E-‐grocery delivery
Consumer preferences and willingness to pay
Author: Aneta Rabljenovic
E-‐grocery delivery
Consumer preferences and willingness to pay
Aneta Rabljenovic
Faculty of Economics and Business
MSc. Marketing Intelligence
January 11, 2016
Groningen, the Netherlands
Nieuweweg 25a, 9711TB Groningen
+31 646187699
Student number: 2604620
a.rabljenovic@student.rug.nl
First supervisor: Erjen van Nierop (j.e.m.van.nierop@rug.nl)
Second supervisor: Keyvan Dehmamy (k.dehmamy@rug.nl)
MANAGEMENT SUMMARY
After many successful years of online shopping for clothes and electronics, it was not the question whether but when we would also be able to shop for groceries online. For some people it is still a bit odd, but online grocery shopping is becoming more and more popular. One retailer after the other is entering the online grocery business and almost every day a new development in this field is shared with the world. So it speaks for itself that this topic is very recent and new research is more than welcome. Therefore the aim of this study is to get more insight in online grocery shopping and especially the link with the delivery of the groceries. Determining the willingness to pay for the delivery of groceries and providing managers with useful implications will help them when targeting different segments in this field.
With help of the literature and news articles, important variables and their relationships with the willingness to pay for the delivery of groceries are determined. After that, a choice based conjoint analysis is performed to test the formulated hypotheses. These hypotheses have been tested for different segments. According to the literature, assortment size has a positive effect on peoples’ willingness to pay for the delivery of groceries. However, the assortment should not be endless, because people cannot deal with too much choice. This is only confirmed by one segment. Other three segments indicate that they want a large assortment. Delivery time and delivery timeframe should, according to the literature be as short as possible, which will ensure for a higher willingness to pay. This is confirmed by almost all segments, simply because people do not want to wait and expect the service to be as best as possible. The other two variables are the distance from the supermarket to a consumers’ house and the amount of scheduled grocery trips per week. These two variables turned out not to be significant and therefore no conclusion could be made about them in this study. However, according to the literature, the distance from the supermarket to a consumers’ house should have a positive effect on the willingness to pay of the delivery of groceries. The amount of scheduled grocery trips per week should weaken/strengthen the negative relationship between delivery time and willingness to pay for the delivery of groceries.
Finally, the willingness to pay for the delivery of groceries is calculated four times, for four different segments. The average willingness to pay is €8.80.
PREFACE
“It always seems impossible, until it’s done!” – Nelson MandelaThis thesis is written during the last phase of my Master Marketing Intelligence at the University of Groningen. I have experienced these past five months as very educational, stressful and yet very organized. Obviously, after all the hours I have put into writing my thesis, I am very glad to present you this final project!
First of all, I would like to thank my first supervisor, Erjen van Nierop, for giving me very useful feedback and guidance, and being very patient. Furthermore, a great thank you goes to my group members Jojanneke Dierssen and Jan Nies for their feedback and helpful discussions. Last but not least, I would like to thank my family and friends for putting up with me and supporting me through this busy period.
TABLE OF CONTENTS
1. INTRODUCTION ... 5
2. THEORETICAL FRAMEWORK ... 7
2.1
Online grocery shopping ... 7
2.2
Delivery of groceries ... 8
2.2.1 Important drivers of delivery service ... 8
2.2.2 Important hurdles of delivery service ... 9
2.3
Willingness to pay (WTP) ... 9
2.4
Willingness to pay for the delivery of groceries ... 10
2.4.1 Assortment size ... 11
2.4.2 Delivery time ... 12
2.4.3 Delivery timeframe ... 13
2.4.4 Distance to the nearest offline supermarket ... 14
2.5
Conceptual model ... 15
3. RESEARCH DESIGN ... 16
3.1
Research methods ... 16
3.1.1 In-‐depth interviews ... 16
3.1.2 Conjoint analysis ... 16
3.2
Data collection: conjoint analysis ... 17
3.2.1 Segmentation ... 17
3.2.2 Attribute levels ... 17
3.2.3 Choice design ... 18
3.3
Plan of analysis ... 20
3.3.1 Assessment of model fit ... 21
3.3.2 Latent class analysis ... 21
4. RESULTS ... 22
4.1
In-‐depth interviews ... 22
4.2
Descriptive data: all respondents ... 23
4.3
Descriptive data: ‘only no-‐choice respondents’ excluded ... 23
4.4
Conjoint analysis ... 25
4.4.1 Parameters ... 26
4.4.2 Model specification ... 27
4.4.4 Model fit ... 29 4.4.5 Segment parameters ... 29 4.4.6 Covariates ... 30 4.4.7 WTP per segment ... 32 4.4.8 Segment profiles ... 33 4.4.9 Hypotheses testing ... 36
5. CONCLUSIONS & RECOMMENDATIONS ... 37
5.1
Theoretical implications ... 37
5.2
Managerial implications ... 39
5.3
Limitations and suggestions for further research ... 40
REFERENCES ... 42
1. INTRODUCTION
Twenty years ago, only futurists had an idea of what the Internet would mean for the world today. Now, almost everybody is familiar with this big network. Over 40% of the world population has access to the Internet today (Internet Live Stats, 2014) and in the Netherlands alone, this number is 96%. This percentage of Internet users gives retailers large opportunities to sell their products everywhere (Zhang, Farris, Irvin, Kushwaha, Steenburgh, & Weitz, 2010), as a result 71% of the Dutch population shops online. Although online shopping is very developed in the Netherlands, online grocery shopping is lagging behind. When comparing the Dutch e-‐grocery market to the UK, Germany and France, the Netherlands lays behind with 1.5% of the total grocery retail market in the country. However, with a growth rate of 55% in 2014, the Dutch e-‐grocery market was the fastest growing among these four European countries (Syndy, 2015). According to Gorczynski and Kooijman (2015), the online grocery market in the Netherlands will continue to increase for the next five years due to home delivery services and further introduction of pick-‐up points. Although retailers prefer pick-‐up points because of the lower cost, 80% of the Dutch consumers prefers home delivery services (Morganosky & Cude, 2002; Syndy, 2015). Even though retailers prefer pick-‐up points, it is also important to understand why consumers prefer home delivery and respond to this preference. For this reason, this thesis will only focus on home delivery of groceries.
Since such a big percentage of the Dutch population is willing to order its groceries online, it is very striking that the Dutch e-‐grocery market is only 1.5% of the total grocery retail market in the country. The advantages and disadvantages of online grocery shopping play a major role here. Multiple studies (Childers, Carr, Peck, & Carson, 2001; Galante, López, & Monroe, 2013; Gorczynski, & Kooijman, 2015; Seitz, 2013; Syndy, 2015; Zhang et al., 2010) have mentioned that convenience is the most important factor for people to order groceries online. Hence, people can order from their own home and whenever they want to. It also saves time, because the customer does not need to go to the store when the groceries are being delivered at home (Bakos, 1997; Burke, 1997). On the other hand, one of the biggest reasons not to order groceries online are the delivery costs. Even though 80% of the Dutch consumers is open to home delivery and is willing to pay approximately €4.00 for it, this is still too low. This because most of the delivery costs for groceries in the Netherlands are much higher (Syndy, 2015). Also, the fact that people are ‘not able to touch, feel or smell the groceries’ is seen as a big disadvantage of online shopping. Consumers don’t have the same opportunity to check the quality of the product as when they are shopping in a traditional brick-‐and-‐mortar store (Keh, & Shieh, 2001).
large part, on the delivery timeframe that indicates the time window in which the groceries are being delivered (Elsevier, 2015). As mentioned before, Dutch consumers are only willing to pay approximately €4.00 for the delivery of their groceries (Syndy, 2015). Willingness to pay indicates the maximum amount of money a customer wants to pay for the product or service and, in this case, the home delivery (Miller, Hofstetter, & Krohmer, 2011). To cover the costs, some retailers ask more for the delivery (Deloitte, 2014), which clashes with the willingness to pay of the customer. On the other hand, the retailers that do ask less for the delivery find it very hard to maintain their business profitable. Although it is very hard to make profit on the home delivery of groceries, more and more retailers are still entering this market. Some of them do not even charge delivery fees at all (Zakelijk Nieuws & Ondernemerstools, 2015). Therefore it is necessary to analyze the situation and to find out what is important for companies when entering the e-‐grocery delivery business. In addition, the topic of home delivery for e-‐grocery retail has not been studied extensively, which is a major reason to focus on this subject.
Having introduced online grocery shopping and home delivery service, the following research question is derived:
What amount of money are consumers willing to pay for the delivery of their groceries?
In order to answer this main research question, 3 sub-‐questions have been used.
1. What are the drivers and hurdles for customers when having their groceries delivered? 2. Which are the most important variables that affect the willingness to pay for the delivery of
groceries?
3. Which factors have an impact on the relationship between the variables mentioned in sub-‐ question 2 and willingness to pay for the delivery of groceries?
The structure of this thesis is as follows: chapter two contains the theoretical framework, which addresses the research questions and the theory to create the conceptual framework. The chapter thereafter explains the choice of the research and data collection method, followed by chapter four that continues the analysis process and discusses the outcome. The thesis ends with chapter five, which answers research questions to be able to make conclusions and give recommendations.
2. THEORETICAL FRAMEWORK
To address the different research questions and bring theories together, this chapter consists of a literature review. Because this field of research is relatively new and has not been studied thoroughly, practical reports have been used as an addition to scientific studies about this subject. First the important definitions and their relations will be explained, which will result in various hypotheses. The outcomes will all help and build the conceptual framework, which is displayed at the end of this chapter.
2.1 Online grocery shopping
Gorczysnki and Kooijman (2015) state that online grocery shopping is expected to increase due to the growing amount of pick-‐up points and home delivery services. If this growth would endlessly continue, it would be at the expense of offline grocery shopping. According to interviewed managers in the same study, this would only be the case when online sales are 10-‐15% of the total grocery share. This statement is being rejected in the same paper. First of all, the majority of the consumers will always prefer going to a physical supermarket. Only select groups of people will prefer the online offer, for example disabled consumers, businesses and the younger generation, because they grew up with the Internet and are more familiar with online shopping. Secondly, when one switches to the online store, most customers will still go to the same retailer. When people start ordering their groceries online, they will first rely on their offline experience because online shopping is new to them. It could be difficult to sort out how the website works, so they will choose the familiar retailer and rely on their previous experience when it comes to the assortment (Melis, Campo, Breugelmans, & Lamey, 2015). Third, a big part of the distribution will happen from the supermarkets, which means that a big part of the online revenue will be kept there.
2.2 Delivery of groceries
Even though online grocery shopping is still in its infancy, consumers have a clear preference for the delivery of their groceries. As mentioned in chapter one, 80% of the Dutch consumers prefers home delivery of their groceries over a pick-‐up point (Morganosky & Cude, 2002; Syndy, 2015). Compared to the previous year, preference for picking up the groceries at a supermarket or pick-‐up point has decreased while the preference for delivery service has actually increased (Deloitte, 2014; Morganosky, & Cude, 2002). The next two paragraphs will briefly give an overview of drivers and hurdles of delivery service. The section thereafter will discuss most of the drivers and hurdles in more depth.
2.2.1 Important drivers of delivery service
As mentioned in the previous section, convenience is the major reason for consumers to order their groceries online. It is also a big advantage for the home delivery service of groceries because people do not need to go out for their groceries. People with a busy schedule can save time because they do not need to go to the supermarket. There are no long queues at the checkout and, for the disabled customers, no carrying of heavy bags (Gorczysnki, & Kooijman, 2015). Therefore, next to saving time, people are also free from the physical activities.
seems that four-‐membered families with the oldest child around 12 years have a high need for convenience. Thereby, the age of the household head lies between 35 and 54 years old. This would fit in the picture that families, with parents that work and have young children, have a busy life so they would rather use the delivery service because they do not have enough time or do not want to waste time in the supermarket. Also according to Verhoef and Langerak (2001), dual-‐incomes and single parent households experience time pressure, which is a valid reason to make use of the home delivery service. Burke (1997) adds to this that it is appealing to elderly, sick or disabled people as well. Furthermore, people who do not enjoy going to the supermarket would prefer ordering their groceries and having the convenience of the delivery service.
2.2.2 Important hurdles of delivery service
The most important reason for people not to use the home delivery service is because of the extra costs (Galante et al., 2013). Many customers are willing to pay approximately €4.00 but not much more (Syndy, 2015), which is often not profitable for retailers. Also, waiting for the groceries to be delivered is seen as a major restriction of the delivery service. People have to indicate the timeframe when they have the opportunity to be at home and wait for the delivery (Colla, & Lapoule, 2012; Kumar, Kalwani, & Dada, 1997).
2.3 Willingness to pay (WTP)
As shortly explained in the introduction, willingness to pay indicates that the consumer will purchase the product or service, providing that the price is less than, or equal to, a stated amount (Miller et al., 2011). In other words, the willingness to pay suggests the value of the product or service to the customer (Cameron, & James, 1987). Miller et al. (2011) also state that indicating consumers’ willingness to pay is very important when formulating strategies and developing new products and services. In this case, retailers need to know how valuable several aspects of the home delivery service are for consumers, in order to know what their overall willingness to pay is.
involvement of consumers is an important factor that informs retailers on how people value its products and services. Of course, expensive products have a higher level of involvement, but expensive does not necessarily refer to high involvement (Fennis & Stroebe, 2010). Some people associate high quality with high price and make their decisions based on this assumption (Steenkamp et al., 2010). Price is still the most important factor when choosing a supermarket. This year, 38% of the consumers suggests that price is more important than the quality, location, assortment or the atmosphere in the supermarket. However, this percentage has decreased by 4% compared to last year, as quality, location and the atmosphere have become more important for customers (ING, 2015). In their paper about price sensitivity, Chu et al. (2008) have discussed different studies. On one hand, customers seem to be more price sensitive online compared to offline. This is due to the easier comparison of prices and the lower combined effect of price and promotion online than offline. On the other hand, the authors state that households experience lower price sensitivity online compared to offline (also Melis et al., 2015). One of the reasons is the time pressure that people could feel when shopping online. Also, shopping lists and non-‐price information on the website would distract the attention from price.
2.4 Willingness to pay for the delivery of groceries
Guiffrida, 2014). This also creates problems in terms of profitability. Therefore, the upcoming section will address the factors that influence willingness to pay for the delivery of groceries in more detail.
2.4.1 Assortment size
According to Zhang (2010), customers choose different channels depending on the assortment. These include the number of brands within a product category and the amount of product categories, resulting in the depth and breadth of the assortment. Likewise, customers, who are looking for very different products, might use more channels as long as one channel does not provide everything that they need. Because convenience is such an important factor of ordering groceries online (Childers et al., 2001; Galante et al., 2013; Gorczynski, & Kooijman, 2015; Seitz, 2013; Zhang et al, 2010), people will not put effort into ordering their groceries at multiple stores. Also, as delivery costs are seen as a big problem (Galante et al., 2013), consumers will especially not be willing to pay a few times for the delivery costs as a result of ordering at multiple stores. Thus, a bigger assortment offers more products, which gives consumers a better opportunity to actually find the product(s) they are looking for. Melis et al. (2015) confirmed some previous research, stating that assortment attractiveness is an important factor when it comes to the purchase intention in an online shopping environment. So, one could say that the bigger the assortment, the better. This statement has partially been confirmed by a study of Mogilner, Rudnick and Iyengar (2008). They suggest that on one hand consumers have a big need for variety and thus as many choice options as possible.
customers the feeling that they control what they choose, which is very important for a lot of people (Mogilner et al., 2008). As a result, next hypothesis is formulated:
H1: Assortment size has a positive effect on the willingness to pay for the delivery of groceries, which becomes less positive after a certain point.
2.4.2 Delivery time
According to Seitz (2013) one household spends circa 200 hours per year on grocery shopping. Morganosky and Cude (2002) noted that “time saving” is an important reason for consumers to order groceries online. So, when ordering their groceries, consumers don’t need to go to the store, which makes them save time. Ordering online of course also takes time, because people still need to place their order. This is often seen as a complex task that, especially the first (few) time(s), tends to be time-‐consuming (Boston Consulting Group, 2015; Hansen, 2005). Even though online shopping for groceries also takes time, it is much less time-‐consuming than going to the store and buying your groceries there (Morganosky, & Cude, 2002).
Delivery time refers to the time from ordering the groceries until the groceries are delivered at the customers’ door. Just like low delivery costs, short delivery time is always desired (Han et al., 2015). If necessary, it could even be possible to deliver the groceries within one or two hours (Zakelijk Nieuws & Ondernemerstools, 2014), but this would be very expensive for the retailers. According to a consumer research from Deloitte (2014), approximately 54% of the Dutch consumers prefers the groceries to be delivered in the evening. One-‐third of the respondents favour Friday and 41% does not have a preference for a delivery day. Especially people who work, have a busy life and/or need convenience probably prefer delivery in de evening or on their day off (De Ondernemer, 2008). According to Chintagunta et al. (2012), fast delivery is most important for fresh products. For example, when someone orders fruit, the retailer cannot wait days to deliver after having the order packaged. This is not the case for longer shelf life products, as they can easily be stored before delivering it. As already mentioned, convenience is one of the main reasons to choose for home delivery of groceries. According to Berry et al. (2002) consumers need convenience because they want to save both time and physical effort. They also suggest that it is commonly acknowledged that people are willing to pay more money for convenience (also Seitz, 2013). This results in the following hypothesis:
H2a: Longer delivery time causes lower willingness to pay for the delivery of groceries.
easily fall back on the traditional grocery shopping. They will not order this one pack of milk, but rush to the nearest supermarket, because it is more realistic to assume that going to the supermarket is faster than the extra fast desirable delivery time of the milk. This fast delivery time is not important when consumers still have groceries in stock. The big challenge here is that a lot of consumers do not think this way. They shop when they run out of groceries, which means that they need their delivery as soon as possible. This results in 60% of Dutch consumers doing groceries two to four times per week, while once per week is seen as the most economical way (ING, 2015). The small group of people that does plan their groceries once a week is less frequently exposed to the products, which leads to less unplanned purchases (Colla, & Lapoule, 2012). Of course, a customer needs to be able to estimate very well how much and what groceries he/she needs for the whole week to order the right amount of certain products. Hansen (2008) refers to the third component of the Theory of Planned Behaviour, which is called “perceived behavioural control”. This defines a consumers’ personal belief whether he/she thinks that he/she is capable of carrying out that particular behaviour. In this case, the behaviour is estimating how many and what groceries the customer will need for the rest of the week. When this is the case, doing groceries once a week has a different effect on consumers’ preferred delivery time, compared to when a customer orders groceries every day. This preferred delivery time is expected to be much lower, because he/she plans in advance and does not need the groceries as soon as possible. This results in the following hypothesis:
H2b: The negative relationship between delivery time and willingness to pay for the delivery of groceries is less negative for consumers who schedule their groceries into one major trip per week compared to consumers who do not plan their grocery trips.
2.4.3 Delivery timeframe
Most people do not need their groceries to be delivered within a few hours, but rather make an appointment so that they know when they have to home (Hansen, 2008). People can indicate the timeframe that suits them the best, which is especially beneficial for customers who have a busy life. The big disadvantage is that people have to be at home at that point of time during the day, which means that they cannot be somewhere else (Colla, & Lapoule, 2012). Therefore, a small as possible timeframe is most convenient for the customers.
groceries to be delivered is not only seen as an economic, but also a psychological cost (Berry et al., 2002). According to Kumar et al. (1997) waiting leads to stress and anxiety, which results in lower customers’ satisfaction. They also examine the traditional queuing theory, which is applicable here as well. For instance, if the queue is too long, and with this the perceived waiting time, people will leave the queue. Therefore, if the timeframe for the groceries to be delivered is too large, people will less likely choose for the home delivery. Thus, consumers at least have to have the option to choose for a shorter delivery timeframe. Essentially, because of the perceived waiting time, the following hypothesis has been formulated:
H3: Willingness to pay for the delivery of groceries is higher for a smaller delivery timeframe, compared to a wider delivery timeframe.
2.4.4 Distance to the nearest offline supermarket
According to Gorczynski and Kooijman (2015), the accessibility of supermarkets in the Netherlands is very good. When comparing Netherlands to the UK, which is the European leader in the field of e-‐ commerce revenues (Syndy, 2014), the distance to a supermarket in the Netherlands is approximately .9 km, while the average distance in the UK is 2.3 km. This means that the average distance in the UK is more than twice as long compared to the Netherlands. Also, per million Dutch citizens there are 220 supermarkets available. In the UK, this number is 90 per million citizens. This shows that the relative timesaving in the Netherlands is much smaller compared to large countries, that do not have this many supermarkets per given number of inhabitants. Chu, Chintagunta and Cebollada (2008) complement this by suggesting that the higher the distance from a household to a store, the lower the price sensitivity of this household. The probable reasoning behind this is that households that visit the store more often also have been exposed to price and promotion activities of that certain supermarket more often, which gives them a better picture of the situation. Pate and Loomis (1997) also found that the larger someone’s distance from the store, the less likely this customer will buy at that particular store. In this case, the online channel offers a great opportunity to shop for groceries. The distance to the online store is after all 0 km. This is exactly what Melis et al. (2015) suggest in their study about multi-‐channel retailing. This results in the following hypothesis:
H4: The larger the distance to the nearest supermarket, the higher the willingness to pay for the delivery of groceries.
offline store. Because these people still have the opportunity to go to the nearest supermarket, they will most likely not order their groceries online and sit and wait for them for a few hours. Therefore, the following hypothesis has been formulated:
H5: The closer the distance between the consumer and the nearest store, the stronger the negative relationship between the delivery timeframe and consumers’ willingness to pay for the delivery of groceries.
2.5 Conceptual model
With help of the previous literature, abovementioned hypotheses have been drawn. Figure 1 below summarizes this theory and the hypotheses in the conceptual model.
FIGURE 1
Conceptual model of effects on WTP for delivery of groceries
3. RESEARCH DESIGN
This chapter explains the chosen research methods and the collection of data. It will end with an analysis plan, which is used for analysing the data.
3.1 Research methods
To be able to answer various sub-‐questions and the research question, it has been decided to conduct two studies: in-‐depth interviews and conjoint analysis. Before starting with the conjoint analysis, in-‐depth interviews are conducted. Processing and analysing of data is done with help of SPSS, Excel and Latent Gold.
3.1.1 In-‐depth interviews
The in-‐depth interviews have been conducted to complement the secondary data about customers’ hurdles and drivers when having their groceries delivered. This method is chosen to discover underlying motivations, beliefs, attitudes and feelings towards delivery of groceries (Malhotra, 2010). The big advantage of depth interviews is that there is no social pressure. Depth interviews focus on one respondent, so they can discover more and give a greater understanding of the topic from the respondents’ point of view. It was important to interview diverse people, to obtain greater likelihood of diverse information. To represent different segments, five people have been selected according to their age, gender, marital status and household size. The five interviewees are:
-‐ A student;
-‐ A woman living with her boyfriend. They both work and don’t have children; -‐ A married woman. Both of the couple work and they have two little children; -‐ A married man. Both of the couple work and their children are already grown up; -‐ An elderly grandmother.
This data has been used to confirm and complement the already acquired information about different drivers and hurdles of online grocery shopping and the delivery service, which is covered in the beginning of chapter two.
3.1.2 Conjoint analysis
analysis also shows underlying motives for respondents’ actions, which give insights into predicting their behaviour (Eggers, & Sattler, 2011; Elrod, Louviere, & Davey, 1992). Measuring customers’ preference also has the advantage that it predicts which characteristics and attributes have the biggest influence on consumers’ choice and their willingness to pay. The performed analysis is choice-‐based conjoint analysis (CBC), which presented a number of attributes and asked respondents for their most preferred option. The CBC has shown to be effective, because people make a lot of choices every day (Eggers, & Sattler, 2011).
3.2 Data collection: conjoint analysis
The collection of the data was done via the Internet and face-‐to-‐face. The survey, which was created in Qualtrics, was distributed via e-‐mail and social media. Also, friends, relatives, old colleagues and internship supervisors were asked to spread the survey among their network to obtain a higher response rate.
3.2.1 Segmentation
Before the respondents started choosing their preferred alternatives, they have filled in some demographic characteristics like gender, age, household size and marital status. After that, they have been asked about their grocery shopping behaviour. The variables distance from the store and
amount of organized grocery shopping trips are translated into questions instead of using them as an attribute in the conjoint analysis. This results in choice sets with less choice options, which are easier to understand compared to choice options with more attributes.
3.2.2 Attribute levels
The attributes are found by means of structured processes, namely existing theories in chapter two and depth interviews that are being conducted prior to the conjoint analysis. Attribute levels of
People can choose to have their groceries delivered the same day, which might be appreciated by a lot of people that are not ordering now because of the longer delivery times. Because this “extra fast” delivery is possible, as suggested in the previous chapter, it has been included in the conjoint analysis. The other two levels, which respondents can choose from, are that their groceries are delivered the next day or the day after tomorrow. Attribute levels of delivery timeframe are also based on the actual time window. The website of Albert Heijn (ah.nl, October 2015) has been used here. The timeframes of Albert Heijn range from 2 until 6 hours, which is why this is the chosen range for the attribute levels in the analysis. The attributes and corresponding levels are summarized in table 1 below.
TABLE 1
Attributes and attribute levels used in conjoint analysis
Attributes Attribute levels
Delivery costs € 2.50 € 5.00 € 7.50 Assortment size Small Medium Large Delivery time Same day Next day Two days Delivery timeframe 2 hours 4 hours 6 hours
As table 1 shows, there are three attribute levels for each chosen attribute. Reasoning behind this is to avoid the number-‐of-‐levels effect, which suggests that the attribute levels are unequal for the various attributes (Eggers, & Sattler, 2011; Steenkamp, & Wittink, 1994). This would bias the results, as respondents would perceive the relative importance of an attribute, with more levels, higher compared to an attribute with fewer levels. According to Eggers and Sattler (2009 and 2011), the number of levels should also not be more than seven, which is not a problem here. Three levels (compared to more attribute levels) means that less data is required to have reliable estimates.
3.2.3 Choice design
The amount of possible choice sets is 81 (4 attributes, with each 3 levels). The optimal solution would be to show the respondent all these possible attribute level combinations, which would result in a full factorial experimental design. As this is not possible and would result in fatigue effects, which lead to decreasing attention (Eggers, & Sattler, 2011), the fractional factorial has been chosen (Fontana, 2014). The number of choice sets should be between 12 and 15, but when using more than 12 choice sets, the respondent needs to be motivated. For this reason, the number of shown choice sets is 12 (Eggers, & Sattler, 2011).
overlapping levels or dominating alternatives). This was kept in mind, when creating the choice sets. Because each level is shown an equal number of times and every combination of attribute levels appears an equal number of times, the design is balanced and orthogonal.
The no-‐choice option is the fourth alternative of the choice set, which has a utility of zero (Liu & Tang, 2015). Even though some information could have been lost, because respondents might have chosen the ‘none’ option to avoid difficult decisions, this option is important when measuring the willingness to pay. It gave respondents the alternative to reject all three alternatives when their prices are too high (Eggers, & Sattler, 2011). The ‘none’ option is displayed as ‘I would rather go to the supermarket myself’. Figure 2 below is an example of a choice set that is shown. The whole survey in Dutch can be found in appendix I.
FIGURE 2
Example choice set conjoint analysis
Option 1 Option 2 Option 3 Option 4
Delivery costs: 5.00 Delivery costs: €2.50 Delivery costs: €7.50
I would rather go to the supermarket myself. Assortment size: Large Assortment size: Medium Assortment size: Small Delivery day:
Day after tomorrow
3.3 Plan of analysis
The section of the research design will address the procedure to obtain the results. To begin with, choices are based on overall utilities of alternatives. The utility of respondent n for alternative i is:
𝑈!" = 𝑉!"+ 𝜀!" With the two components:
𝑉!" = systematic utility component, rational utility 𝜀!" = stochastic utility component, error term
The general assumption of the utility function is that goods and services are combinations of attributes. Consumers attach preferences to each attribute. Systematic utility of respondent n for alternative i is sum of the part-‐worth utilities:
𝑉!"= !
!!! 𝛽!"𝑋!" With:
k = (1, …, K) number of attributes
𝑋!" = dummy indicating the specific attribute level of product i
𝛽!" = part-‐worth utility (preferences, to be estimated) of consumer n for attribute k
As the respondents’ chosen option can be any alternative from a choice set, the dependent variable can exhibit multiple states. For this reason a multinomial logit model has been used. Complementing this function with the attributes from table 1 on page 18, it gives the following formula:
𝑉!"= 𝛽!!𝐷𝑒𝑙𝐶𝑜𝑠𝑡𝑠!+ 𝛽!!𝐷𝑒𝑙𝐶𝑜𝑠𝑡𝑠!+ 𝛽!!𝐷𝑒𝑙𝐶𝑜𝑠𝑡𝑠!+ 𝛽!!𝐴𝑠𝑠𝑆𝑖𝑧𝑒!"#$$+ 𝛽!!𝐴𝑠𝑠𝑆𝑖𝑧𝑒!"# + 𝛽!!𝐴𝑠𝑠𝑆𝑖𝑧𝑒!"#$%+ 𝛽!!𝐷𝑒𝑙𝑇𝑖𝑚𝑒!"#$%"&+ 𝛽!!𝐷𝑒𝑙𝑇𝑖𝑚𝑒!"#$%&'
+ 𝛽!!𝐷𝑒𝑙𝑇𝑖𝑚𝑒!"#$%&'+ 𝛽!"!𝑇𝑖𝑚𝑒𝐹𝑟𝑎𝑚𝑒!"!"#$+ 𝛽!!!𝑇𝑖𝑚𝑒𝐹𝑟𝑎𝑚𝑒!"!"#$ + 𝛽!"!𝑇𝑖𝑚𝑒𝐹𝑟𝑎𝑚𝑒!"!"#$
With:
𝑉!" = Utility for the chosen option
𝛽!!𝐷𝑒𝑙𝑇𝑖𝑚𝑒!"#$%&'= Dummy next day delivery for alternative i 𝛽!!𝐷𝑒𝑙𝑇𝑖𝑚𝑒!"#$%&'= Dummy delivery in two days for alternative i
𝛽!"!𝑇𝑖𝑚𝑒𝐹𝑟𝑎𝑚𝑒!"!"#$= Dummy delivery timeframe within 2 hours for alternative i 𝛽!!!𝑇𝑖𝑚𝑒𝐹𝑟𝑎𝑚𝑒!"!"#$= Dummy delivery timeframe within 4 hours for alternative i 𝛽!"!𝑇𝑖𝑚𝑒𝐹𝑟𝑎𝑚𝑒!"!"#$= Dummy delivery timeframe within 6 hours for alternative i
Because the first model that is estimated is totally part-‐worth, this model form is used when composing the formula above. Based on the utility function above, the utility function is also determined, which shows a prediction of the probabilities that alternative i is chosen out of j choice sets. 𝑝𝑟𝑜𝑏 𝑗 𝐶) = 𝐸𝑥𝑝 (𝑉!") 𝐸𝑥𝑝 (𝑉!") ! !!!
3.3.1 Assessment of model fit
The model fit will be determined with help of the Likelihood ratio test, adjusted R2 and the hit rate. Here, the model will be compared to the aggregate model and the NULL model to see whether the estimation of the chosen model is better.
3.3.2 Latent class analysis
4. RESULTS
This chapter discusses the results of the analyses and their main findings. First the in-‐depth interviews will be addressed, followed by sample characteristics and conjoint analysis. It will end with a segmentation and hypotheses overview.
4.1 In-‐depth interviews
As mentioned in chapter three, to complement the hurdles and drivers that are obtained with help of secondary data, in-‐ depth interviews are conducted. The interviewees were:
-‐ A student;
-‐ A woman living with her boyfriend. They both work and don’t have children; -‐ A married woman. Both of the couple work and they have two little children; -‐ A married man. Both of the couple work and their children are already grown up; -‐ An elderly grandmother.
The oldest interviewee is the elderly woman, who regularly orders her groceries via the website of Albert Heijn. She orders her groceries approximately every two weeks and now and then has to visit the traditional supermarket to pick up the groceries she had forgotten to order. The main reason for her to order groceries online is that she doesn’t have to lift the heavy bags. The main drawback she mentions is that it is impossible to see, feel and touch the products, which results in the fact that some products, mostly fruit and vegetables, are often disappointing.
The married man (with grown up children) and married woman (with little children) both suggest that they perceive doing groceries as a relaxing activity. Because the interviewees and their partners work fulltime, it is nice to go shopping together, run into neighbours and have a chat. In addition, they are not very enthusiastic about the delivery costs.
woman likes to be spontaneous and inspired in the supermarket. This statement is confirmed by the fifth interviewee (the student) as well. Lastly, even though online shopping has been seen as time saving, it also takes time to order the groceries. The interviewee tells that, the last time she ordered online, it took a lot of time because she had been adding to and removing products from the list. Altogether, these two young women think it is a very good solution when you have a family, fulltime job, not much time and in need of a lot of groceries, online grocery shopping might be a good option for them in the future.
Shortly summarized, these interviews have helped to confirm the information stated in chapter two. These five people, who are in different stages of their lives, have not really given new information. Most of the attributes are mentioned by the interviewees, which is a good sign because it confirms the found literature. This information was kept in mind when drafting the survey questions.
4.2 Descriptive data: all respondents
In total, 481 people have started filling out the survey. In the end, 329 finished surveys were useful. 39.2% of the respondents is male and 60.8% is female. Most of the respondents (38.3%) are between 18 and 25 years old, and the highest proportion (69.9%) lives in the North of the Netherlands. The main explanation for this is that the biggest part of the network is student or has just started working and lives in Groningen or Friesland. In fact, almost 30% of the respondents is studying.
4.3 Descriptive data: ‘only no-‐choice respondents’ excluded
