BRIDGING THE LAST-MILE OF E-COMMERCE

(1)

BRIDGING THE LAST-MILE OF E-COMMERCE

Matching last-mile delivery options with product types and customer characteristics by means of hypothesis testing

Master’s thesis Supply Chain Management

University of Groningen, Faculty of Economics and Business June 21, 2013

MARIËLLA JOANNE SETTELAAR Student number: 1831720 e-mail: m.j.settelaar@student.rug.nl Supervisors E. Ursavas K.J. Roodbergen ABSTRACT

Keywords: E-commerce; Last-mile delivery; Home delivery; Alternative delivery locations; Product types; Customer characteristics.

(2)

MSc Thesis Supply Chain Management – Mariëlla Joanne Settelaar Page 2

Most articles about e-commerce start with sentences that stress the importance and growth of e-commerce. In 2012, the Central Bureau of Statistics reported for example that seven out of ten people in The Netherlands have shopped online in 2011 and that online shopping accounted for more than ten percent of total sales in 2012. This suggests a bright future for e-commerce, and more firms to enter this way of doing business. E-commerce enables businesses to sell their products and services directly to consumers without establishing a physical point of sale (Park & Regan, 2004). However, from a logistics point of view, this involves high volumes of small deliveries and returns to and from a customer’s location, which is more costly and time consuming than delivering to retailers (Tarn, Razi, Wen, & Perez Jr., 2003; Park & Regan, 2004). This implies that the increase in e-commerce also puts increased pressure on managing the logistics cost component of a firm.

To stay profitable, delivery of the product to the individual customer has to be viable. Therefore, this study investigates whether offering some or all last-mile delivery options is desirable among different products types and among customer segments. As earlier research has suggested that differences can be expected for the choice of last-mile delivery option per product category (Keeney, 1999; Maldberger & Sester, 2005; Netwon, 2001; Thirumalai & Sinha, 2005), this research builds on to these notions. Furthermore, the purpose of marketing is to understand customer preferences and determine what products to offer, how to promote them, what prices to charge, and how to best deliver the product to the customer (Allenby & Rossi, 1999). Often, heterogeneity in customer preferences exist, which gives rise to differentiated offerings and market segments (Allenby & Rossi, 1999). Markets can be segmented in different ways, for example based on demographics, geographic region, or psychographics. Customer demographics such as age, gender, education, and income show a relationship with customer satisfaction with e-commerce (Lightner, 2003), but have not been related specifically to the last-mile issue. Furthermore, research concerning other market segmentation approaches in relation to last-mile delivery options is lacking. The customer characteristics that will be investigated in this study relate to demographic and geographic characteristics, more specifically income, household size, and type of residential environment.

(4)

MSc Thesis Supply Chain Management – Mariëlla Joanne Settelaar Page 4 2 RESEARCH QUESTION

The fact that last-mile delivery options are more costly than traditional warehousing indicates a necessity for e-commerce businesses to look for more efficient practices. Because online shopping is increasing not only in retail but also in other industries, the necessity of these delivery methods is also increasing. In the available literature, research mainly focuses on issues such as what types of last-mile delivery options exist (Agatz, Fleischmann, & Van Nunen, 2008; Weltevreden, 2008; Gevaers, Van de Voorde, & Vanelslander, 2010). Punakivi, Yrjölä, and Holmström (2001) go one step further and compare attended and unattended home delivery solutions in a simulation study, and Punakivi and Saranen (2001) also do this but specifically for the grocery industry. However, as stated before, some studies indicate that there are differences to be explored in last-mile delivery, particularly they describe that a difference can be expected for the choice of last-mile delivery per product category (Keeney, 1999; Maldberger & Sester, 2005; Thirumalai & Sinha, 2005). Maldberger and Sester (2005) hypothesize that customers expect different delivery modes among product categories and this hypothesis is supported by their data. However, their study uses a different classification of last-mile delivery options than most other studies that are described in this study (for example, the difference between ‘home’, ‘mail’ and ‘messenger’ delivery is not discussed in other articles, but these options are used in the study of Maldberger and Sester (2005) without explanation).

Furthermore, as stated in the introduction, research concerning market segmentation approaches in relation to last-mile delivery options is lacking. Customer demographics have only been related to overall customer satisfaction with e-commerce, but not specific specifically to the mile issue (Lightner, 2003). This indicates a gap between what is available in last-mile delivery options and how it can be better applied for certain products and customers, suggesting a contingency approach. Therefore, the following research question is proposed:

“Can the different last-mile delivery options be matched with product types and customer characteristics by means of hypothesis testing?”

(5)

MSc Thesis Supply Chain Management – Mariëlla Joanne Settelaar Page 5 3 LITERATURE

The critical link between consumer-based Internet ordering and the delivery of the product to the customer is often referred to as the last-mile (Esper, Jensen, Turnipseed, & Burton, 2003). It is that portion of the supply chain that deals with delivering products directly to the consumer (Kull, Boyer, & Calantone, 2007). Gevaers et al. (2010) define it as the final leg in a B2C delivery service, where the “consignment is delivered to the recipient, either at the recipient’s home or at a collection point.”

Traditional distribution of goods to retail shops involves deliveries of one or more boxes, pallets, or containers filled with homogenous goods. In contrast, e-commerce delivery has only one item for each address and requires a different service than traditional freight transportation (Dennis, 2011). The most significant impact of e-commerce on transportation is the increase in direct home delivery of smaller shipments. It has been one of the key factors leading to large losses for pioneering companies (Punakivi et al., 2001). Moreover, shipping costs are one of the biggest concerns for online customers (Allen, Thorne, & Browne, 2007). So, the provision of fast, reliable, and flexible delivery service at a reasonable price is the key to the success of online retail applications (Ehmke, 2012).

However, problems arise with the traditional home-delivery model of e-commerce if the customer is not at home. Park and Regan (2004) state that this is the most critical factor leading to higher operating costs for the e-tailer and lower customer satisfaction. The not-at-home problem is increasing due to inflexible working patterns of customers, increase in working women, and a growth of single-person households (Park & Regan, 2004; Weltevreden, 2008). Moreover, Dennis (2011) states that delivery costs are increased by the fact that residential areas have lower density, which increases travel times between stops. Customers also demand faster delivery, but there is no evidence that they are willing to pay extra for this service (Dennis, 2011).

Consequently, the likelihood of delivery problems should be reduced. This would lead to lower operating costs on the e-tailers’ side and higher customer satisfaction, which can help in smoothing the progress of e-commerce (Park & Regan, 2004). Because the traditional home-delivery model often encountered problems in home-delivery, other options have appeared throughout the past years. These other delivery options are called ‘last-mile delivery options’ and will be discussed more in-depth in the following section.

3.1 Types of last-mile delivery options

Punakivi and Saranen (2001) distinguish between the following types of last-mile delivery options: (1) attended home-delivery; (2) unattended private reception boxes (e.g. at the customer’s garage or yard); (3) attended pickup points (e.g. fuel stations, convenience stores); and (4) unattended pickup points (e.g. secured locker boxes at public places, such as railway stations or offices).

(6)

MSc Thesis Supply Chain Management – Mariëlla Joanne Settelaar Page 6 stations or convenience stores), or the customer’s home. This overview is adapted and supplemented with the concepts mentioned by Punakivi and Saranen (2001), which results in figure 3.1.

Figure 3.1 Types of last-mile delivery options (edited figure of Gevaers et al., 2010). 3.2 Trade-offs in choosing last-mile delivery options

There is a trade-off in choosing the type of last-mile delivery option. This trade-off is between consumer convenience and provider efficiency (Ehmke, 2012). For example, using a collection point is very convenient for the e-tailer as it overcomes not-at-home problems and has low delivery costs, however the customer has to travel to collect or return a package. This trade-off is depicted in table 3.1 below.

Home delivery

Clustering E-tailers’ DC or in-store delivery Distance or effort for e-tailer High Medium Low Distance or effort for customer Low Medium High

(7)

MSc Thesis Supply Chain Management – Mariëlla Joanne Settelaar Page 7 Delivery point Attended home delivery Private dropbox Neighbors Secured locker box Pickup from store or collection point Who covers the last mile? Logistics service provider Logistics service provider Logistics service provider Consumer Consumer Consumer at home problems Yes No Neighbor possibly not at home No No Failed

deliveries High Virtually none Medium Virtually none Virtually none Collection

times Not appropriate 24h When neighbor is at home 24h Collection point opening times Retrieval time for customers

None Very short Short Short-long Short-long Drop-off

time for service provider

Long Short Same as

attended home delivery

Very short Very short

Initial

investment Low High/medium Low Medium Low/medium

Delivery

costs High Low Medium Lowest Lowest

Operational

problems High number of failed deliveries Large number of boxes needed Failed deliveries if neighbor is not at home Customer

has to travel Customer has to travel

Table 3.2 Characteristics of last-mile delivery options (edited figure of Allen et al., 2007). As mentioned before, the traditional home-delivery model is expensive due to a high number of delivery failures (Park & Regan, 2004). Especially for low-value items such as groceries, cost-efficient processing of small transactions is a major challenge (Agatz et al., 2008). Home delivery is only convenient for the customer if someone is at home at the time of delivery as delivery failure leads to lower customer satisfaction (Park & Regan, 2004). For the e-tailer, investment in home delivery is initially low, but operating costs are high. To make this delivery option more viable, the use of time-windows (e.g. half an hour – an hour that specifies the time of delivery) has been introduced for attended home delivery (Punakivi & Saranen, 2001). This way, the customer can make sure that he or she is at home for the time of delivery. However, the use of time windows is not the norm for most e-tailers as the pre-arrangement of delivery time slots with customer increases the inflexibility in carriers’ fleet operations, leading to an extremely expensive delivery systems for e-tailers (Park & Regan, 2004).

(8)

MSc Thesis Supply Chain Management – Mariëlla Joanne Settelaar Page 8 investments in reception solutions at the consumers end (Punakivi et al, 2001). They state that unattended delivery is convenient, because customers are independent of delivery time windows (Punakivi et al, 2001). However, they require effort on the customers’ side, as they might have to travel.

Attended pickup points (e.g., fuel stations and convenience stores) might have benefits for all parties involved (Park & Regan, 2004). Weltevreden (2007) describes the benefits compared to traditional home delivery as less delivery failures, reduction of theft, time and cost savings for customers, carriers, and retailers, reduction in freight transport in residential areas, and more in-store traffic which can increase revenues in shops. However, the downside is that the customer might perceive a lower service level, and he or she might need to travel far to the particular pickup point. Also, the opening hours of the collection point should be taken into account, as the customers favor to collect a package in the evening, after working hours (Song, Cherrett, McLeod, & Guan, 2009; Thuiswinkel Markt Monitor, 2012).

3.3 Product types and the relationship with last-mile delivery options

A common way to classify products is proposed by Copeland in 1923. He distinguishes between convenience goods, shopping goods, and specialty goods. This classification is based on two dimensions: effort and risk (Murphy & Enis, 1986). In this definition, effort is considered as the amount of money, time, and energy the buyer is willing to expend to acquire a product, and risk is considered as the possibility that the product will not deliver the benefits sought (Murphy & Enis, 1986). Although this classification has been subject to criticism because of the minimal inclusion of social characteristics of goods and social psychology of buyer behavior (Mason, 2005), it was the first one to be well documented (Sheth, Garner, & Garrett, 1988;Mason, 2005). Furthermore, the classification of Copeland survived throughout the years and is endorsed by the American Marketing Association and the UK Chartered Institute of Marketing (Mason, 2005).

Aspinwall (1961) also made a product classification but uses other characteristics such as replacement rate, gross margin, time of consumption, and searching time. He uses three colors (red, orange, yellow) to describe three different product types, however these three product types align with Copeland’s three product classes (Taylor & Mujtaba, 2005). Table 3.3 shows the similarities between the classifications of Copeland and Aspinwall as proposed by Taylor & Mujtaba (2005).

Red goods (similar to Copeland’s

convenience goods)

Orange goods (similar to Copeland’s shopping goods)

Yellow goods (similar to Copeland’s specialty goods)

Replacement rate High Medium Low

Gross margin Low Medium High

Time of consumption

Low Medium High

Searching time Low Medium High

(9)

MSc Thesis Supply Chain Management – Mariëlla Joanne Settelaar Page 9 Furthermore, Bucklin (1963) also uses the traditional terms convenience, shopping, and specialty goods in his classification. However, his classification was based on the importance of the purchase to the customer and the experience and information available to the customer (Taylor & Mujtaba, 2005). Because Copeland his classification resonates throughout these other classifications and because this study builds onto earlier research of Thirumalai and Sinha (2005) that use Copeland as a reference, this study will also distinguish between convenience, shopping, and specialty goods.

Copeland (1923) proposed that convenience goods are goods that the consumer usually desires to purchase frequently, immediately, and with a minimum effort. It is the type of product that is lowest in both effort and risk (Murphy & Enis, 1986). The consumer is not willing to travel a great distance from his place of residence to the place of purchase, not visit more than one store, and not devote much time negotiating a purchase (Kleimenhagen, 1966). According to Bucklin (1963), the product is preferred to be obtained from the most accessible store. Furthermore, in selection and purchase of convenience goods, customers do not experience significant levels of risk (Thirumalai & Sinha, 2005). Examples are groceries, home supplies, and office supplies. Shopping goods are defined as those goods that the consumer usually wishes to purchase only after comparing quality, price, and style in a number of stores (Kleimenhagen, 1966). The customer would travel a considerable distance and is willing to visit more than one store to compare items (Kleimenhagen, 1966). Shopping stores are those stores that the customer will go to when he or she does not have a completely defined preference for the product, and will go to these stores to gain information to complete product preference (Bucklin, 1963). Consumers also experience moderate levels of risk with these products (Thirumalai & Sinha, 2005). An example is men’s and women’s apparel. These products (mostly apparel) are often characterized by the need to touch, feel, and try on (Grewal, Iyer, & Levy, 2004).

Specialty goods are those goods that have a particular attraction for the consumer so that he is willing to make a special purchasing effort (Kleimenhagen, 1966). The consumer has full knowledge of the product he wishes to purchase and is willing to expend considerable effort in order to get the desired item (Kleimenhagen, 1966). It is defined to be highest on both the risk and effort dimensions, in terms of time and money spend (Murphy & Enis, 1986). The customer is willing to travel a great distance and is likely to visit only one store. Bucklin (1963) calls this specialty stores as customers will go there even though the store may not be the most accessible. The customer would not accept a substitute and would postpone purchase until the particular item can be found (Kleimenhagen, 1966). Within this the product category, the customer perceives a high marginal product value, in either owning or consuming the product (Taylor & Mujtaba, 2005). Examples include notebooks, jewelry, and personalized or customized products.

(10)

MSc Thesis Supply Chain Management – Mariëlla Joanne Settelaar Page 10 per product type. The result was that customer expectations and satisfaction with order fulfillment was different for specialty goods than they were for convenience or shopping goods. They posit that there is a need to develop customized approaches to order fulfillment across product types in general (Thirumalai & Shina, 2005). Madlberger and Sester (2005) hypothesize that consumers expect different modes of delivery for different product categories. This hypothesis was supported, but overall home delivery was preferred for all product categories. They did not look at what customers prefer if home delivery is not possible, which might also be interesting for e-tailers due to the experienced not-at-home problems as suggested by Dennis (2011). Zhou et al. (2004) state that customer acceptance of online shopping may vary across product types. However, they did not look at how products are delivered to customers. Furthermore, they argue that most research in online shopping either deals with a single product type for the sake of simplicity and controllability or failed to examine the moderating effect of product type when more than one product type was used, which makes their findings incomparable (Zhou, Dai, & Zhang, 2007). This study, however, includes three different product types.

3.4 Customer characteristics and the relationship with last-mile delivery options

A challenge in managing logistics is to develop target segments of customers that can be served profitably by distinct, rationalized pipelines (Fuller, O’Conor, & Rawlinson, 1993). They state that “one size does not fit all.” According to Gitman and McDaniel (2008), markets can be segmented in different ways, examples include demographic segmentation (e.g. age, income, household size), geographic segmentation (e.g. regional location, population density), and psychographic segmentation (e.g. attitudes, values, lifestyle, personality).

(11)

MSc Thesis Supply Chain Management – Mariëlla Joanne Settelaar Page 11 people without a car might have more difficulty in accessing an alternative delivery location. Furthermore, people that are time-poor might not be at home often. This can be explained by the increasing not-at-home problems due to customer work patterns, increase in working women, and growth in single-person households (Park & Regan, 2004; Weltevreden,2008). As delivery failures influence customer satisfaction, these changes in customer characteristics can also influence the delivery option that a customer prefers.

Moreover, the customer’s geographic location in relation to last-mile delivery has not been addressed in current literature so far. As already mentioned in the previous section, a move away from traditional home delivery leads to an increase in customer effort, in the sense that he or she needs to travel to the alternative delivery location. People who do not have a car or live in rural areas might not have sufficient access to the available alternative delivery locations, thus might prefer home delivery. Also, the distance to convenience stores or supermarkets can influence the customer’s decision whether he or she wants it delivered there. Farag, Weltevreden, Van Rietbergen, Dijst, and Van Oort (2006) state that little geographical research has been conducted concerning the spatial distribution of e-shoppers, therefore they have studied whether spatial variables (residential environment and shop accessibility) have an influence on shopping. They have found that spatial attributes have an influence on e-shopping, but this effect varies for the different stages of the e-shopping process (e.g., searching or buying) and for the type of product. Two hypotheses were supported, namely people in strongly urbanized areas are more likely to search and buy online, and people with low shop accessibility (e.g., rural areas) buy more products online (Farag et al., 2006). These statements seem contradicting, but the support for these two findings depends on the type of product (book, CD, clothing, travel tickets, hardware and software). For travel tickets, the first finding is supported that people in urban areas search and buy more online. For CDs and similar products, the latter hypothesis stating that people in rural areas buy more products online is supported. However, their study did not look at how these products were delivered to customers. Johnson, Yoo, Rhee, Lennon, Jasper, and Damhorst (2006) also have studied the channel use among rural customers. However, both of these studies did not focus on the last-mile delivery decision, but rather on the likelihood of adopting e-commerce in general.

4 HYPOTHESES DEVELOPMENT

4.1 Product types and last-mile delivery options

(12)

MSc Thesis Supply Chain Management – Mariëlla Joanne Settelaar Page 12 E-commerce is particularly convenient for customers in that it can save time and money (Punj, 2012). Convenience goods are characterized by customers as being low effort and low risk products. Examples include groceries and home and office supplies. They are often consumed or used shortly after buying (Maldberger & Sester, 2005). Moreover, convenience goods are most often not high value items (Miracle, 1965). That is why we think that, translated to an e-commerce environment; this means that the customer is most likely not using e-e-commerce for buying convenience goods to save money, but rather to save time. The last-mile delivery option that fits this description is traditional home delivery as the customer has the product delivered closest to their home. Furthermore, according to Bucklin (1963), convenience goods are preferred to be obtained from the most accessible store. In case of last-mile delivery options, the option that is most accessible regarding opening hours is the locker box, which might be a reason to expect that picking up from a locker box is preferred when home delivery is not possible.

For shopping goods, customers are willing to put in more effort and perceived risk is moderate. They require more extensive buying decisions, are more durable and usually not consumed immediately (Li & Gery, 2000). Examples include apparel and books. They are also expected to be purchased more frequently online, because the risk is moderate, the customer likes to ‘see and feel’ the product (e.g., ordering the product to see and feel it before deciding to keep the product), and the customer has access to a wider range of products (Thirumalai & Shina, 2005; Dennis, 2011). As mentioned before, home delivery is preferred for all product types (Maldberger & Sester, 2005). However, because shopping goods are usually not consumed immediately, we believe that customers might consider other delivery options if this saves delivery fees, for example.

For specialty goods, the customer experiences significant levels of risk and the products are relatively high in value (Miracle, 1965; Murphy & Enis, 1986; Thirumalai & Shina, 2005). Examples include notebooks or a personalized item. Because the customer experiences a significant level of risk, we think that customers want to have it delivered at home or pick the item up from an attended alternative delivery location, which are the physical store of the brand or chain or the supermarket. The point of this discussion is to make clear that the actual relationship between the product type and last-mile delivery option is not clear yet. Therefore, the following hypothesis has been formulated.

= There is a relationship between the product type and the choice of last-mile delivery option.

4.2 Customer characteristics and last-mile delivery options

(13)

MSc Thesis Supply Chain Management – Mariëlla Joanne Settelaar Page 13 the least effort delivery method for him or her, being it home delivery, even if delivery fees are being charged for this. Customers that have a low income can value saving money more than saving time, thus they might choose for a lower delivery cost option for which they have to put in more effort, e.g. pick the product up from an alternative location. However, customers with a lower income can also be time poor. Consider students; they need to be at school, which means there is no time to work full-time. They can choose for a delivery option that saves them time and effort (e.g., home delivery, delivery at neighbors), or a delivery option that saves them money but takes time to collect the order (e.g., in-store pickup). Therefore, we propose that there is a relationship between the customer’s income and the last-mile delivery option that is opted.

= There is a relationship between the customer’s income and the choice of last-mile delivery option.

Furthermore, we propose that the structure of a customer’s household has an influence on the last-mile delivery option that is chosen. Farag et al. (2006) did not found significant effects of the household composition (singles, households without children, and households with children) on e-shopping (searching and buying); however, this research did not look at the delivery of these goods. Online shoppers prefer to receive their order in the evening (Thuiswinkel Markt Monitor, 2012), after working hours (Song et al., 2009). This indicates that most people expect not to be home during the daytime, due to work, study, or other activities. As Dennis (2011: 32) states, “a key factor that determines the success of a home delivery operation is the availability of someone to receive the delivery.” With a growing number of houses empty during the daytime and with delivery times between 8AM and 5PM, difficulties are predictable, he says.

We believe that the size of the household can influence the choice of the last-mile delivery option, which is consistent with a statement of Gould and Golob (1997). However, Gould and Golob (1997) did not study this. They only mention that a decrease in family unit size might deter home delivery. If the customer expects that someone could be home at the time of delivery, attended home delivery is more likely to succeed, hence reducing delivery failures. More people living in a household might increase the chance of successful attended home delivery, thus customer’s might choose more often for home delivery if there are more members in a household. However, concluding this might be distorted, as there are situations that signal the opposite. Consider student houses and households with children that go to school during the daytime. Therefore, before claiming such a statement, we propose the following.

= There is a relationship between the customer’s household size and the choice of last-mile delivery option.

(14)

MSc Thesis Supply Chain Management – Mariëlla Joanne Settelaar Page 14 especially value the convenience of online shopping, being it having the product delivered at home as opposed to picking it up somewhere else.

Furthermore, in urban areas there are likely to be more alternative delivery locations than in a small town. Farag et al. (2006) mention that consumers in urban areas are more likely to shop online because they are younger, better educated, have higher incomes, and are more time constrained. On the other hand, Farag et al. (2006) also state that when people have a lower accessibility to shops, the number of products bought online increases. These two statements both indicate an increase in using e-commerce in urban and rural areas; however nothing is said about how the last-mile is covered. Consistent with the reasons why people in rural areas shop online, namely lower shop accessibility, we propose that accessibility to alternative delivery locations such as post offices, kiosks, gas stations et cetera, is also lower. Therefore, we propose that there is a relationship between the customer’s geographic location and the last-mile delivery option the customer prefers.

= There is a relationship between the customer’s residential area and the choice of last-mile delivery option.

4.3 Conceptual model

While the previous sections suggest that only the product type and some customer characteristics have an influence on the last-mile option decision, it should be noted that this is not the suggestion of this thesis. There can be a lot of other reasons why a certain last-mile delivery option is preferred over others, for example customer personality, customer experience with e-commerce, and product size and shape. However, these options are outside the scope of this study. A pilot study has been conducted to make sure other variables (such as product size and shape, monetary value of the product) are excluded in the survey questions. What is within the scope of this study are the two components discussed above, the product type and customer characteristics. The results of this study are still relevant as they can be used (to build hypotheses) for future research that includes other variables.

(15)

MSc Thesis Supply Chain Management – Mariëlla Joanne Settelaar Page 15 Figure 4.1 Conceptual model.

5 METHODOLOGY

The purpose of this study is to find out whether a customer’s choice for a last-mile delivery option has a relationship with product type and customer characteristics. This section will elaborate on how to get an answer on the research question. The structure of this section is consistent with what Cooper and Schindler (2008) propose: research design, sampling design, pilot test, data collection, and data analysis and interpretation.

5.1 Research design

First of all, Cooper and Schindler (2008) distinguish between exploratory and formal studies. The difference is that exploratory studies have the objective of discovering future research tasks while a formal research design is used to test the hypothesis or research questions posed. Based on this distinction and the research question that is formulated, this research is positioned as formal research. This does not imply that there are no exploratory elements to be found in this study, the results can direct future research towards additional hypotheses.

(16)

MSc Thesis Supply Chain Management – Mariëlla Joanne Settelaar Page 16 if there is a webshop with a database available from which information could be used to test the hypotheses. In the communication study approach, the researcher questions the subjects and collects their responses by personal or impersonal means (Cooper & Schindler, 2008). Because information in databases of webshops can have limitations regarding the last-mile delivery options they are offering (e.g., how can we research whether a person would like to pick his or her order up from the supermarket or a locker box if the webshop is not offering all last-mile delivery options) and because of time and resource limitations, the communication study approach is opted for. More specifically, the survey method is chosen due to its advantages to obtain large amounts of quantative data that can be used to test the proposed hypotheses. Another advantage is that by using scenarios in the survey, other factors that can influence customer’s decision to choose for last-mile delivery options can be excluded.

A survey with closed-ended questions is used, because respondents might be unfamiliar with the types of last-mile delivery options they can possibly choose from. Also, when completing an online order in real-life, customers need to make a choice between the e-tailer’s predefined delivery options. This real-life situation is reflected by presenting scenarios (the three different product types) to respondents and having them choose among a predefined set of last-mile delivery options.

5.2 Sampling design

The target population of this study includes all Dutch Internet users. This study is limited to The Netherlands for several reasons. First of all, extending the survey to a larger international population would require translating the survey, and any inconsistencies between the two versions would harm internal validity. Secondly, cultural or infrastructural aspects might cause countries to differ regarding customer preferences for collecting a package; for example more than half of the shopping trips in The Netherlands are made by foot or bicycle (Dielemen et al., 2002). Pucher and Buehler (2008) state that physical infrastructure and policies of countries such as Germany and The Netherlands are the ingredients for a higher bike use compared to car dominated countries such as the USA. Rietveld and Daniel (2004) add to this that this is not only caused by physical infrastructure and policies, but also due to cultural differences.

(17)

MSc Thesis Supply Chain Management – Mariëlla Joanne Settelaar Page 17 according to the last count in 2011, the overall Dutch population consisted of 49% males and 51% females (Wobma, 2011). As these percentages are comparable, the sample also is selected by ensuring approximately equal proportions in both the male and female group to guard against skewed results based on gender. This is done by asking participants to identify other possible respondents based on their gender.

5.3 Pilot test

The survey was created in consultation with two professors from the University of Groningen. For a period between two to three weeks, the survey was updated and revised based on discussions with these professors. Moreover, a pilot test has been done with target respondents and industry experts. This is consistent with Karlsson (2009), who suggests that a pre-test of a questionnaire should be submitted to three types of people: colleagues, industry experts, and target respondents. This allowed us finding weaknesses in the questionnaire, hence improving the reliability of the measuring instrument (Baarda & De Goede, 2006).

The feedback obtained from the pilot tests was used to update and revise the survey. Not only regarding the survey length, clarity and wording, and the ease of answering the questions, but also regarding to the product types that were proposed in the survey. After a discussion, the following three products were selected: a bag of groceries (convenience), a pair of shoes (shopping), and a personalized photo album (specialty). Moreover, pilot test respondents mentioned that in the survey it should be emphasized that these products all fit in the same size box and do not differ regarding the monetary value. Also, to enhance clarity of the survey, at each page of the Web-based questionnaire it was decided to show a picture of these three products. The final product after the pilot test was the survey that is used in this study, which can be found in Appendix A.

5.4 Data collection

A self-administered Web-based questionnaire in Qualtrics has been used to collect primary data, for the following reasons: low costs, rapid data collection, respondent anonymity, and larger geographic coverage. Anonymity is emphasized at the start of the survey to reduce non-response (Cho & Larose, 1999). The survey was both distributed and returned via the Internet, which has the advantage that large numbers of quantitative information can be downloaded. Furthermore, a web based survey has the advantage that people who are reached by the e-mail or electronic link inherently use the Internet. The respondents do not have to be regular online shoppers, because they have to make decisions in the survey based on scenarios that are given in the study.

(18)

MSc Thesis Supply Chain Management – Mariëlla Joanne Settelaar Page 18 Variable Values N % Sex M 100 49,5 F 101 50,5 Age <25 88 44,0 25-40 74 37,0 >40 38 19,0

Education Primary school 2 1,0

Secondary education 32 15,9

Intermediate vocational education

(MBO) 28 13,9

Higher education or university Bachelor 91 45,3

University Master 48 23,9 Household size 1-2 92 47,5 3-4 80 40,0 5-more 25 12,5 Income <€1500 106 53,0 €1500-€3000 47 23,5 >€3000 27 13,5

Do not want to say 20 10,0

Residential

area Urban 122 61,0

Suburban 21 10,5

Rural 57 28,5

Table 5.1 Sample characteristics. 5.5 Data analysis and interpretation

(19)

Variable Type Values How in the survey

1 Product type Nominal Convenience (bag of

groceries), shopping (shoes), specialty (personalized photo album)

Framed in the question as scenario

2 Last-mile delivery

option Nominal Home delivery, picking up from alternative location Multiple-choice for each product type 3 Delivery fees Interval No difference, €4 for home

delivery, €8 for home delivery

Framed in the question 4 Alternative delivery

location

Nominal Physical store of the brand/chain where the product is ordered,

supermarket, public locker box

Multiple-choice for each product type

5 Opening hours Nominal 09.00-17.00 (physical store), 08.00-20.00 (supermarket), and 24 hours per day (locker box)

Framed in the question

6 Income Interval <€1500, €1500-€3000,

>€3000, do not want to say Multiple-choice 7 Household size Interval 1, 2, 3, 4, 5 or more Multiple-choice 8 Residential area Nominal Urban, suburban, rural Multiple-choice

Table 5.2 Variables in this study.

These variables are not complex or abstract, but rather ‘observable’ (unlike for example customer satisfaction). The different product types have been specified in discussion with pilot test participants. The last-mile delivery options and alternative delivery locations also include concrete measures, as they fall in the theoretical domain of this topic. Because this classification is already made in literature (see literature section), content validity is increased (Karlsson, 2009). A notion should be made regarding the variable ‘alternative delivery location’, as in table 5.2 it can be seen that this variable has only three possible values. In the literature section, more alternative delivery options have been discussed, however this study only focuses on these three categories as adding more options can lead to a primacy or recency effect (Cooper & Schindler, 2008). Also, other last-mile delivery options (such as a kiosk or tobacco store instead of a supermarket) do not differ from an operational perspective for the e-tailer.

Because these measures are concrete, chances of measuring something else than what is meant is small, which increases construct validity (Baarda & De Goede, 2006). The measures appear to be valid “on the face of it”. According to Cozby (2005), face validity is the simplest way to argue that a measure is valid, as the evidence involves only involves a judgment of whether the content of the measure appears to actually measure the variable. The questions in have been made concrete and other factors such as product monetary value, size of the box or distance to alternative delivery locations have been made explicit, to avoid the respondent interpreting the question from his or her specific situation.

(20)

MSc Thesis Supply Chain Management – Mariëlla Joanne Settelaar Page 20 and the mean of these values will depend on how many members each category has (Field, 2009). Therefore, when only nominal variables are measured, frequencies need to be analyzed. As we deal with nominal and interval data, a nonparametric test is necessary to test the hypotheses. However, some variables such as ‘product type’ and ‘last-mile delivery option’ include more than two values, which makes a binomial test inappropriate. Other tests that can be applied for nominal data are the Pearson chi-square, Fishers exact test, the likelihood ratio, or loglinear analysis (Field, 2009). Fishers exact test and the likelihood ratio are discarded, as they are both preferred for small sample sizes. Furthermore, Fishers exact test is normally used for 2x2 tables. Loglinear analysis is applicable for tests that involve more than two categorical variables, however as some variables are embedded in the question of the survey, loglinear is not possible. The Pearson chi-square test can be used to test a relationship between two categorical variables that have more than two values (Keller, 2009; Field, 2009). Moreover, Pearson chi-square is useful in cases of one-sample analysis (Cooper & Schindler, 2008). Therefore, this test is used. The Pearson Chi-square test results in contingency tables such as provided in table 5.3.

Product

type Home delivery Picking up alternative location Total Convenience Count Expected count Shopping Count Expected count Specialty Count Total Count Expected count

Table 5.3 Example contingency table Pearson chi-square.

The idea behind the Pearson Chi-square test is that it compares the observed frequencies (O) in certain categories to the expected frequencies (E) in those categories as calculated by chance (Field, 2009). The expected frequency per cell is calculated as follows, where N is the total amount of valid cases in a test:

E = (observed row total * observed column total) / N; The chi-square value is calculated by the following formula:

= ;

To interpret the chi-square value, the degrees of freedom of the particular table has to be taken into account. The degrees of freedom (Df) is calculated as follows:

(21)

MSc Thesis Supply Chain Management – Mariëlla Joanne Settelaar Page 21 As an example, table 5.3 has (3-1)*(2-1) = 2 degrees of freedom. To determine whether there is a significant relationship between the two variables as proposed, the value of the chi-square statistic needs to be compared with the critical chi-square value. In most statistics books, a table with critical chi-square values can be found in the appendix. The degrees of freedom and the alpha level are needed to determine the critical chi-square value. The most common level of alpha is .05 (Cooper & Schindler, 2008), therefore this alpha level is chosen for this study. When the calculated chi-square is bigger the critical chi-square, it can be concluded that there is a significant relationship between the two variables. When using SPSS, a p-value is directly calculated. In case the p-value is lower than the proposed alpha level, we will also conclude that there is a significant relationship between the two variables.

Assumptions that must be adhered to before performing the Pearson chi-square test are that each case contributes only to one cell of the contingency table (Field, 2009) and the sample size n must be large enough so that the expected count in each cell is greater than or equal to 5 (Moore & McCabe, 2006). Therefore, in the data collection phase there was no predefined number of respondents, but data collection was finished once these assumptions were met. 6 ANALYSIS AND RESULTS

In this section, the results will be presented. Each section starts with an overview with the outcomes of the tests with that particular independent variable (product type, income, household size, and residential area). An asterisk in the column ‘p-value’ means that the relationship is significant on an alpha level of 0,05.

For the tests with income as an independent variable, the respondents that answered ‘I do not want to say’ had to be filtered out before analysis. For income, household size, and residential area, the tests had to be performed while keeping the product type constant. Otherwise, the same respondent’s characteristics would be measured three times in one analysis.

6.1 Product types and last-mile delivery options Test Independent

variable

Dependent variable

p-value

N df 1 Product type Last-mile delivery option (home delivery versus

pickup)

0,040* 628 2 2 Product type Last-mile delivery option (home delivery €4

versus pickup)

0,072 625 2 3 Product type Last-mile delivery option (home delivery €8

versus pickup)

0,552 621 2 4 Product type Alternative delivery location 0,000* 607 4 5 Product type Alternative delivery location with opening hours

given

(22)

MSc Thesis Supply Chain Management – Mariëlla Joanne Settelaar Page 22 alternative location when there is no difference in delivery fees (p = 0,040) or choosing between alternative delivery locations (p = 0,000). When choosing between home delivery and picking up from an alternative location, home delivery is preferred among all product types (table 6.2). Convenience and shopping goods, however, are more often picked up from an alternative location compared to specialty goods.

Product type Home delivery Picking up from an alternative location Convenience Count 162 (77,5%) 47 (22,5%) Expected count 169,4 (81,1%) 39,6 (18,9%) Shopping Count 166 (79,0%) 44 (21,0%) Expected count 170,2 (81,0%) 39,8 (19,0%) Specialty Count 181 (86,6%) 28 (13,4%) Expected count 169,4 (81,1%) 39,6 (18,9%)

Table 6.2 Product types and last-mile delivery options.

The relationship between the product type and last-mile delivery option is only significant when there is no difference in delivery fees for home delivery and picking up from an alternative location. When delivery fees of €4 are being charged for home delivery, the relationship between the product type and last-mile delivery option being home delivery versus picking up from an alternative location is not significant anymore (p = 0,072). Only 29,3% of the respondents has chosen for home delivery for all product types when delivery fees are €4. In case delivery fees for home delivery increase to €8, only 8,2% of the respondents opts for home delivery for all product types. The effect of delivery fees on the choice of last-mile delivery option for all product types combined is shown in table 6.3. Because of this, we conclude that the product type does have a relationship with the last-mile delivery option only when there is no difference in delivery fees for the two proposed options (home delivery versus picking up from an alternative location).

Home delivery Picking up from alternative location No price difference 509 (81,1%) 119 (18,9%)

€4 home delivery fee 183 (29,3%) 442 (70,7%) €8 home delivery fee 51 (8,2%) 570 (91,8%)

Table 6.3 Delivery fees and last-mile delivery options.

(23)

MSc Thesis Supply Chain Management – Mariëlla Joanne Settelaar Page 23 Product type Physical store brand or chain Supermarket Locker box

Convenience Count 58 (28,7%) 125 (61,9%) 19 (9,4%) Expected count 85,9 (42,5%) 74,5 (36,9%) 41,6 (20,6%) Shopping Count Expected count 109 (53,7%) 86,3 (42,5%) 45 (22,2%) 74,9 (36,9%) 49 (24,1%) 41,8 (20,6%) Specialty Count Expected count 91 (45,1%) 85,9 (42,5%) 54 (26,7%) 74,5 (36,9%) 57 (28,2%) 41,6 (20,6%) Table 6.4 Product types and alternative delivery locations.

In case opening hours are provided, the relationship between the product type and alternative delivery location is still significant (p = 0,000). However, the results show a decrease for all product types in choosing for a physical store (decreased by more than half), and an increase for choosing a locker box for shopping and specialty goods (almost doubled). This can be seen in table 6.5. The physical store of the brand or chain is least preferred among all product types, because this option has the most limited opening hours (09.00-17.00, compared to 08.00-20.00 of the supermarket and 24 hours per day of the locker box).

(24)

MSc Thesis Supply Chain Management – Mariëlla Joanne Settelaar Page 24 6.2 Income and last-mile delivery options

Test Independent variable

Dependent variable Constant p-value

N df 6(a) Income Last-mile delivery option (home

delivery versus pickup)

Convenience 0,381 180 2 6(b) Income Last-mile delivery option Shopping 0,075 180 2 6(c) Income Last-mile delivery option Specialty 0,130 180 2 7(a) Income Last-mile delivery option Convenience 0,130 180 2 7(b) Income Last-mile delivery option Shopping 0,341 180 2 7(c) Income Last-mile delivery option Specialty 0,258 180 2 8(a) Income Alternative delivery location Convenience 0,024* 180 4 8(b) Income Alternative delivery location Shopping 0,092 180 4 8(c) Income Alternative delivery location Specialty 0,009* 180 4 9(a) Income Alternative delivery location with

opening hours given

Convenience 0,080 180 4 9(b) Income Alternative delivery location with

opening hours given

Shopping 0,051 180 4 9(c) Income Alternative delivery location with

opening hours given

Specialty 0,081 180 4 Table 6.6 Outcomes chi-square tests with income as independent variable.

The customer’s income does not show to have a significant relationship with the last-mile delivery option, being it home delivery versus picking up from an alternative location, for any of the product types (table 6.6). Home delivery is chosen more often among all product types. However, when delivery fees are taken into account, an increase in picking up from an alternative location can be seen for all product types across all income groups. This observation is shown in table 6.7 and is consistent with what can be seen in an earlier analysis in table 6.3. In this table, all product types have been combined.

Income Home delivery and picking up same Home delivery €4 Home

delivery

Picking up from an alternative location

Home delivery Picking up from an alternative location Less than €1500 271 (85,2%) 47 (14,8%) 85 (26,7%) 233 (73,3%) Between €1500-€3000 102 (72,3%) 39 (17,7%) 38 (26,9%) 103 (73,1%) More than €3000 62 (76,5%) 19 (23,5%) 34 (42,0%) 47 (58,0%) Table 6.7 Income and delivery fees: all product types combined.

(25)

MSc Thesis Supply Chain Management – Mariëlla Joanne Settelaar Page 25 6.9). However, higher incomes use the locker box more (51,9%) compared to lower (20,7%) and middle (24,0%) incomes.

Income Physical store of

brand or chain

Supermarket Locker box Less than €1500 Count Expected count 34 (32,0%) 30 (28,3%) 68 (64,2%) 66 (62,3%) 4 (3,8%) 10 (9,4%) Between €1500-€3000 Count Expected count 12 (25,5%) 13,3 (28,3%) 26 (55,3%) 29,2 (62,2%) 9 (19,2%) 4,4 (9,4%) More than €3000 Count Expected count 5 (18,5%) 7,7 (28,4%) 18 (66,7%) 16,8 (62,0%) 4 (14,8%) 2,6 (9,6%) Table 6.8 Convenience goods: income and alternative delivery locations.

Income Physical store of

brand or chain

Supermarket Locker box Less than €1500 Count Expected count 54 (51,0%) 47,1 (44,4%) 30 (28,3%) 28,3 (26,7%) 22 (20,7%) 30,6 (28,9%) Between €1500-€3000 Count Expected count 21 (44,7%) 20,9 (44,5%) 10 (21,3%) 12,5 (26,6%) 16 (34,0%) 13,6 (28,9%) More than €3000 Count Expected count 5 (18,5%) 12,0 (44,4%) 8 (29,6%) 7,2 (26,7%) 14 (51,9%) 7,8 (28,9%) Table 6.9 Specialty goods: income and alternative delivery locations.

This relationship is no longer significant when opening hours are given for convenience goods (p = 0,080) and specialty goods (p = 0,081). What can be seen is that the choice for a locker box as a delivery option increases for all income groups for convenience and specialty goods when opening hours are given (compare table 6.8 with table 6.10 and table 6.9 with table 6.11). This suggests that the opening hours have an influence on the choice of the alternative delivery location.

Income Physical store of

brand or chain

(26)

Income Physical store of

brand or chain

Supermarket Locker box Less than €1500 Count Expected count 25 (23,6%) 21,2 (20,0%) 41 (38,7%) 37,7 (35,6%) 40 (37,7%) 47,1 (44,4%) Between €1500-€3000 Count Expected count 10 (21,3%) 9,4 (20,0%) 13 (27,7%) 16,7 (35,5%) 24 (51,0%) 20,9 (44,5%) More than €3000 Count Expected count 1 (3,7%) 5,4 (20,0%) 10 (37,0%) 9,6 (35,6%) 16 (59,3%) 12 (44,4%) Table 6.11 Specialty goods: alternative delivery locations with opening hours given. Based on the outcomes of the tests performed, we conclude that income does not have a direct effect on the last-mile delivery option (home delivery versus picking up from an alternative location). Even when delivery fees are taken into account, the customer’s income does not show a significant relationship. This does not support the expectation of Punj (2012) that lower income groups value money-saving aspects (e.g. choosing the lower cost delivery option) more than higher income groups. The relationship between income and alternative delivery location that was found to be significant for convenience goods might be deterred as the physical store of brand or chain is in fact the same as a supermarket as the convenience good in this study was a bag of groceries. Income does show a significant relationship with alternative delivery locations for specialty goods, where higher incomes use the locker box more often compared to the other two groups. However, this relationship is not significant anymore when opening hours of alternative delivery locations are given. While the opening hours make the relationship between income and alternative delivery location insignificant, comparing table 6.8 with table 6.9 and table 6.10 with table 6.11 shows that specialty goods are more often chosen to be picked up from a locker box than convenience goods, regardless of opening hours. This contradicts with the expectation that specialty goods are more likely to be picked up from attended alternative delivery locations compared to shopping and specialty goods. However, as the decision to pick up the product from a locker box increases when opening hours are given, the notion of Song et al. (2009) and Thuiswinkel Markt Monitor (2012) is supported that opening hours influence the decision of the last-mile delivery option. Furthermore, because it differs per product type whether a relationship is significant, this supports the notions of Keeney (1999), Maldberger and Sester (2005), Newton (2001), and Thirumalai and Sinha (2005) that differences among product types regarding last-mile delivery options should be expected.

6.3 Household size and last-mile delivery options Test Independent

variable

Dependent variable Constant

p-value

N df 10(a) Household size Last-mile delivery option (home

Convenience 0,276 199 4 10(b) Household size Last-mile delivery option (home

Shopping 0,237 200 4 10(c) Household size Last-mile delivery option (home

Specialty 0,615 199 4 11(a) Household size Last-mile delivery option (home

delivery €4 versus pickup)

(27)

MSc Thesis Supply Chain Management – Mariëlla Joanne Settelaar Page 27 11(b) Household size Last-mile delivery option (home

Shopping 0,247 200 4 11(c) Household size Last-mile delivery option (home

Specialty 0,886 200 4 12(a) Household size Alternative delivery location Convenience 0,406 200 8 12(b) Household size Alternative delivery location Shopping 0,634 200 8 12(c) Household size Alternative delivery location Specialty 0,654 200 8 13(a) Household size Alternative with opening hours Convenience 0,075 200 8 13(b) Household size Alternative delivery location

with opening hours given

Shopping 0,212 200 8 13(c) Household size Alternative delivery location

with opening hours given

Specialty 0,101 200 8 Table 6.12 Outcomes chi-square tests with household size as independent variable. Among all product types, no significant relationship between household size and the last-mile delivery option has been found (table 6.12). Home delivery was chosen most frequently among all product types. However, when delivery fees are taken into account, an increase in picking up from an alternative location can be seen for all product types across all income groups, which can be seen in table 6.13 below (in this table, all product types have been combined). This observation is consistent with what can be seen in table 6.3.

Household size Home delivery and picking up same Home delivery €4 Home delivery Picking up from an

alternative location Home delivery Picking up from an alternative location 1 64 (79,0%) 17 (21,0%) 23 (28,4%) 58 (71,6%) 2 141 (75,1%) 41 (24,9%) 51 (25,0%) 153 (75,0%) 3 91 (84,3%) 17 (15,7%) 31 (28,7%) 77 (71,3%) 5 108 (84,1%) 21 (15,9%) 43 (32,6%) 89 (67,4%) 5 or more 68 (90,7%) 7 (9,3%) 26 (35,7%) 49 (64,3%) Table 6.13 Household size and delivery fees: all products combined.

(28)

MSc Thesis Supply Chain Management – Mariëlla Joanne Settelaar Page 28 6.4 Residential area and last-mile delivery options

Test Independent variable

Dependent variable Constant

p-value

N df 14(a) Residential area Last-mile delivery option (home

Convenience 0,054 199 2 14(b) Residential area Last-mile delivery option (home

Shopping 0,843 200 2 14(c) Residential area Last-mile delivery option (home

Specialty 0,691 199 2 15(a) Residential area Last-mile delivery option (home

Convenience 0,697 200 2 15(b) Residential area Last-mile delivery option (home

Shopping 0,232 200 2 15(c) Residential area Last-mile delivery option (home

Specialty 0,370 200 2 16(a) Residential area Alternative delivery location Convenience 0,085 200 4 16(b) Residential area Alternative delivery location Shopping 0,262 200 4 16(c) Residential area Alternative delivery location Specialty 0,046* 200 4 17(a) Residential area Alternative with opening hours Convenience 0,060 200 4 17(b) Residential area Alternative delivery location with

opening hours given

Shopping 0,014* 200 4 17(c) Residential area Alternative delivery location with

opening hours given

Specialty 0,005* 200 4 Table 6.14 Outcomes chi-square tests with residential area as independent variable. The choice of the last-mile delivery option (home delivery versus picking it up from an alternative delivery location) does not have a significant relationship with a customer’s residential area for any of the product types (table 6.14). Home delivery is chosen more often among all product types. However, when delivery fees are taken into account, an increase in picking up from an alternative location can be seen for all product types across all income groups (table 6.15). This observation is consistent with what can be seen in table 6.3.

Residential area Home delivery and picking up same Home delivery €4 Home delivery Picking up from an alternative location Home delivery Picking up from an alternative location Urban 299 (82,1%) 65 (17,9%) 100 (27,3%) 266 (72,7%) Suburban 45 (84,9%) 18 (15,1%) 14 (22,2%) 49 (77,8%) Rural 141 (82,5%) 30 (17,5%) 60 (35,1%) 111 (64,9%)

Table 6.15 Residential area and delivery fees: all product types combined.

(29)

MSc Thesis Supply Chain Management – Mariëlla Joanne Settelaar Page 29 in urban areas and people in rural areas prefer the physical store of the brand or chain. On the other hand, customers that live in the suburbs prefer to pick up the product from a locker box. Residential

area

Physical store of brand or chain

Supermarket Locker box

Urban Count Expected count 51 (41,8%) 53,7 (44,0%) 39 (32,0%) 33,6 (27,5%) 32 (26,2%) 34,8 (28,5%) Suburban Count Expected count 6 (28,6%) 9,2 (43,8%) 4 (19,0%) 5,8 (27,6%) 11 (52,4%) 6 (28,6%) Rural Count Expected count 31 (54,4%) 25,1 (44,0%) 12 (21,0%) 15,7 (27,6%) 14 (24,6%) 16,2 (28,4%) Table 6.16 Specialty goods: residential area and alternative delivery locations. When opening hours are given, shopping goods (table 6.17; p = 0,014) and specialty goods (table 6.18; p = 0,005) show a significant relationship between a customer’s residential area and last-mile delivery option.

Residential area Physical store of brand or chain

Urban Count Expected count 28 (22,9%) 29,9 (24,5%) 49 (40,2%) 42,1 (34,5%) 45 (36,9%) 50 (41,0%) Suburban Count Expected count 2 (9,5%) 5,1 (24,4%) 4 (19,1%) 7,2 (34,5%) 15 (71,4%) 8,6 (41,1%) Rural Count Expected count 19 (33,3%) 14 (24,5%) 16 (28,1%) 19,7 (34,5%) 22 (38,6%) 23,4 (41,0%) Table 6.17 Shopping goods: residential area and alternative delivery locations when

opening hours are given. Residential area Physical store of brand

or chain

Urban Count Expected count 21 (17,2%) 25,6 (21,0%) 51 (41,8) 42,7 (35,0%) 50 (41,0%) 53,7 (44,0%) Suburban Count Expected count 2 (9,5%) 4,4 (21,0%) 4 (19,1%) 7,4 (35,2%) 15 (71,4%) 9,2 (43,8%) Rural Count Expected count 19 (33,3%) 12 (21,0%) 15 (26,3%) 20 (35,0%) 23 (40,4%) 25,1 (44,0%) Table 6.18 Specialty goods: residential area and alternative delivery locations when

opening hours are given.

(30)

MSc Thesis Supply Chain Management – Mariëlla Joanne Settelaar Page 30 In current literature, not much has been found regarding this specific topic. It was mentioned that there is a difference in online shopping behavior (searching versus buying online) for people in urban and rural areas (Farag et al., 2006). However this research has not been linked to the delivery aspect of online shopping. Keeney (1999), Maldberger and Sester (2005), Newton (2001), and Thirumalai and Sinha (2005) expected differences among product types regarding last-mile delivery options. This analysis shows that when residential area is taken into account, it is also dependent per product type whether this relationship is significant which supports the prior expectations. Furthermore, while specialty goods already show a significant relationship when opening hours are not mentioned, both shopping and specialty goods show to have a significant relationship with the last-mile delivery option when opening hours are mentioned. This is consistent with what Song et al. (2009) and Thuiswinkel Markt Monitor (2012) state, namely that the influence of opening hours need to be taken into account when choosing for last-mile delivery options.

7 CONCLUSIONS AND RECOMMENDATIONS

The research question posed in this study was: “Can the different last-mile delivery options be matched with product types and customer characteristics by means of hypothesis testing?” First of all, we conclude that the product type does have an influence on the last-mile delivery option, which is consistent with Keeney (1999), Maldberger and Sester (2005), Newton (2001) and Thirumalai and Sinha (2005). This study shows that the product type has a relationship with the last mile delivery option, being it home delivery versus picking up from an alternative location or choosing between alternative delivery options. Consistent with Maldberger & Sester (2005), home delivery seems to be the preferred delivery mode among all product types. However, this is only the case when the delivery fees for both options are the same, which supports the expectation of Dennis (2011) that there is no evidence that customers are willing to pay extra for (faster) delivery.

When delivery fees are taken into account, so for charging €4 or €8 for home delivery, no significant relationship between the product type and last-mile delivery option exists anymore. Among all product types and among all tests, customers increasingly choose to pick up the product elsewhere. Therefore, delivery fees seem to influence the choice of last-mile delivery option more than the product type. Future research could study the price sensitiveness of customers among last-mile delivery options more in-depth. This can prove valuable for e-tailers as they can use this information to develop appropriate pricing schemes for the last-mile delivery options they are offering. If it is known what the customer wants, the price difference among the last-mile delivery options can ‘steer the customer to the right channel.’

BRIDGING THE LAST-MILE OF E-COMMERCE

Table of contents