, I have the highest standards for myself.
Manipulation checks 1. This shelf of specialty beers contained a lot of variety (choice options).
2. The information shown was factual.
3. The information shown was an opinion of an expert
4. The information shown was an opinion of peer consumers.
Control question 1. What is your age? 2. What is your gender?
3.5.1 Perceived usefulness of information
26 consumer (Park & Gretzel, 2010). This is somehow comparable with my research on different ways of conveying attribute information. Park & Gretzel (2010) differ from my research, because it is
conducted in an online environment. Nevertheless, I think the measurement scale they used, is also applicable for my research. To analyze if the scale is applicable for my research, I will execute a Factor Analysis and Reliability Analysis for the 5 items (see Table 4) in this scale. Factor Analysis is appropriate, given KMO = 0.824 and Bartlett’s test is significant (p=0.000). The scale to measure the perceived usefulness of information can merge to 1 factor, the total EV is 3.13 and explains 62.7% of the variance. The reliability analysis shows an alpha of 0.849. So, the dependent variable ‘perceived usefulness of information’ consist out of five items, which are represented in one factor.
Perceived usefulness has a mean of 5.1 on a Likert scale of 7. There is variance in the responses, with a standard deviation of 1.048. However, it shows more loading on the left side of the distribution, indicated by the mean of 5.1. Looking at the skewness of the distribution, it shows a Skewness is - 0.935, the standard error of Skewness is 0.204. The Z-score deriving from this is -4.583. Hence, it can be concluded that the distribution of the dependent variable is left side skewed, i.e. in every
condition, the perceived usefulness scored relatively high.
3.5.2 Decision making strategy
The decision-making strategy of a shopper can influence the causal relation of conceptual model. Therefore, I’m going to measure the decision-making strategy of every respondent to see if differences in causal relationships between satisfiers and maximizers exist. To measure a
respondent’s decision making strategy, I use the same questionnaire as Malone et al. (2019), they research the decision making process of consumers in a bar, they manipulate the variety and measure the influence of the decision making on the buying behavior of respondents (Malone et al., 2019). The scale is based on extensive research from Simon (1978), Schwartz (2002) and
27 3.5.3 Manipulation Checks
To check if the manipulations are perceived as intended, I will add manipulation checks to my survey; questions about how respondents perceive the manipulations as I wanted them to perceive it
(Morton & William, 2010). I will make use of factual manipulations checks because I want to know if they perceived the manipulation in a right way (Kane & Barabas, 2019). Four different manipulations have been checked: the perceived level of variety, the perceived level of getting information from search attributes, the perceived level of getting information via the opinion of an expert or the perceived level of getting information via the opinion of peer consumers. Respondents can rank the manipulation checks from disagree to agree on a 7-point Likert scale.
3.5.4 Control variables
In order to control for possible confounding effects of personal characteristics I will measure the age and gender of the respondents. Age and gender are common control variables.
To fill in the survey, the respondent has to be 18 years or older. Because I expect a lot of people from my own network to fill in the survey, their age categories should not be leading in the research. This applies also on gender, because I expect more women to fill in the survey, compared to men.
3.6 Manipulation checks
The outcomes of the manipulations checks on variety, type of information and type of sender are described in table 6, followed by the test descriptions per manipulation check.
Table 6: ANOVA analysis on manipulation checks
Mean Standard Deviation t-test
Perceived level of variety 4.54 1.68 -1.9 *
Perceived level of information on
search attributes 4.57 1.40 0.582
Perceived level of information via
opinion of expert 4.42 1.45 0.622
Perceived level of information via
opinion of peer consumer 3.70 1.51 -3.887**
28 3.6.1 Variety
To check whether the respondents have perceived the variety as intended, I conducted an independent samples t-test (see table 6) to test whether difference answers on the manipulation check between the two groups within variety (high or low) exist. This test was 10% significant (p = 0.059). Hence, the manipulation check for variety shows a significant difference between people perceiving variety in low variety and high variety condition. So, respondents in the high variety condition perceived significant more variety than respondents in the low variety condition.
3.6.2 Search attribute information
To check whether the respondents have perceived the search attribute information as intended, I conducted an independent samples t-test (see table 6) to test whether difference answers on the manipulation check between the two groups within type of information (search or experiential) exist. This test was not significant (p = 0.561). Hence, the manipulation check for search attribute
information does not show a significant difference between respondents in the different conditions of type of information. The mean of respondents in the search condition is 4.66, the mean of respondents in the experiential condition is 4.53. This might implicate people in the experiential condition to perceive the online consumer review they see as factual information. Remarkable is the significant difference (p = 0.027, t=2.230) between respondents who saw information coming from a peer consumer or expert of the manipulation check of search attribute information. The mean of this question, when the sender of information is a peer is 4.79, the mean when the sender is an expert is 4.28. So, people in the experiential condition, given the sender is a peer, perceive the information as more factual, compared to the sender being an expert.
3.6.3 Experiential attribute information
To check whether the respondents have perceived the sender as intended, I conducted an independent samples t-test (see table 6) to test whether difference answers on the manipulation check between the two groups within the sender of information (a peer consumer or an expert) exist. The test for the manipulation check for the peer consumer was significant (p = 0.000). Hence, a significant difference of the perceived sender has been shown between the different conditions of experiential information. Respondents who saw a peer consumer review perceived this significantly more as a peer consumer than respondents who saw an expert review do.
29 mean of the respondents in the Peer consumer condition is 4.59. So, respondents might perceive peer consumers as experts.
3.7 Plan of analysis
In order to analyze the data, derived from the survey, several steps should take place. In this section I will explain the conducted steps and the several terms and conditions of the analysis.
To use the measurement scales as variable, Factor Analysis needs to be conducted to see if the different items could be merge into 1 factor. Factor Analysis is appropriate if KMO > 0.5 and Bartlett’s test of Sphericity is significant with a confidence interval of 95% (Maholtra & Birks, 2007). The
number of factors deriving from the items, depends on specific criteria, namely; all Eigenvalues which are higher than 1, or when the total variance explained is more than 60%, or those factors which explain more than 5% each (Maholtra & Birks, 2007). In this research, I will examine the Eigenvalues > 1 and the total variance explained >60%. After conducting Factor Analysis, the new factor needs to be checked with a Reliability Analysis, the scale is reliable if the alpha is higher dan 0.6. (Maholtra & Birks, 2007). For the dependent variable; perceived usefulness of information, a skewness analysis will be conducted. To analyze the Z-score, I will assume a Z-score bigger than – 3 to be indicated as negatively skewed (Tabachnik & Fidell, 2005).
30 if the type of information is on experiential attribute. By making two conceptual models, this
separation will be made clear. Figure 6 illustrates the first conceptual model and figure 7 illustrates the second conceptual, which contains the nested variable type of sender of experiential attribute information. Again, this only appears if the information is on experiential attribute.
Figure 6: First conceptual model, containing type of information
Figure 7: Second Conceptual model, containing type of sender
Hence, to analyze this model, two regressions have been conducted. Subsequently, one big regression to measure the whole model has been conducted. To analyze the full model, the
31 variables have been made for the 3 independent variables. The first dummy consists of 0 = search attribute information and 1 = sender of the experiential attribute information is an expert. The other dummy consists of 0 = search attribute information and 1 = the sender of experiential attribute information is a peer consumer.
Figure 8: Full conceptual model, containing three levels of type of information.
Every regression analysis will be divided in different steps; first I will look at the main effect from the independent variables perceived usefulness of information. Secondly, I will look at the same relation and I will add the interaction-effect of these independent variables. The next step is to look at the interaction-effect of the moderator; choice making difficulty. Lastly, I will analyze the full model, including the control variables.
The first conceptual model contains the analysis of type of information and variety as independent variables. The second conceptual model contains the analysis of sender of experiential attribute information and variety as independent variables. Lastly, for the full conceptual model, I will start with the main relations from variety and type of information (3 levels) on perceived usefulness. The different steps will be defined by the following formulas, described in table 7.
To deal with the multicollinearity that arises by the addition of the moderator Choice Making
32 Table 7: formulas to describe the regression analysis models
First analysis Second analysis Full analysis Model 1 Y = a +b1X1 + b2X2 Y = a + b3X3 + b2X2 Y = a + b4X4 + b2X2 Model 2 Y = a +b1X1 + b2X2 + b5(X1X2) Y = a + b3X3 + b2X2 + b5(X1X2) Y = a + b4X4 + b2X2 + b5(X1X2) Model 3 Y = a +b1X1 + b2X2 + b5(X1X2) + b6M1 X1 Y = a + b3X3 + b1X1 + b5(X1X2) + b6M1 X1 Y = a + b4X4 + b1X1 + b5(X1X2) + b6M1 X1 Full Model Y = a +b1X1 + b2X2 + b5(X1X2) + b6M1 X1 + A + G Y = a + b3X3 + b1X1 + b5(X1X2) + b6M1 X1 + A + G Y = a + b4X4 + b1X1 + b5(X1X2) + b6M1 X1 + A + G
X1 = Type of Information: search attribute information versus experiential attribute information
X2 = Variety: high versus low
X3 = Type of sender: expert versus peer consumers
X4 = Type of information: search attribute information versus expert (experiential attribute information) versus
peer consumers (experiential attribute information) M1 = Choice Making Difficulty
33
4. Results
This chapter contains a summary of the results of the research. First, I will give an overview of the different conditions and how they relate to the dependent variable: the perceived usefulness of information. Thereafter, with the results of the regression analysis, I will answer the different hypothesis.
4.1 Insights in the data
Table 2 shows the average score on the dependent variable per condition, i.e. to what extent people perceive the information next to the shelf as useful, when in different variety conditions.
Table 7: Means on Perceived Usefulness of Information, based on different manipulation conditions Variety level Search Attributes Experiential attributes Total
Expert Peer
Low variety 4.93 5.08 5.04 5.01
High variety 5.21 5.21 5.11 5.18
Total 5.06 5.15 5.07
The two-way ANOVA analysis shows no significant (p = 0.574) interaction effect between the type of information and the level of variety. So, the average scores on perceived usefulness of information did not differ between the different conditions of variety and type of information.
Furthermore, two-way ANOVA analysis shows no significant interaction effect (p = 0.850) between the two different senders of information within the experiential attribute information (peer
consumers or experts). Hence, average scores on perceived usefulness of information did not differ between the conditions of different senders or different levels of variety.
This can be explained by the left side skewness of distribution of the dependent variable; perceived usefulness of information. This skewness and average score per scenario showing a relatively high perceived usefulness of information in every scenario.
4.2 Regression analysis
34 The second conceptual model consists of the sender of information, variety, the moderator choice making difficulty and the control variables age and gender. Figure 10 illustrates this model. The full conceptual model consists of the type of information (3 levels, including the sender), variety, the moderator choice making difficulty and the control variables age and gender. Figure 11 illustrates this model.
Figure 9 Figure 10
Figure 11
4.2.1 First Conceptual Model
Table 8: Regression analysis on first conceptual model
Model 1 Model 2 Model 3 Full model
Type of Information B = 0.039 B = 0.126 B = 0.174 B = 0.219
Variety B = 0.162 B = 0.281 B = 0.230 B = 0.163
Type of Info * Variety B = 0.176 B = - 0.124 B = - 0.076 Choice making
difficulty
B = 0.141 B = 0.147
35 * = 90% significant
** = 95% significant
To analyze if experiential type of info and high variety will lead to a higher perceived usefulness of information (see figure 9), regression from type of information and variety on perceived usefulness of information has been conducted. The results show no significant main effect, with a R2 = 0.006 and F = 0.664. After adding the interaction effect of variety on the type of info, the regression has been conducted again. Still the results are not significant, with R2 = 0.008 and F = 0.541. The interaction-effect causes more collinearity (VIF = 4.235) but still doesn’t explain if experiential information will cause a higher perceived usefulness. The full regression from Type of Information to the Perceived Usefulness of Information with moderators Variety and Choice making Difficulty, still show no
significant effect, with R2= 0.030 and F = 1.240. In this model, Type of Information and the interaction between Type of Information and Choice Making Difficulty have both a high VIF-score; respectively 11.11 and 12.87. This indicates multicollinearity and it causes standard errors to be higher. Again, both Type of info and the interaction between Type of info and Choice making Difficulty shows respectively a VIF of 11.19 and 12,99. By adding the control variables in this model, it turned out to be that age does not have a significant (p = 0.848) effect, with B = 0.001 on perceived usefulness of information. The other control variable, gender, does have a significant effect (p = 0.041), with B = -0.316 on the perceived usefulness of information. Gender has a VIF score of 1.03. Hence, it does not contribute to multicollinearity and does not cohere with other variables. The effect from gender on perceived usefulness of information stands on his own. Although, the full model is still not significant with R2= 0.050 and F = 1.507.
4.2.2 Second Conceptual Model
Table 9: Regression analysis on second conceptual model
Model 1 Model 2 Model 3 Full Model
Type of Sender B = - 0.071 B = 0.134 B = - 0.459 B = - 0.452 Variety B = 0.098 B = - 0.033 B = 0.122 B = 0.075 Type of Sender * Variety B = 0.071 B = - 0.029 B = - 0.030 Choice making difficulty B = 0.062 B = 0.078 Type of Sender * Choice making diff.
36 * = 90% significant
** = 95% significant
To analyze if experts as sender of the experiential information or high variety will lead to a higher perceived usefulness of information (see figure 10), regression from the sender of information and variety on perceived usefulness of information has been conducted. These results show no significant effect on the main relations, R2 = 0.003 and F = 0.230. By adding the interaction effect of variety, again no significant effect has been shown, R2 = 0.004 and F = 0.164. By including the Choice Making Difficulty as moderator to this regression, the relation is not significant with R2= 0.027 and F = 0.754. In this model, the VIF of Type of Sender and the VIF of the interaction between the Type of Sender and Choice Making difficulty are respectively 10.3 and 10.7. So, both causing multicollinearity. When adding the control variables in this model, and therefore testing the whole model it results in both gender and age not having a significant effect on perceived usefulness of information. With respectively p = 0.170 and B = -0.286 and p = 0.529 and B = 0.005. The full model is not significant with R2= 0.044 and F = 0.868.
4.2.3 Full Conceptual Model
After the two independent regressions, I have attempt to analyze the full model in one regression analysis. A new conceptual model has been structured for this regression to redistribute the IV’s (see figure 11). It becomes a 2x3 research design, with two levels of variety and 3 levels of type of
information. Because it changed to a 2x3 research design, ANOVA can be conducted. ANOVA has been conducted with Variety and Type of Information (including sender). The test was not significant (p = 0.873). Figure 12 shows the distribution of the different means of information on search
attribute, information from an expert and information from a peer consumer. Although not
significant, this figure gives some interesting insights. It shows information from experts to perceived as most useful. Information on search attribute makes the greatest step up in when confronted with a high variety situation. And finally, information on all types of information attributes is perceived as more useful in high-variety situations.
Age B = 0.005
Gender B = - 0.268
R2 0.003 0.004 0.027 0.044
37 Figure 12: Means on Perceived Usefulness of Information, based on different manipulation conditions according to conceptual model in figure 11.
0.00 = search attribute information
1.00 = experiential attribute information, sender: expert
2.00 = experiential attribute information, sender: peer consumers
* = 90% significant ** = 95% significant
Model 1 Model 2 Model 3 Full Model
Type of information a B = 0.007 B = 0.112 B = - 0.038 B = - 0.009 Type of information b B = 0.072 B = 0.144 B = 0.421 B = 0.483 Variety B = 0.157 B = 0.281 B = 0.230 B = 0.161 Type of information a * variety B = - 0.218 B = - 0.136 B = - 0.051 Type of information b * variety B = - 0.147 B = - 0.108 B = - 0.101 Choice making difficulty B = 0.141 B = 0.147 Type of information a *Choice making diff.
B = 0.035 B = 0.020
Type of information b * Choice making diff.
B = - 0.079 B = - 0.086
Age B = 0.001
Gender B = -0.320**
R2 0.007 0.009 0.034 0.054
38 To analyze the full model and research which kind of information will lead to the highest perceived usefulness, regression has been conducted from the different types of information and variety to the perceived usefulness of information. Table 10 shows these results. The main effect from the type of information or variety on perceived usefulness of information is not significant, with R2 = 0.007 and F = 0.484. By adding the interaction effect between the different types of information and variety, the relation remains not significant, with R2= 0.009 and F = 0.364. By adding the choice making difficulty, the model is still not significant with R2 = 0.034 and F = 0.868. The multicollinearity of this model is high, driven by the different information variables and their interaction with the choice making difficulty variable. Due to the coherence between choice making difficulty and the independent variables; type of information and sender of experiential attribute information, the model contained less information and B is harder to explain. In order to analyze the direct effect from the moderator choice making difficulty on perceived usefulness without the multicollinearity effect it causes, an ANOVA analysis with variety, type of information and the median-split dummy of choice making difficulty. ANOVA shows a significant difference in choice making difficulty on perceived usefulness of information (p = 0.012). It seems to have a direct effect on perceived usefulness of information, nevertheless, this has not been proven in the regression analysis. The analysis of the full model, including the control variables gender and age, is still not significant, with R2 = 0.054 and F = 1.128. The direct relation from age is not significant, but the direct relation from gender on perceived usefulness is significant (p = 0.041). The correlation analysis of gender and the dummy of choice making difficulty shows no significant effect (p = 0.917).
4.3 Discussion of results
Based on the results, it’s not possible to accepted any of the hypotheses of this research (see table 11), as none of the tested relations to perceived usefulness of information are significant. An interesting, serendipity result, is the significant direct effect form gender on perceived usefulness in both the first conceptual model as the full conceptual model and close to significance in the second conceptual model. It means women perceive information as more useful, compared to men. Besides, the direct significant ANOVA effect from choice making difficulty on perceived usefulness of
39 Although the interaction effect of variety was not significant, it can be cautious mentioned that people in a high variety condition benefit more from information than people in a low variety condition. Figure 12 shows the perceived usefulness of information te be higher for every type of information in the high variety condition, compared to low variety. Unfortunately, this relation is not significant, so real conclusions cannot be made. Moreover, according to the manipulation checks, people in the low-variety condition perceived this as relatively high variety, i.e. they perceived variety is higher than intended, as shown by a mean of 4.32 and no significant difference in response
between the two groups (high and low variety). Hence, it makes sense they perceive information to be useful. If they perceive it as high variety they benefit more from the information, as the
hypothesis experts. Unfortunately, this conclusion is nothing more than speculations based on insignificant data and graphs.
Table 11: Hypotheses outcomes
Hypothesis Accepted / rejected
1. The perceived usefulness of information will be higher when the information is based on experiential attributes, compared to search attributes.
Rejected
2. The perceived usefulness of information will be higher when the sender of the experiential information is an expert, compared to a peer consumer.
3. The perceived usefulness of information will be higher when there is high variety in the assortment, compared to low variety.
Rejected
4. The positive effect of experiential attribute information versus search attribute information on the perceived usefulness of information will be stronger when there is more variety.
Rejected
5. The positive effect of the sender (of experiential attribute information) being an expert, compared to a peer consumer, on the perceived usefulness of information will be stronger when there is more variety.
Rejected
6. The positive effect of experiential information versus search attribute
information on the perceived usefulness of information will be stronger when the respondent has a maximizing strategy compared to a satisfying strategy.
Rejected
7. The positive effect of the sender (of experiential attribute information) being an expert, compared to a peer consumer, on perceived usefulness will be weaker when the respondent has a maximizing strategy compared to a satisfying strategy.
40
5. Conclusion & Discussion
In this chapter I will derive conclusions from the results of the research. First, I am going to answer the main question, thereafter I will explain it by answering the separate sub-questions.
The main question is: “What is the effect of information on different types of attributes given by either peers or experts on the perceived usefulness of information, when making a product choice in a situation of a high variety assortment of experiential products in an in-store environment?”
Although none of the hypotheses can be accepted, some cautious conclusion might be made. All different types of information attributes show positive effects on perceived usefulness, unfortunately not significant. The main effect to perceived usefulness of information remains unclear, so it will not be clear which type of information attribute, given by either peers or experts, will influence the perceived usefulness of information the most. Both expert reviews and peer consumer reviews were perceived as moderately useful. This is in line with the existing contradicting literature about this subject (Zhang, 2010; Willemsen et al., 2011; Friberg & Grönqvist, 2012; Hilger et al, 2011). Both search attribute information as experiential attribute information, i.e. experts and peer consumers as sender of the information could be perceived as useful senders of information for experiential products. These outcomes assume any kind of information, by a shelf with experiential products, to be perceived as useful. This might be explained through the development in consumers that has been take place the 21st. There has never been more access to information than now. Due to the digital developments and media, consumers are always informed, about everything they are interested in. This created a ‘new consumer’ who is really demanding lot of information to make a deliberate choice. These shoppers use different touchpoints to get information about the product and base their decision on the availability of information and variation (Mosquera et al., 2017). They are informed, connected and interacting with both the company and other customers to get
information (Mosquera et al., 2017). Because customers demand more information and are more satisfied with their decision-making process if information is available, it could explain why all different types of information are perceived as relatively useful.
Every type of attribute information is perceived as more useful in a high variety situation, this is not significant, so real conclusions cannot be made. But a cautious derived conclusion should be;
