The Future of Grocery Shopping

(1)

0

The Future of Grocery Shopping

An Incentive Aligned Choice Based Conjoint Analysis : Assessing the Willingness to Pay for Delivery Services of Online Grocers

Master Thesis

MSc. Marketing Intelligence & MSc. Marketing Management

(2)

1

The Future of Grocery Shopping

An Incentive Aligned Choice Based Conjoint Analysis: Assessing the Willingness to Pay for Delivery Services of Online Grocers

Master thesis

MSc. Marketing Intelligence & MSc. Marketing Management

University of Groningen

Faculty of Economics and Business

J. (Jan) Nies

Papengang 38

9711 PA Groningen

T: +31 6 44744 003

Jan_nies@hotmail.com

Student number: 2611015

Supervisors:

(3)

i

Management summary

More and more consumers nowadays order products and services online. It is the online retailers, such as Coolblue or Bol.com, that change market dynamics drastically. Old fashioned brick-and-mortar stores such as V&D and Macintosh, which do not keep up with their online competitors, are losing ground. Online grocery shopping, despite the growing popularity, is still lagging compared with the ‘regular’ online retail. Research suggests that the delivery fees asked by the online grocers are the main hurdle for not ordering groceries online.

This study (n=731) helps management to optimize their online grocery services by providing concrete advise about the willingness to pay (WTP) of consumers, for the different delivery services available now in the market. It shows which elements stimulate consumers to shop groceries online and which form hurdles for consumers and their impact on the WTP for online grocery delivery services.

The vast majority of the consumers consider price, delivery day and the retail brand to be the most important elements of online grocery shopping, in the order as mentioned. Although one must not neglect the timeframe, delivery option and assortment of the online service, consumers find these elements less important. The majority of the consumers are willing to pay a considerable amount of money for the services provided by online grocers, where the WTP ranges from 60% of the consumers from €9.86 to €10.56 for having the groceries delivered at home. A large share of all consumers, however, is more price sensitive and is not willing to pay for delivery of online groceries. For picking up groceries, 46% is willing to pay €10.41 whereas 54% is willing to pay a delivery fee ranging between €0.74 and €1.62.

(4)

ii

Preface

I would never believe, if 11 years ago after graduating at the VMBO and starting at the MBO International business studies, future me said that I would be writing my master thesis now. He would even have had to explain what a master thesis would be. I would like to thank everybody that helped me during this 11 year period. Especially I would like to thank Erjen for his support.

(5)

1 Introduction

Since the introduction of the first (modern) supermarket, this branch of activity has radically changed. In roughly hundred years, small grocery shops transformed in large (inter)national grocery chains. The internet has created great opportunities for both firms and consumers. Consumers now can buy everything they want, whenever they want. The internet has made the world more transparent and made the consumer by far the strongest party. Where retailers decided what you were buying in 1920, nowadays consumers decide what the retailer sells (Sloot, 2015a).

In the Netherlands, e-commerce grew with 9.8% in 2014 and accounted for 7.1% of the total retail sales (CBS, 2015a; Syndy, 2015). However, online grocery shopping is still in its infancy in the Netherlands. 2.7% of the total grocery spending can be allocated to online sales (GFK, 2015). This signifies that online grocery spending is relatively low. Although the UK is at the forefront with 4.4% of the total groceries shopped online, even there, online grocery shopping is still lagging compared to general online spending (Syndy, 2015).

Firms have tried to set up a profitable online grocery branch, but many have been declared bankrupt or decided to withdraw from the market. Some Dutch examples are Superdirect.com and truus.nl. The ‘last mile’, the delivery process of groceries, appears to be one of the highest hurdles when it comes to make an e-grocer profitable (Boyer and Hult, 2005; Punakivi, Yrjölä, and Holmström, 2001). The already low margins of this sector are under severe pressure when providing a delivery service at competitive prices. For example, truus.nl order fulfilments costs accounted for 40% of the total costs (Sloot, 2015b). In order to make at home delivery attractive for grocers, these costs need to be lowered drastically. A new concept in the Netherlands, Picnic, has a more promising business model (Sloot, 2015a). Picnic has no goods in stock, which saves storage costs and product loss. Furthermore, to streamline their logistics, they deliver their products according a specific delivery route (Distrifood, 2015a). Albert Heijn (AH) is by far the largest e-grocer in the Netherlands. With an online revenue of €240 MLN in 2014, it outperforms supermarkets like Spar (21 MLN), Plus (16 MLN), Deen (7 MLN) and other (pure) players such as Hello-Fresh and Picnic (Syndy, 2015). Jumbo recently announced that it will deliver products at home in the near future, besides the already existing pickup points (Distrifood, 2015b). With Jumbo entering as the second largest player in ‘offline’ groceries in the Netherlands, competition is likely to increase which will change market dynamics drastically (Syndy, 2015).

(8)

2 more remarkable that 58% of the households do not even consider shopping groceries online (CBS, 2015c).

What causes the slow adoption of online grocery shopping? Research suggests that delivery fees are one of the most important factors of rejecting online grocery shopping (Chintagunta, Chu, and Cebollada, 2012; Hand et al., 2009; Huang and Oppewal, 2006; Morganosky and Cude, 2000). In the Netherlands, the fees of home delivery of groceries can add up to €12.95 at AH, starting at €4.95. Other online grocers use flat rates ranging from 4.95 (Spar) to free delivery (e.g. Deen). Also, the delivery fees can be relatively high compared with the costs of the groceries itself. For example, in the extreme case of paying €12.95 for at home delivery for a shopping list of €70, almost 20% is accounted to delivery costs.

It is widely acknowledged that price directly influences demand and thus influences revenues (Bijmolt, Heerde, and Pieters, 2005). Furthermore, price is the instrument for marketers that is most easily adjusted, however, can have large implications. Price can influence consumers’ willingness to buy or the perceived value of a product, and can work as a quality and prestige indicator (Brucks, Zeithaml, and Naylor, 2000; Dodds, Monroe, and Grewal, 1991). Not only does price influence consumers, it can also seriously influence dynamics on macro level and it can disrupt market dynamics. For example, in 2003, Albert Heijn started a price war, which caused for a lot of turmoil, increases price sensitivity and a permanent change in consumer shopping behavior (Van Heerde, Gijsbrechts, and Pauwels, 2007; NRC, 2013). As the price of the delivery of the online groceries is one of the main obstacles to use the service, it is important to know for both theorists and practitioners what price they should charge for processing and delivering orders to stimulate acceptance of the e-grocer. There is scarce literature on drivers of adopting and not adopting the service of online grocers and does not provide theorists or practitioners a concrete answer of what consumers are willing to pay (hereafter WTP) for these services. Also, as this is a relatively new field of research, most of the research is more practical in nature.

The focus of this study is presented in the following problem statement:

‘What is the customers’ willingness to pay for the delivery of online groceries?’

(9)

3 The following sub questions are formulated:

1. What does the adoption process of innovations look like?

2. What factors drive the adoption/rejection of online grocery shopping?

3. What factors influence the consumers’ willingness to pay delivery of online groceries?

(10)

4

2 Theoretical framework

2.1 Adoption of innovations

In an ever faster changing world, firms need to innovate to keep their customer satisfied and to keep up with competition. Constantly meeting new (latent and manifest) customer needs and providing benefits and features that were not available before, are key drivers of product success (Henard and Szymanski, 2001). The bulk of the literature suggests that a firms’ innovation effort is positively related to the performance of a firm by having an impact on e.g. sales, profitability or stock price returns (see e.g. Ataman, Van Heerde, and Mela, 2010; Brynjolfsson and Smith, 2000; Frambach and Schillewaert, 2002; Joskow et al., 1993; Srinivasan et al., 2009; Zhang, Ko, and Lee, 2013). This suggests that firms need to innovate in order to keep its right to stay in the game.

Rogers defined adoption as: ‘to make full use of an innovation as the best course of action

available’ (Rogers, 1995, p. 21). It seldom occurs, however, that innovations are adopted instantly.

An important factor whether someone adopts an innovation, is the ‘complexity’ of the innovation. A meta-analysis of Garcia and Calantone (2001), provides a framework of innovation typology which consists of three types of innovations: incremental innovations, really new product innovations and

radical innovations with each having an increasingly complex adoption process over the previous

one, for example in terms of learning costs, and increases adoption time (Mugge and Dahl, 2013; Slater, Mohr, and Sengupta, 2014).

(11)

5 Relative advantage “The degree to which an innovation is perceived as consistent with the existing

values, past experiences, and needs of potential adopters.”

Compatibility “The degree to which an innovation is perceived as consistent with the existing values, past experiences, and needs of potential adopters.”

Complexity The degree to which an innovation is perceived as relatively difficult to understand and use.”

Trialability “Is the degree to which an innovation may be experimented with on a limited

basis.”

Observability “The degree to which the results of an innovation are visible to others.”

Table 1, Innovation characteristics

2.2 Consumer adoption of online grocery shopping

In this chapter the framework of innovation characteristics as suggested by Rogers (1995) are used to identify relevant characteristics for the adoption of online grocery shopping.

2.2.1 Relative advantage

2.2.1.1 Convenience

Convenience has been one of the most important underlying motivations why consumers adopt online shopping, for both shopping in general and for shopping for groceries (Chintagunta, Chu and Cebollada, 2012; Chu et al., 2010; Colla and Lapoule, 2012; Jiang, Yang and Jun, 2013; Rohm and Swaminathan, 2004; Verhoef and Langerak, 2001). When shopping online, one can stay at home, does not have to cycle through the rain. Chintagunta, Chu, and Cebollada (2012) suggested that the physical costs of offline shopping positively influences online shopping. This especially holds for large basket shoppers and/or heavy bulk items as it can save a lot of energy, of which effect becomes stronger with an increase in distance to the shop. Other examples of convenience that online shopping provides are: parents that do not have to drag their children through the supermarket anymore or for the (temporary) physical disabled consumers who find it hard to get out of their homes (Morganosky and Cude, 2000; Ramus and Nielsen, 2005).

2.2.1.2 Time saving

(12)

6 saves time as searching for products on (most) websites is easier, once a customer has experience on a grocers’ website. This is especially true during weekdays, when most people have tight work schedules and do not want to spend their free time on grocery shopping.

2.2.1.3 Product range

Ramus and Nielsen (2005) found in their study that participants perceive a large range of products in their online assortment is a serious advantage of shopping online. Also in the offline retail setting this is an important factor for store choice. According to Ailawadi and Keller (2004), a larger breadth and depth of an assortment can increase consumer recall and the intention to visit a shop. Also, the one-stop-shop concept provides consumers a lot of convenience and consumers can buy products at the focal retailer instead of at other retailers and leads to enhanced store evaluations (Oppewal, Koelemeijer, 2005). A large assortment thus seems to attract customers. However, research is divided on this topic. Authors have discussed ‘choice overload’ extensively (see for an extensive overview Scheibehenne, Greifeneder, and Todd (2010)). For example, Iyengar and Lepper (2000) show that although a large assortment is initially attractive to customers, but when participants were exposed to a large assortment vs. smaller assortment, they were less vs. more likely to buy products. They further show that participants exposed to a large assortment are less satisfied with their product choice afterwards. However, meta-analysis of Scheibehenne, Greifeneder, and Todd (2010) confirms the notion of ‘more is better’ with respect to assortment. Only exposed to certain preconditions consumers perceive a large assortment as negative, such as when products are not categorized, if products have non-comparable/unique features or when consumers are under time pressure during their shopping trip.

2.2.2 Compatibility

2.2.2.1 Delivery price

Many researchers agree on the finding that delivery price is the most important factor of rejecting online grocery shopping (e.g. Chintagunta, Chu, and Cebollada, 2012; Colla and Lapoule, 2012). People find that the delivery costs are a substantial amount of the total grocery shopping costs when shopping online, especially for small orders (Ramus and Nielsen, 2005).

2.2.2.2 Ability to ‘touch and feel’ products

(13)

7 Nielsen, 2005). For example, some consumer might have a strong preference for nearly green bananas, and do not want to receive over-ripe yellow/brown ones. The inability to judge these quality aspects make that consumers prefer to shop in the offline store. This is related to the perceived risk and reducing uncertainty of online shopping as consumers are not in control in these situations (Herhausen et al., 2015).

2.2.2.3 Fun and enjoyment

People who shop in regular supermarkets often state that the interaction with other customers or shopping with friends and/or family is important to them and is also one of the main reasons why they do not shop online (Chu et al., 2010; Colla and Lapoule, 2012; Ramus and Nielsen, 2005; Verhoef and Langerak, 2001). In line with these findings, Arnold and Reynolds (2003) found in their study in hedonic shopping motivations, that the social interaction and enjoyment with others is one of the reasons why people shop. (Some) people have the desire to interact outside their own home with relatives of other customer who have similar interests and affiliate with reference groups. Furthermore, it is not only friends/family or other customers that satisfy the social needs when shopping, (loyal) customer do have frequently good relations with store owners and employees. Interaction with firms’ employees or store owners is part of the shopping experience (Reynolds and Beatty, 1999). Other interesting findings in this study are that this shopping segment are the heaviest purchasers, generate the highest word of mouth and are most loyal. Several studies (e.g. Cha, 2009; Wang, Yu, and Wei (2012)) suggest that social media or other social add-ins can stimulate and simulate social interaction. However, this can never fully replace the real life social interactions, just for the fact that these interactions include physical interactions and intimacy, such as a handshake or kiss.

2.2.2.4 Pickup- or delivery time window is defined by the service provider

(14)

8

2.2.3 Complexity

Prior research focused on the ease of use of web shops and learning costs of shopping online for groceries. Although some of these studies are less relevant as the majority of consumers now are digitally trained and online shopping is common, ease of use or the quality of the webshop still is relevant (Boyer, 2007; Elliot and Fowell, 2000). Perceived ease of use of a website even appears one of the most important drivers of the evaluations of websites (Dickinger and Stangl, 2013). When people are dissatisfied with the functioning of a website, or can’t even find the information they are looking for, rejection is likely to follow.

2.2.4 Trialability

Consumers can easily try online grocery shopping without being obliged to be subscribed. Exceptions are automatic-subscriptions such as Hello-fresh, however, these services also have a trail period.

2.2.5 Observability

As the majority of the diffusion process is triggered by social aspects (Bass, 2004), this is an important component. With online services, there is no need for, however also no possibility anymore for social interaction. Also, according to Rogers (1995), innovations with a lower observability also have slower adoption rates.

In Table 2, an overview is presented of the innovation characteristics relevant for this study. Innovation characteristic Aspects

Relative advantage Convenience, time savings and assortment

Compatibility Delivery costs, ability to touch the products, fun and enjoyment and time windows

Complexity Ease of use

Trialability No subscriptions Observability Social interaction Table 2, Innovation Characteristics

Although complexity, ability to try and observability are important factors in the adoption process, out of scope for this study as the focus is on the WTP of these services. Also some of these

(15)

9

3 Factors of influence on the WTP of delivery of online groceries

As mentioned in chapter 2.2.2, the compatibility of an innovation can influence the adoption speed of innovations. Unlike shopping in stores, online grocery shopping are associated with an extra delivery fee. The delivery costs of online delivery are perceived as the most important negative compatibility of online grocery shopping. Therefore it is of key importance to know what consumers are willing to pay for delivery of their groceries and what factors influences the WTP. In this study, WTP is defined as the maximum amount of money a customer is willing to pay for a product or service (Koschate-Fischer, Stefan, and Hoyer, 2012).

In recent years, the Omni-channel retailing literature has shown what dramatic transformations the retail sector has undergone (see for an overview Verhoef et al. (2015). One of these changes is the way how firms and customers now interact due to technological advances (Van Bruggen et al., 2010). Two of these recent developments are showrooming and web rooming, which have a profound effect on WTP (Rapp et al., 2015). These technological advances increase market transparency, which is beneficial for consumers, as they are fully informed about pricing and can pick the cheapest product among several stores. For firms however, this is less beneficial as price has come under severe pressure. There is empirical evidence that ‘the internet’ has lowered prices. To illustrate, Zettelmeyer, Morton, and Silva-Risso (2006) found that internet transparency, lowered car prices on average with 1.5% or in terms of a car dealers gross vehicle profit, with 22%. Brown and Goolsbee (2002) found that insurance prices were reduced with 8-15% in a two year period due to internet competition. In addition, other research showed that increased transparency, for example with price agents (e.g. kieskeurig.nl), increases price sensitivity (Degeratu, Rangaswamy, and Wu, 2000; Diehl, Kornish, and Lynch, Jr., 2003).

(16)

10 In the next section, the factors that influence the WTP of online delivery services of grocers. First the internal factors that influence the WTP of delivery services are discussed. These are factors that can be influenced by the firm.

3.1 Retail brands

Although some online retailers provide free delivery services, the majority of online grocery shoppers buy at AH. This can be the result of early movement advantages of AH and the fact that AH’s delivery services are available everywhere in the Netherlands. Another explanation might be that the brand of AH might play a role in the formation of WTP of consumers. Many authors suggest high equity brands can ask their customers a price premium and thus can increase the WTP for a brand (Ailawadi, Lehmann, and Neslin, 2003; Buil, Martínez, and de Chernatony, 2013; Netemeyer et al., 2004). (Customer based) brand equity (BE) is defined as ‘the differential effect that brand knowledge

has on consumer response to the marketing of that brand’. (Keller, 2008, p.69) A brand has a positive

(negative) BE when consumers react more (less) favorable to the brand than when the product is not branded. This definition has three important ingredients. The first is ‘the differential effect’ of the response of the consumer, which holds that when there is no effect, the product can be seen as generic product. The second is ‘brand knowledge’, which implies that the consumer has learned, experienced and has a certain knowledge of the brand. The third is the ‘customers’ response to the marketing’ which are the preferences and behavior towards the focal brand, such as recall or brand evaluations. These findings suggest that brands with a high equity can ask a (higher) price premium on the delivery costs than low brand equity brands can do. Therefore the following hypothesis is proposed:

H1: The WTP for delivery of online groceries is higher for high equity retail brands than for low equity retail brands

3.2 Time window of delivery

(17)

11 a period of two or four hours. One possible underlying reason why consumers perceive this as negative is that the perceived waiting time often is longer than the actual waiting time (Antonides, Verhoef, and van Aalst, 2002; Baker and Cameron, 1996; Kumar, Kalwani, and Dada, 1997). When people have to wait at home, people will feel less happy and useless. As a consequence, longer waiting times will lead to lower satisfaction (Antonides, Verhoef, and van Aalst, 2002; Kumar, Kalwani, and Dada, 1997). Therefore it is expected that:

H2: The smaller the time window of the delivery of groceries, the higher the willingness to pay for the delivery costs.

3.3 Shipping time

The most important relative advantage of online grocery shopping is that it provides the consumer higher level of convenience compared with ‘offline’ grocery shopping as it saves time and energy (Choi and Bell, 2011). Savings in time thus seems to be highly important for consumers when shopping groceries online. In general, shipping times of online orders become increasingly shorter. In the Netherlands for example, Coolblue offers consumer same day delivery, if they order before 15:00. In a grocery setting, at Picnic, customers can order until 22:00 and receive their products the next day. Shorter shipping time can provide serious benefits in a grocery setting. For example, most consumers do not buy groceries in bulk, and it is more convenient for consumers to decide at the point of purchase what to eat that day rather than having to plan 3 days in advance. Therefore it is expected that:

H3: The shorter the shipping time, the higher the willingness to pay for delivery

3.4 Assortment

(18)

12 en Cultureel Planbureau, 2013). It is reasonable to assume that as people want to spend the sparse free time left on activities they like and become ever more polychromic. Grocery shopping thus comes with high opportunity costs. Therefore consumers can minimize shopping costs with one-stop-shop websites and increase customer satisfaction (Campo and Breugelmans, 2015; Chintagunta, Chu, and Cebollada, 2012; Felker et al., 2013). Therefore it is expected that:

H4: The larger the assortment, the higher the willingness to pay for the delivery of groceries.

3.5 Delivery options

As mentioned before, there are several delivery options to choose for a customer. These options, at home delivery, pick-up points, automatic subscription or a grocery boy, all have different characteristics and provides different convenience levels that serve as a relative advantage compared with ‘offline’ grocery shopping. Pick-up points still ask some time and effort from people as they still need to travel to a pick-up point. Often this is also the least expensive option when grocers also provide delivery services (see e.g. AlbertHeijn.nl). Home delivery of products provides more convenience than pick-up points as there is no need to travel, and thus the relative advantage of home delivery can be perceived as higher. Although the time frame of delivery can be perceived as negative, it still is expected that:

H5: The WTP for home delivery is higher than for pick-up points

Automatic subscription is not yet available in the Netherlands. Amazon Prime, for example, is

available for €299 a year in the US. The high initial investment might scare consumers away. However there is evidence that consumers prefer flat rate pricing over pay-per-use. Most shoppers overestimate their product/service use and therefore prefer a flat rate rather than pay-per-use (Lambrecht and Skiera, 2006). However, for methodological (see chapter 4) reasons and to keep this study concise, the WTP of automatic subscriptions and ‘grocery boy’s’ are not of interest in this study.

(19)

13

3.6 Brand loyalty

A closely related concept to brand equity as discussed in chapter 3.1, is brand loyalty. Consumers who are loyal to a brand are less price sensitive than non-loyal consumers (Aaker, 1991; Allender and Richards, 2012; Krishnamurthi and Raj, 1991). Loyal customers tend to stick with their preferred brand when exposed to alternatives, and a substantial amount of risk should be involved when a loyal customer switches brands. Loyalty thus can prevent customers to switch from their preferred brand even when the price level of the substitute is lower. Therefore it is expected:

H6a: Loyal consumers are willing to pay more for delivery costs than non-loyal consumers

Brand loyalty can be distinguished between behavioral (habitual) and attitudinal loyalty (Kumar, Pozza, and Ganesh, 2013; Liu-Thompkins and Tam, 2013). Both kinds of loyalty result in similar purchase behavior, but are different constructs. Behavioral loyalty refers to the phenomenon that consumers buy products as if there were no competitive offers or being loyal to the location of the supermarket rather than the retail brand. Attitudinal loyalty, however, refers to the phenomenon that a consumer buys a product as he/she genuine believes that this product has superior attributes compared with alternatives and has positive feelings to that particular brand (Liu-Thompkins and Tam, 2013). As consumers with a high attitudinal loyalty are more intrinsically motivated to buy at one firm than another, it is necessary to distinguish between these effects. It therefore is expected that:

H6b: Consumers that have a higher attitudinal loyalty towards a brand are more WTP than consumers who have a higher behavioral loyalty

3.7 Experience with online shopping

According to Melis et al. (2015), when people start shopping online for groceries, consumers prefer shopping at their preferred offline supermarket. When people gain more online shopping experience, this effect weakens and offline preferences disappears and become more price sensitive (Diehl, Kornish, and Lynch, Jr., 2003). Therefore it is expected that:

(20)

14

Figure 1, Conceptual model

(21)

15

5 Methodology

5.1 Qualitative study

A qualitative study was conducted in order to validate the findings of chapter 2 and test whether the attributes in proposed in chapter 3 were indeed the most important regarding online grocery shopping. Research suggests to conduct 6-30 interviews, depending on the nature of the study (Cresswell and Miller, 2000; Denzin and Lincoln, 1994; Kuzel, 1992). According to Kuzel (1992, p41) 12-20 data sources are needed “when looking for disconfirming evidence or trying to achieve maximum variation”. Lastly, Guest, Bunce and Johnson (2006) suggest that saturation occurs within the first twelve interviews. The 19 data sources in this study should provide the information needed. In total, 6 in-depth interviews and five group discussions with 2-5 persons were conducted. In total, 19 people were interviewed. Qualitative studies can extend previous findings with insights not present in current literature or show insights of relations between current findings (Malhotra, 2010). Group discussions can increase synergism, which provides a wider range of information than when using solely using private interviews, and snowballing, triggering chain reactions of other participants. Private interviews can reveal deeper insights, however, take more time to conduct.

(22)

16

5.1.1 Procedure

During the interviews, open ended questions were asked to find out the drivers and hurdles of online grocery shopping. If some of the drivers or hurdles were not directly mentioned, these were suggested and interviewees were asked whether these are relevant to them. The interviews were semi-structured, and open-ended questions were asked provide more in-depth insights. Thereafter, respondents were asked whether the attributes in chapter 3 are important attributes regarding the WTP of online grocery shopping. During the group discussions, also a semi-structured method was. First the participants were asked what the drivers and hurdles of online grocery shopping were for them, which followed with a discussion between the participants. When a discussion stopped or participants had nothing more to say about a specific topic, a new statement was introduced. Lastly, the most important attributes of online grocery shopping in relation to delivery charges were asked.

5.1.2 Results

The results were in line with the results provided in chapter 2 and 3. The results of the interviews and group discussions are nearly identical and are therefore discussed simultaneously. Generally, more educated people are more experienced with online grocery shopping than less educated people. Some even indicated not having ever thought of shopping online groceries. Also, higher educated participants suggested that the pick-up point was more attractive. Moreover, picking up groceries after work or let them deliver right after work is perceived as most beneficial. Every participant agreed more or less that online grocery shopping can provide the customer convenience. Most participants mentioned online grocery shopping as being convenient to them as they do not have to carry heavy grocery bags and not having to travel cycle through the bad weather anymore. Furthermore, respondents argued that when being physical disabled, online delivery can be very convenient (or even a necessity). Time savings was also mentioned by nearly every respondent. Only two participants mentioned that assortment was more attractive online. Some participants stated that they shop at different stores for specific products. For example, two respondents (who never bought groceries online) found that online shopping in general is less preferred and especially preferred shopping personal care such as shampoo and toothpaste at hard discount stores (OP=OP) as these are much cheaper. They stated that it is not desirable to shop at two or three different shop online as every e-tailer has different time windows for delivering their products (if they even own a webshop).

(23)

17 perceived as too high, especially for the small household sizes. Not being in control of the quality of the products was also mentioned by nearly everyone. Especially the perceived quality and the hassle and uncertainty of returning products was mentioned often. Also the absence of scents of freshly baked products was a big disadvantage. Many participants really enjoyed these experiences.

Nearly every respondent mentioned that they liked the social aspects of shopping groceries. One participant mentioned that he worked at home a few days per week and has little social contacts during these days (except for mail and telephone) and therefore grocery shopping fulfilled partly his social needs.

Longer waiting times in terms of delivery day and time frames were also considered negative. However, most participants mentioned that when they didn’t want to wait at home for their products, they would use a pick-up point and which would solve this problem. A last interesting finding is that many participants use store cues for determining what products they buy, for dinner or browse through the isles for items they otherwise would forget.

5.2 Choice Based Conjoint Analysis

Choice based conjoint analysis, or more narrowly, incentive aligned choice based conjoint analysis,

was used to measure respondents’ utilities for online grocery shopping and WTP for online delivery of their groceries. With Choice Based Conjoint analysis (hereafter CBC), the relative importance of different attributes of the delivery of online grocery shopping, can be estimated. Each respondent is presented several stimuli that consist of different combinations of attribute levels and has to evaluate these stimuli with respect to their attractiveness. The most important assumption of ICBC is that each product is perceived as attribute bundles. The goal of CBC is to quantify the utilities for product attributes (attribute bundles) and attributes levels (Malhotra, 2010).

(24)

18 rating or ranking a product or service, which makes ICBC the most preferred option (Toubia, Hauser and Simester, 2004).

Respondents were instructed to imagine a situation where they had to buy online groceries. Respondents are presented with ten choice sets with three attribute bundles. According to Eggers and Sattler (2009), fewer choice sets reduce utility evolution and boredom, which can increase the validity. Participants are asked at every choice set which option they would choose this option if this would be presented in real life. The incentive in this study a coupon of €50 at the supermarket which they prefer most. Also a no-choice option was included to make the choice option more realistic.

5.2.1 Measures

5.2.1.1 Attributes and attribute levels

The price attribute included three levels, €2.50, €5, €7.50. Although some supermarkets deliver for free, logically, free delivery would be the most preferred option by far; this would heavily bias the outcomes. Although a free option would be the most close to the real life situation, the grocers who provide free deliveries are not located everywhere in the Netherlands, which justified these measures.

Two delivery options were studied in this study, namely at home delivery and pick-up points. As mentioned in chapter 3.5, there are more delivery options available; however, as these options are not well known by the Dutch population, only delivery at home and pick-up points are included. Although literature warns for the number of levels effect, where respondents attach a relatively high value to attributes that have more attribute levels and leads to biased results, other delivery options were left out of this study (Steenkamp and Wittink, 1994; De Wilde, Cooke, and Janiszewski, 2008). Also, if other delivery options were included, this would conflict with other attributes. If a ‘grocery boy’ is included, delivery day and time windows are irrelevant as grocery boys deliver within an hour. Automatic subscriptions are delivered at specific moments, for example once every three weeks on Thursday and the attribute delivery day and time windows would be irrelevant as well. Brand equity was manipulated with thee (national) grocery chains. A distinction was made between high (Jumbo), medium (Albert Heijn) and low BE (Spar) (Distrifood, 2015c).

The attribute levels of the time frame included three levels, 1, 3 and 5 hour intervals. In the Dutch online grocery market of you can choose between 1 hour-5 hour intervals (see e.g. ah.nl). For pick-up points, this differs per local supermarket and chain. Some Jumbo’s provide pick-up points which can be visited during regular opening hours but most have specific hours (mostly 16:00-19:00) to pick-up the groceries. To keep the choices as realistic and feasible as possible and the 1, 3 and 5 hour interval is chosen.

(25)

19 in the Netherlands there is no grocer who provides the same-day delivery service, in other sectors such as clothing and electronics, this is quite common and this might be the standard in a few years as well in the grocery sector.

Three attribute levels for assortment were used, a ‘normal’ supermarket assortment (which includes fresh products, packaged food, diapers, toilet paper etc.), an extended assortment of drugstores as second level, and as third level, both ‘normal’ ‘drugs’ extended with ‘liquor’ were used.

In order to measure the WTP of respondents for the delivery of online groceries, a no-choice option is added (see e.g. Gensler et al. (2012)). The no-choice option also increases the representativeness of the study as consumers in real life also can choose to reject the option presented and look for other options, although some information could be lost as some respondents chose the no-choice option as they find choosing an alternative to difficult. The no-choice option was displayed as ‘With these options provided, I would search for other alternatives or would go to a physical supermarket’. A summary of all attributes and attribute levels and an example of a choice set can be seen in Table 3 and Table 4 (for the survey see appendix a).

Attribute Attribute levels

Retail brand Jumbo, Albert Heijn and Spar Delivery price €2,50, €5 and €7,50

‘Delivery option’ Home delivery and Pick-up point Time window 1 hour, 3 hours and 5 hours

Shipping time Same day, next day and two day delivery

Assortment Normal, Normal + drugs, Normal + drugs + liquor

Table 3, Attributes and attribute levels

Table 4, Example choice-set

Retail brand

Delivery option

Pick-up

Delivery fee € 7,50 € 5,00 € 2,50

Delivery / pick-up day One day after

ordering

Two days after

ordering Same day

Time frame 5 hours 1 hour 3 hours

Assortment

(26)

20

5.2.1.2 Loyalty measures

Behavioral loyalty was measured by two statements on a seven point Likert scale: "I will buy at (retailer brand) the next time I buy groceries" and "I intend to keep purchasing at (retail brand)” (see e.g. Algesheimer, Dholakia, and Herrmann (2005) and Chaudhuri and Holbrook (2001)). Attitudinal loyalty was measured by a two statements on a seven point Likert scale: “I am committed to (retail brand)” and “I would be willing to pay a higher price for this (retail brand) over other brands” (see e.g. Chaudhuri and Holbrook (2001). The overall brand loyalty was measured averaging these four statements. Experience with online grocery shopping was measured with a statement where respondents indicate whether how much percent of the total groceries they shop could be accounted to online shopping. Lastly, to control for household size, the number of people in a respondent’s household was asked.

5.2.1.3 Choice design

Using a full factorial design would lead to a maximum number of choice sets in this study. In this study that would result in 4861_{choice sets per respondent. Although this would provide much} information about a respondent preferences, a full factorial design leads to fatigue effects and a decrease in attention and interest (Eggers & Sattler, 2011). Therefore, a fractional factorial design was chosen. The optimal number of choice sets is according to Egger and Sattler (2011) between 12 and 15. In order to keep respondents motivated and keep the reliability of the answers as high as possible, a 10 choice set design was chosen. Furthermore, minimal overlap, dominating design, a balanced design and orthogonally were controlled for by showing each level an equal times (Egger and Sattler, 2011). A dominating profile is according to Wiley and Chitturi (2010) a profile that ‘has

more of every benefit attribute and less of every cost attribute than every other profile in the choice set’.

The discriminant validity is the extent to which a construct is truly distinct from other constructs (Hair et al., 2010). High discriminant validity provides evidence that a construct is unique and captures some phenomena other measures do not. One could argue that some respondents confuse brand with ‘price’ as at Jumbo, the lowest price guarantee is provided and Albert Heijn and especially Spar is often perceived as expensive, however, it was mentioned that the focus is on delivery prices which should assert that respondents were fully informed about the different attributes and their meanings.

(27)

21

5.2.2 Mathematical specification

In this chapter, the conceptual model is translated in formula. With ICBC, choices of respondents are based on the overall utilities of the alternatives. Thus, the utility for respondent i for delivery option j is:

𝑈𝑖𝑗 = 𝑉𝑖𝑗+ 𝜀𝑖𝑗

Where:

V= the systematic utility component, the rational utility ɛ= the stochastic utility component, the error term.

As mentioned before, it is assumed that a delivery option is a combination of different attributes. Consumers attach linear, part worth or quadratic utilities to attributes, the systematic utility for respondent i for delivery option j is:

𝑉𝑖𝑗= ∑ 𝛽𝑖𝑘 𝑥𝑖𝑘 𝐾

𝑘=1

Where:

k=the number of attributes

β=part-worth utility of consumer i for attribute k x=dummy for the level of attribute i

A multinomial logit model was used to predict the probability that a respondent is selected from a given choice set:

𝑝𝑟𝑜𝑏(𝑗 | 𝐶) = 𝐸𝑥𝑝(𝑉𝑖𝑗) ∑𝑚𝑐=1𝐸𝑥𝑝(𝑉𝑖𝑗)

Where:

j= the alternative from choice set C m = number of choice sets

(28)

22

5.3 Data collection

The survey was created in Qualtrics and distributed solely via internet by discussion forums as Tweakers.nl, FOK.nl, Radar.nl and webwinkelforum.nl. The survey was further distributed via Facebook, using several groups to promote the survey. Tweakers.nl is a tech-forum and users might be interested in internet and new (technological) services, and might like the concept online grocery shopping. The same reason hold for the FOK.nl and webwinkelforum.nl forum that were used to distribute the survey. Actively engaging in discussions on Tweakers.nl and FOK.nl resulted in a lot of interest of the forums users. The same holds for the distribution on Facebook groups, where people were triggered to start a discussion. 15 Facebook groups were used to distribute the survey. The survey was distributed during an eleven day period, starting the 9th_{of November of 2015 until} November 20th_{of 2015.}

5.4 Plan of analysis

(29)

23

6 Results

In this chapter the results are presented. First a short description of the sample is provided, followed by descriptive statistics. Lastly, the results of the CBC analysis are described.

6.1 Sample description

Five days were used to actively promote the survey and resulted in roughly 95% of the responses, respectively 193, 452, 233, 180 and 100 responses. In total, 1185 people had started the survey.

The data set was tested for outliers, inconsistencies and respondents that only chose the no-choice option. 27 outliers2_{regarding income were found, of which the most extreme case is}

removed3_{. In order to cope with these outliers, this variable was binned and transformed into a} discrete variable4_{, as outliers can influence estimates of subsequent analysis. 341 respondents had} missing data and were deleted from the data set. 122 respondents chose solely the no-choice option, of which moreover in chapter 6.2. 49 respondents had missing data regarding income. For example, some filled in ‘something’ when was asked what monthly income they had. These missing values were predicted by regressing education, age and size of household on income (F=20,333; p=0.00; R2_{=0.1). The predicted values are used for these respondents. A total of 731 responses were useful} for further analysis. After cleaning the data set, the data set is prepared for the conjoint analysis. Transformation of the data was necessary as the conjoint study was set up with multiple choice questions for each choice set. Excels’ VBA was used to transform the data (see appendix B for the code used for the data transformation process).

6.2 Strictly no-choice option respondents

In total, 122 respondents chose the no-choice option. This could indicate that this group prefers offline grocery shopping over online shopping, but also could be the result of fatigue. This group spent significantly less time than respondents that also chose other options than the no-choice option (p=0.00). This was measured in two separate parts of the survey (42 vs. 46 seconds for the first six questions5_{and 68 vs. 72 seconds for question 7-12). Besides this, the respondents that only} chose the no-choice option found themselves less experienced with online shopping than other respondents (p.000; mean 4.4 vs. 5.1 on a seven point Likert scale). This might explain why they constantly preferred the ‘offline’ supermarket over the online options. The comments of this group may also serve as an indication why they chose the no-choice option consistently over the other options. Some argued that they do not like the idea of online grocery shopping, and mostly reasoned

2_{An outlier is defined when the value is 2.5 times the standard deviation of the mean (€1648)} 3_{This respondent filled in data that seemed highly implausible.}

(30)

(31)

25

6.3 Descriptive statistics

6.3.1 Demographics

The descriptive statistics are summarized in Table 5. The sample consisted of 21% males and 79% females. According the CBS (2015d), the population is divided in 49,5% males and 50,5% females, however, this sample was more in line with the distribution of EFMI (2014) who targets the shopping population. This deviation could be explained as the majority of females still do all grocery shopping. (EFMI, 2013). Therefore these deviations of the national average are justified.

Table 5, Descriptive statistics

The difference in age distribution could be explained by the fact that survey was distributed on the internet on social media and forums. The age group of 18 – 34 years was over represented, which might be an effect of these age groups being more active on social media and forums than the age group of 55+ is. Also, as people younger than 18 years do not shop groceries regularly, this might explained the low response rate in this group. As the groups 35-54 years old and 55 > years old were underrepresented, a weighting was applied to control for over- or under reporting an age group. The weighting factors for the age groups <18, 18-34, 35-54 and >55 are respectively 0.47, 0.19, 0.64 and 2.04. Generally it is not advised to assign extreme weights to groups of the weighting variable. As

(32)

26 majority of this data set was in the age group 18-34, it can be argued that this is not necessarily ‘wrong’ as this group might be more interested in online grocery shopping than older groups. Also, to make services more aligned with the preferences of the ‘future shoppers’, the age group <18 can be very interesting. However, in the sample also needs to be a good representation of the total shopper population, which is achieved with the weighting factors presented above. In Table 5, the effects of the weighting are displayed. The percentages were after weighting nearly identical to the shopper population.

The household size was larger compared with CBS (2014) and EFMI (2013, 2014) (mean 2.6 vs. 2.2). The sample has on average a higher education level than the Dutch average. One

explanation for this is that the age of the sample is lower than the Dutch average (see Table 6). Older generations, have enjoyed less educated than younger people (Vis and Moldenhauer, 2000). As this sample consists of more younger than older people, this might be an explanation. Other explanations might be that higher educated people are more interested in online grocery shopping and therefore had a higher intrinsic motivation to respond on this survey (Morganosky and Cude, 2000).

Onderwijscijfers (2014) Sample (N=731) Education None/unknown 2% 1% Primary school 7% < 1% Secondary school 23% 13% MBO 35% 41% HBO 21% 31% WO (+) 13% 8%

Table 6, Education level

This sample had a lower income than the Dutch average (mean: €1854), which could be explained by the fact the average age level was lower than the shopper population. However, this deviation is not extremely large (see Table 5). Although it was not measured directly, it could be that due to the lower age, more respondents were still studying, and thus have relatively low income levels.

6.3.2 Online shopping descriptives

The high supermarket penetration in the Netherlands has also been found in the results. 65% of this sample lived less than one kilometer away from a supermarket and 82% lived less than 1.5

(33)

27 The experience with online shopping of the respondents was higher than the Dutch average. 61.3%6 of all Dutch people have ever shopped online according to CBS (2015) and in this sample 82% of respondents had some experience with online shopping7_{(see Figure 2). This survey was conducted} online and people who do not shop online probably spend less time on the internet. As a

consequence, this group may be less likely to have filled in this survey. Furthermore, 46.5% of the Dutch people regularly buy online, which in this dataset was 50%8_.

Figure 2, Experience with online shopping

With respect to online grocery shopping, 30% of the respondents indicated to shop some of their groceries online. On average 6.4%9_{of all groceries are bought online, which is relatively high} compared with the Dutch average of 2.7% (GFK, 2015). When asked to indicate how much of their total grocery spending was online in five years’ time, 71% of the respondents indicated to shop some of their groceries online and on average 26% of all groceries are expected to be bought online. Besides this, the percentage of buying more than half of the groceries online shifted from 2.6% currently, to 15.3% in five years’ time.

6_{10.9000.000/16.972.336 = 61,3%}

7_{Only people who responded ‘disagree’ or ‘strongly disagree’ are regarded as having no experience with online}

shopping.

8_{People that responded with ‘agree’ and ‘totally agree’ are regarded to buy frequently online.}

9_{A respondent is believed to shop ‘some’ of their groceries online when they indicated to shop more than 5%}

of their groceries online. Some respondents indicated that they couldn’t answer 0% with the slider option and filled in 1%. To be sure to only include respondents who shop online, %5 or more is regarded as ‘buying some of their groceries online’.

0 10 20 30 40 50 60 70 80 90 100 R e sp o d e n ts sh ar e [% ]

Experience with online shopping

(34)

28

Figure 3, Supermarket shares

33% 25% 19% 8% _7% 1% 1% 1% 5% 0% 5% 10% 15% 20% 25% 30% 35%

Albert Heijn Jumbo LIDL ALDI Plus Coop Boni Nettorama Other

Supermarket shares

6.3.3 Store preferences

As can be seen in Figure Albert Heijn (33%), Jumbo (25%) and Lidl (19%) were the most popular supermarkets among the respondents by far. A remarkable result was that none of the respondents selected Spar as the supermarket they visit most. In general discount supermarkets were

overrepresented in this sample. In Figure 3, an overview of supermarket the most frequently visited supermarkets was provided (1%>).

Each respondent had to answer four questions10_{that reflect their loyalty to the supermarket that} they visit most (α = 0.745). The average score of loyalty is 4.79, 5.9 for behavioral loyalty and 3.7 for attitudinal loyalty. The two measures for attitudinal loyalty had a Cronbach’s alpha of 0.624 which makes it questionable. However, the cutoff point of 0.7 is only one guideline and other studies accept α = > 0.6 (Hair, Anderson and Tafham, 1998). Therefore the Cronbach’s alpha of 0.624 is accepted. The behavioral loyalty had a Cronbach’s alpha of α = 0.737 and was therefore accepted. Figure 4 shows an overview of the scores of the supermarkets11_.

All supermarkets were quite similar in terms of behavioral loyalty but differed in terms of attitudinal loyalty. Furthermore, discount supermarkets in this data set were relatively close together in terms of loyalty score, of which Lidls’ customers are most loyal. Also, full-service supermarkets, such as Emté and Plus, scored higher on attitudinal loyalty.

10_{See chapter 5.2.1}

11_{Supermarkets marked with an asterisk have a relatively low response rate (< 2%) and interpretations have to}

(35)

29

Figure 4, Store loyalty

Albert Heijn Jumbo Plus Aldi Lidl Boni* Coop Dirk* Emte* Hoogvliet* Nettorama* Poeisz* 1,0 2,0 3,0 4,0 5,0 6,0 7,0 1,0 2,0 3,0 4,0 5,0 6,0 7,0 Att itu d in al Lo yalt y Behavioral Loyalty

Store loyalty

(36)

30

Figure 5, Attribute form

-1 -0,8 -0,6 -0,4 -0,2 0 0,2 0,4 0,6 0,8 1 2,5 5 7,5 Sam ed ay N ext d ay Tw o d ay s Alb e rt H e ijn Ju m b o Sp ar Be zo rge n Pick-u p 1 h o u r 3 h o u r 5 h o u r N orm al N o rm al + d ru gs to re N o rm al + d r. + liq u o r U tility Attributes

Attribute form

Price Delivery day Brand Delivery option Time frame Assortment

6.4 CBC Analysis

In this section the CBC model is analyzed. After analyzing the socio-demographic characteristics of the respondents, this analysis gives an understanding in what respondents prefer in terms of online grocery shopping.

6.4.1 Model form

In this chapter, the model forms of the attributes were assessed. First, the utility levels of the attributes were visualized, which form an indication for the attribute forms. Secondly, the model fit of the models were assessed according the information criteria and hit rate. Figure 5, provides an indication of the form of the six attributes used in this model. It appears that brand and assortment have a part worth relationship and all other variables a linear relationship.

(37)

31 The following model was used to measure the utility V of delivery option j, for individual i:

𝑉𝑖𝑗= 𝛽1𝑗𝐵𝐸𝐿 + 𝛽2𝑗𝐵𝐸𝑀 + 𝛽3𝑗𝐵𝐸𝐻 + 𝛽4𝑗𝑃𝑟𝑖𝑐𝑒 + 𝛽5𝑗𝑇𝐹 + 𝛽6𝑗 𝐷𝐷 + 𝛽7𝑗 𝐷𝑂𝐻 + 𝛽8𝑗 𝐷𝑂𝑃

+ 𝛽9𝑗 𝐴𝑆𝑆 + 𝛽10𝑗 𝐴𝑆𝑀 + 𝛽11𝑗 𝐴𝑆𝐿

Where:

Vij= Utility for delivery option j

β1jBEL = Dummy variable for brand equity low for delivery option j β2jBEM = Dummy variable for brand equity medium for delivery option j β3jBEH = Dummy variable for brand equity high for delivery option j β4jPrice = Linear variable for price for delivery option j

β5j TF = Linear variable for time frame for delivery option j β6j DD = Linear variable for delivery day for delivery option j

β7j DOH = Dummy variable for at home delivery for delivery option j β8j DOP = Dummy variable for pick-up point for delivery option j β7j ASS = Dummy variable for small assortment for delivery option j β7j ASM = Dummy variable for medium assortment for delivery option j β7j ASL = Dummy variable for large assortment for delivery option j

Model Model form LL Χ2 _P-value _BIC _AIC3 _CAIC _R² _Hit

rate N-par 1 All attributes nominal -9447 -1372 <.001 18976 18931 18988 0.168 50% 12 2 Price as nominal _-9449 _-1368 _<.001 ₁₈₉₇₄ ₁₈₉₃₃ _{18985 0.168} _50% ₁₁

3 Price and delivery day as linear

-9450 -1366 <.001 18969 18931 18979 0.168 50% 10 4 Price, time frame

and delivery day

as linear -9457 -1366 <.001 18975 18942 18984 0.168 50% 9

5 Price, time frame, assortment and delivery day as

linear -9461 -1361 <.001 18976 18947 18984 0.168 50% 8

6 Price, time frame, assortment, delivery option and delivery day

as linear -9461 -1345 <.001 18976 18947 18984 0.168 50% 8

7 Price, time frame, assortment, delivery option, brand and delivery day as

linear -9514 -1239 <.001 19076 19050 19083 0.164 50% 7

(38)

32

6.4.2 Model selection

Standard choice models assume that people are homogenous in preference. However, it seems logical that different people have different preferences regarding online grocery shopping. Some people might have a tight budget and may choose an online grocery store as they provide their services for a low price. Others might have a larger budget and choose for convenience (for example a small time frame), which is more expensive.

Mixture models assume that utilities are distributed across people and assume that he respondents are allocated to certain segments with a certain probability. To test whether latent classes exist within this data set, latent class analysis is performed in Latent Gold 5.0. Models with two to seven segments are estimated to test which model performs best. Information criteria and the (managerial) interpretability of the segments are used to find the best fit. The Bayesian information criteria (BIC) and Consistent Akaike information criteria (CAIC) are used for accessing model fit (Andrews and Currim, 2003). These measures give higher penalties to more complex models and are preferred for large sample sizes and are thus preferred over other information criteria in this analysis. Besides the information criteria, management criteria are used to assess the segment solutions. A ten segment solution might be the best solution according the information criteria and have the highest hit rate; most managers would not advise to target ten separate segments as there are probably no financial resources available to manage 10 separate segments. Therefore both information criteria and management criteria are used (see Figure 6 and Table 8).

Figure 6, Information criteria

(39)

33 BIC(LL) _{AIC3(LL) CAIC(LL) N-par p-value Hit rate Class.Err. R²} 2-segment solution 16558 16391 16603 45 0.00 61% 0.024 0.305 3-segment solution 15966 15665 16047 81 0.00 65% 0.032 0.360 4-segment solution 15704 15270 15821 117 0.00 67% 0.041 0.405 5-segment solution 15466 14898 15619 153 0.00 68% 0.042 0.436 6-segment solution 15410 14709 15599 189 0.00 70% 0.052 0.456 7-segment solution 15367 14531 15592 225 0.00 72% 0.056 0.483

Table 8, Segment solutions

The fit of the segment solutions are discussed below. As can be seen in Figure 6 and Table 8, the five segment model had the lowest BIC, AIC3 and CAIC and thus had the best fit. However, the segment sizes of the 5-segment solution are relatively small (up to 8%) and therefore from a managerial perspective less desired. This also holds for the 6-segment and 7-segment solution. Although a one-to-one strategy probably has the best effect on consumers, in practice it is hard to manage seven different segments. To allocate enough monetary funds to these seven segments to target them effectively and efficiently is hard and unusual. However, the interpretability of a 7-segment becomes hard. Therefore the 7-segment model is not used. Besides this, many covariates and attributes are not significant. As the purpose of segments is to target consumers to satisfy their wants as specific as possible, it would be better to segment those consumers based on more covariates. Subsequently the four segment model was assessed. Again a small segment was found (8%) and given that most attributes were not significant as well, this model was not selected. At last, the three segment solution appears to be the best solution in terms of information criteria, managerial perspective and number of significant attributes and covariates. Hence, the 3-class choice model was used for further analysis. In Table 9, it can be seen that the segmented model performed better regarding BIC, AIC3, CAIC and had a higher hit rate compared to the aggregate model. Therefore, further analyses are executed with the three-class model.

BIC AIC3 CAIC N-par p-value Hit rate R² Class.Err.

Aggregate model 18969 18931 18979 10 0.00 50% 0.168 0 3-segment solution 15966 15665 16047 81 0.00 65% 0.360 0.032

(40)

34 In order to test the robustness of the model, a cross validation test was performed (see Table 10). A random sample of 50% of the respondents is drawn and model performance is compared with the aggregate model. Of the validation sample, the composition of gender and household size were largely the same. As can be seen in Table 10, the validation model underperformed the aggregate model, however, the differences are small and therefore the model was perceived as robust.

BIC AIC3 CAIC N-par p-value Hit rate R²

Aggregate model 18969 18931 18979 10 0.00 50% 0.168 Cross-validation 19035 18994 19045 10 .000 47% 0.145

Table 10, Cross validation

6.4.3 Segment parameters

(41)

35 Parameters Class1 Class2 Class3 Wald p-value Wald(=)

p-value Size 46% 40% 14% Attributes Brand* Albert Heijn 0.113 0.039 0.760 382.745 0.000 65.244 0.000 Jumbo 0.344 0.379 -0.063 Spar -0.457 -0.419 -0.697 Delivery option* Home delivery -0.102 0.228 1.295 290.255 0.000 269.627 0.000 Pick-up 0.102 -0.228 -1.295 Price* -0.376 -0.467 -0.290 1377.355 0.000 19.147 0.000 Delivery day* -0.714 -0.899 -0.258 445.401 0.000 23.458 0.000 Time frame* 0.022 -0.115 -0.097 21.424 0.000 19.920 0.000 Assortment** Normal 0.003 -0.069 0.184 16.105 0.013 6.085 0.19 Normal + drugs -0.088 -0.112 -0.154

Normal + drugs + liquor 0.085 0.179 -0.030

None option* 198.972 0.000

-5.164 -2.327 -2.387 1450.631 0.000

* = significant at 1% ** = significant at 5% *** = significant at 10% (moderately significant)

6.4.4 Relative importance of attributes

In order to see which attributes are most important to segments, the relative importance of attributes was calculated and presented in Table 12. In general, price was perceived as the most important attribute. In many other studies price is perceived as the most important attribute as well, and as discussed in chapter 2.2.2, delivery prices are considered the largest hurdle for online grocery shopping (Keen et al., 2004). Delivery day was considered as second important in general. This makes sense as consumers are looking for convenience and many consumers shop multiple times per week for their groceries and do not have a week worth of supplies at home.

Class 1 found price the most important attribute (43%) followed by delivery day (32%) and brand (18%). Delivery option (5%) assortment (4%) and time frame (2%) were of less importance. Class 2 also found price most important (38%), as well as delivery day (29%) and brand (13%). Again, the other attributes were relatively unimportant, delivery option (7%), time frame (7%) and

assortment (5%) and thus class 1 seemed to focus mostly on price and delivery day.

(42)

36 Class 3 preferences were different from class 1 and 2 as they found the delivery option the most important attribute (38%). This was followed by price (22%) brand (22%) and delivery day (8%). They found time frame (6%) and assortment (5%) the least important.

Rel. imp Class1 (46%) Class2 (40%) Class3 (14%)

Price 43% 38% 22% Delivery day 32% 29% 8% Brand 18% 13% 22% Delivery option 5% 7% 38% Assortment 4% 5% 5% Time frame 2% 7% 6%

Table 12, Relative importance of attributes

6.4.5 Segment WTP and covariates

After assessing whether segments differ in terms of preferences in attributes, the WTP on attribute level is calculated (see Table 13) and it was tested to see whether segments differ in terms of covariates (see Table 14). The most influential attributes regarding the WTP for service profiles are discussed in the segment descriptions. Regarding the covariates, in his study, income, education, experience with online shopping in general, behavioral loyalty, attitudinal loyalty, percentage of groceries shopped online and percentage of groceries expected to buy in five years’ time are used as covariates. All covariates are significant and are displayed in Table 14. Gender, average spending on groceries per week, household size.

WTP Class1 Class2 Class3

Brand Albert Heijn €0.30 €0.08 €2.63 Jumbo €0.92 €0.81 -€0.22 Spar -€1.22 -€0.90 -€2.41 Delivery option Home delivery -€0.27 €0.49 €4.47 Pick-up €0.27 -€0.49 -€4.47 Delivery day -€1.90 -€1.92 -€0.89 Time frame €0.06 -€0.25 -€0.33 Assortment Normal €0.01 -€0.15 €0.64 Normal + drugs -€0.23 -€0.24 -€0.53

Normal + drugs + liquor €0.23 €0.38 -€0.10

None option -€13.75 -€4.98 -€8.24

The Future of Grocery Shopping

The Future of Grocery Shopping

Master Thesis

MSc. Marketing Intelligence & MSc. Marketing Management

The Future of Grocery Shopping

Master thesis

MSc. Marketing Intelligence & MSc. Marketing Management

University of Groningen

Faculty of Economics and Business

J. (Jan) Nies

Papengang 38

9711 PA Groningen

T: +31 6 44744 003

Jan_nies@hotmail.com

Student number: 2611015

Supervisors:

Management summary

Preface

Table of contents

1 Introduction

2 Theoretical framework

2.1 Adoption of innovations

2.2 Consumer adoption of online grocery shopping

2.2.1 Relative advantage

2.2.2 Compatibility

2.2.3 Complexity

2.2.4 Trialability

2.2.5 Observability

3

Factors of influence on the WTP of delivery of online groceries

3.1 Retail brands

3.2 Time window of delivery

3.3 Shipping time

3.4 Assortment

3.5 Delivery options

3.6 Brand loyalty

3.7 Experience with online shopping

5 Methodology

5.1 Qualitative study

5.1.1 Procedure

5.1.2 Results

5.2 Choice Based Conjoint Analysis

5.2.1 Measures

5.2.2 Mathematical specification

5.3 Data collection

5.4 Plan of analysis

6 Results

6.1 Sample description

6.2 Strictly no-choice option respondents

6.3 Descriptive statistics

6.3.1 Demographics

6.3.2 Online shopping descriptives

Experience with online shopping

Supermarket shares

6.3.3 Store preferences

Store loyalty

Attribute form

6.4 CBC Analysis

6.4.1 Model form

6.4.2 Model selection

6.4.3 Segment parameters

6.4.4 Relative importance of attributes

6.4.5 Segment WTP and covariates