Reducing Product Returns in Internet Retailing through Customer Differentiation

(1)

(2)

2

Preface

In your hands, you hold the result of the past half year of doing research for the purpose of my Master Thesis at the University of Groningen. This thesis is also the final piece of work I have to deliver before I graduate on the Master, Business Administration specialization Operations and Supply Chains.

I started this Master past September, now almost a year ago and found that studying can actually be fun as long as you are doing something that you really like. This certainly was the case for me and this also helped me to get much higher grades than I was used to during my Bachelor period. During this year, I got the opportunity to learn from some great minds of the Operations Department, but also from some fellow students. It was really nice to be part of such a small tightly connected group and I look forward to see where all your careers will be heading in the future.

During the process of doing research and writing my thesis, I got help from a couple of people, whom I would really like to thank. I would foremost like to thank prof. dr. Kees Jan Roodbergen, who gave me excellent comments during the whole process of writing my thesis.

Whenever I thought I was stuck, he always gave me new insights and ways of looking at things.

I would also like to thank dr. Tudor Bodea, who gave me useful feedback during the initial stages of my research and urged me to question some of the assumptions that I had made. This urged me to build better arguments for my hypotheses, which made my story much more stronger.

During the phase in which I had to find participating companies for my study, prof. dr Kees Jan Roodbergen together with Tom Steffens, introduced me and my research project to a couple of companies active in internet retailing. I would like to thank Tom Steffens for his assistance in this process.

Of the companies that eventually participated, Thijs Bregman, who was my contact at the first company, needs to be thanked. The visit to your company was really interesting and I appreciated the extent to which you were willing to think along with me on my research. I also appreciated that you were keen to supply me with follow up information after I received the first data.

Further, my thanks go out to Jan-Noud Hutten, my contact at the second company, you were also really helpful in thinking along with my project. I also really appreciated the tour through the company’s warehouse facility; it was really interesting to get a real feel for type of company were I conducted my research.

During the whole process of writing my thesis I got help from a couple of close friends and family members in giving feedback on small parts or ideas for my thesis. I would like to thank all of them, but I specifically want to thank Arend-Jan Wonink for proofreading my thesis at the end of the process.

- Sylvester J. van Dijk, Groningen

(4)

4

Reducing Product Returns in Internet Retailing through Customer Differentiation

Sylvester J. van Dijk^*1

Supervisor: Prof. Dr. Kees Jan Roodbergen Second assessor: Dr. Tudor D. Bodea

July 2012

University of Groningen Faculty of Economics and Business

Master Thesis – BA specialization Operations & Supply Chains

Keywords: E-commerce, Internet Retailing, Marketing-Operations Interface, Product Returns, Return Shipments

Abstract

Reducing the amount of products returned in internet retailing can result in large gains due to high transportation, warehousing, and obsolescence costs. Prior studies have indicated that returns concentrate on a narrow set of products. This study tries to give internet retailers a way to identify products that are returned because of a misrepresentation on their website or a common defect. It is hypothesized that customers with different characteristics vary in their effectiveness to select the right product online. The most effective decision makers mainly return products for reasons beyond their own control such as a defect, instead of because they made an error. These customers are identified through a chi-squared test that compares the amount of products returned by different customer groups at two internet retailers. The results show significant differences between customers with a different order frequency, gender, and age. Males and customer of ages 33 through 44 were found to return significantly less for both e-retailers, while this group was also identified in literature as the most frequent online shopping group. The results can be used to reduce returns by two means. Products returned because of a misrepresentation on the website or a defect can be identified by looking at what effective decision makers do return; adaptations can then be made accordingly. Secondly, customers returning often should be aided; they may for instance be supplied with extra product information to make sure they order the right product.

1. Introduction

Internet retailers (e-retailers) often experience high products returns reaching 45% in some cases (Tarn et al., 2003; Cooke, 2000). According to Gartner Research, returns can cost in excess of $20 per return in warehousing, transportation, and obsolescence costs. Reducing the

*Address: Wassenberghstraat 25, 9718 LH Groningen, Email address: S.J.van.Dijk.1@student.rug.nl Student number:

1633805

(5)

5

amount of products returned can therefore result in significant financial gains for these companies.

Prior research has shown that e-retailers have on average a broader assortment compared to traditional stores; this offers them benefits such as an increase in customer surplus (Brynjolfsson et al., 2003). It is argued however that e-retailers therefore have to manage a broad set of returns for different products, but research has shown that this is not the case, as returns tend to be concentrated on a specific set of products (Rabinovich et al., 2011). This opens up opportunities, because e-retailers mainly have to focus their attention to these products if they want to decrease returns. In their study on products returns, Rabinovich et al.

(2011) investigated how product returns can be differentiated on product characteristics that influence the risk that customers incur while buying online. They opt for the possibility to reduce returns by decreasing the assortment, removing products with characteristics that are returned often. However product variety has been shown to increase a firm’s demand and the internet made it much more cost effective to offer a high variety of products (Ramdas, 2003).

There is a gap in the literature as studies have not focused on how differentiating between customer groups can give a better insights into return reasons. Consider for instance a customer buying and returning a golden watch online, because the customer accidentally ordered the wrong type. This is very different from a situation in which the customer returns the watch because the website specified it as a mechanical watch, while it is actually a quartz watch. In the first scenario the return is beyond the influence of the e-retailer, but in the second scenario the e-retailer can do something about it and change the specification on the website. The possibility that more experienced customers in online shopping can make better choices in product selection (Xiao & Benbasat, 2007), is one of the possible options by which returns may be reduced. It can for instance be argued that e-retailers should pay closer attention when an experienced internet customer returns a product as compared to an inexperienced customer. This is because those experienced online shoppers will be the group that mainly sends products back, because something is wrong with the product, or how it is presented and not because the customer made a wrong decision in the online shopping process.

This study provides a closer insight into the relationships between customer characteristics and product returns, so e-retailers can identify products prone to be returned. The objective of this study therefore is: “Identify customer groups that make the most accurate online purchasing decisions.” This is done by means of constructing a set of hypotheses that discriminate on customer characteristics. The difference in return behaviour amongst groups that differ on those characteristics is then tested through statistical analysis of datasets of two e-retailers. This datasets will consist of information about sales and returns of 2011, tied to individual customer IDs and their relevant demographic information.

The remainder of this paper is structured as follows: in the next section the research questions are defined, in the third section the theoretical background is discussed, in the fourth section the hypotheses under investigation are introduced, in the fifth section the method for this research is explained, the sixth section states the results, in the seventh section the results are discussed, and in the final section the conclusions of this study are presented.

(6)

6 2. Research questions

The following set of research questions are developed to define the scope of the research, determine the relevant variables and drive the research. They will be answered throughout the remainder of this paper.

 What has so far been written in academic literature on product returns?

 What product characteristics drive the return of products?

 What customer characteristics can hypothetically make a difference in the effectiveness to make choices in online shopping?

 What data on these characteristics can realistically be acquired from e-retailers?

 What is the best statistical method to test the hypotheses with the acquired data?

 How can e-retailers benefit from knowing what kind of customers are more and less effective in making product selections online?

 In what ways can this eventually result in less products being returned overall?

3. Theoretical background

3.1. Losses in the product return flow

The return process of firms constitute of many losses of which firms are often at least partially unaware. It is therefore useful to zoom in on those costs and their sources to clarify the potential gains that can be acquired by reducing the amount of products returned.

Returned products account for large asset streams for many firms. However, a large proportion of this value is lost in the reverse supply chain. The study of Blackburn et al.

(2004), found that nearly half the asset value is lost in the return stream. These losses mainly fall into two categories; one share is because the products are downgraded to a lower valued product. This happens when a product must be remanufactured, salvaged for parts, or scrapped. The other share is devalued due to the passage of time as it moves to disposition, through deterioration of value with time and downgrading due to obsolescence.

Reverse supply chains thereby make supply chain management much more complex for companies. It adds to coordination issues such as complex trade-offs in supply chain objectives and conflict of interests among participants. A closed-loop supply chain is a supply chain in which return processes are intended to capture extra value (Guide et al., 2003). The intrinsic complexity is often a reason for companies not to invest in the design of their closed- loop supply chain leading to inefficiencies and non-responsiveness (Krikke et al., 2004).

These inefficiencies in the return supply chain make it more demanding for companies to decrease the amount of products returned.

3.2. Returns and product characteristics

The study in hand focusses on differences in customer groups that vary in the amount of products returned. It is however useful to acknowledge product characteristics that drive the amounts of products that are returned. Extensive research has been done on this subject and the amount of returns is often depended on specific product characteristics.

Most research focuses on price paid. First, Anderson et al. (2008) provide empirical evidence that return rates increase with the price that is paid. Secondly, Rabinovich et al.

(7)

7

(2011) showed that high-priced goods exhibit a high concentration of sales across stock- keeping-units (SKUs), but a wider distribution of returns. This indicates that on the internet, customers when purchasing high-priced items, purchase mainly a small set of popular items, but the high price will also entice them to return the item if it does not match their expectations. Thirdly, it was found that products on sale, where returned less frequently than products that are not on sale (Petersen & Kumar, 2009).

Other characteristics have also been studied. Petersen and Kumar (2009) found that products bought as gifts were less likely to be returned. They also found that products bought during the holidays were more likely to be returned compared to products bought throughout the rest of the year, indicating seasonality in the amount of returns.

The effect of size (here weight and volume) has also been studied. Rabinovich et al.

(2011) found that on the internet for large products, sales are concentrated on a small set of SKUs. Their results however failed to show any difference based of large versus small products in the amount of returns.

Another finding of Rabinovich et al. (2011) was that products that have long been available at the website of the e-retailer distributed a wider distribution of sales and a higher concentration of returns compared to newly available products. This indicates that for these long standing products a portion of these may have some sort of defect or may be misrepresented on the website.

3.3. Product returns on the internet

It is important to acknowledge that product returns on the internet differ from traditional product returns in some fundamental ways. This is due to such factors as breadth of assortment, low searching costs, and the time lag that occurs between ordering an item and receiving it. It is important to understand the implications and constraints these factors pose on the e-retailers’ return policies.

The internet has greatly expanded the consumer’s ability to appropriately select products across a wide variety of products. Information economics theory suggests that the internet’s interoperability and open standards for data transfer decrease search costs that customers incur (Rabinovich, 2007; Rabinovich et al., 2011). In the online shopping process, customers do not have to make sacrifices in time and money that are required in brick-and-mortar stores to overcome location boundaries and to make purchasing decisions. This involves sometimes investing a day to find the right product (Bakos, 1997).

In e-retailing the customer must await product delivery after placing the order; therefore product returns are often the result of explicit shortfalls of SKUs that become apparent after receipt (Forbes et al., 2005). These shortfalls occur more for a small set of SKUs as returns on the internet concentrate on an as compared to sales narrow set of SKUs, which implies that there are certain SKUs inherently prone to be returned (Rabinovich et al., 2011).

After a customer has made a purchase the customer needs to wait for the product to be delivered in order to be able to make a physical inspection. This would traditionally already have occurred before the purchase. Wood (2001) calls this a two-stage decision process, the first being the decision of selecting the product online and the second decision being whether to keep the product after physical inspection upon receipt.

(8)

8

In this situation it was found that more lenient return policies result in a greater probability of products being returned compared to restricted return policies. Lenient return policies, however, also increase order rates and create higher quality perception even after physical inspection (Wood, 2001; Bonifield et al., 2010). Also, the option to return an item, as opposed to not being able to return the item at all, leads to significant increases in demand (Anderson et al., 2009). Because of this important trade-off, e-retailers will in most cases not benefit from decreasing return amounts through restricted return policies.

Further, earlier studies on product returns investigated how the manner at which e-retailers deal with returning products affects their customers’ loyalty intentions. It was found that the perceived value of the returns process and return satisfaction directly affected the customers’

loyalty intentions (Mollenkopf et al., 2007). This further acknowledges the importance of lenient return processes in e-retailing.

The two stage decision process does make product returns in e-retailing more common than in brick-and-mortar stores; however the fact that returns concentrate on a narrow set of SKUs proves that improvements can be made (Rabinovich et al., 2011).

4. Hypotheses formulation

4.1. The effect of order frequency on product returns

It can be expected that as customers have more experience in online shopping they will be better suited in making accurate decisions in this environment. Through the process of buying online, customers gain knowledge that they can use to make effective product selections (Xiao & Benbasat, 2007). Previous research found that when customers purchase unfamiliar products through a new channel such as the internet, they are more likely to return the product (Petersen & Kumar, 2009). Experienced online customers base their product choice on attribute information; while customers with less experience have been found to look more at less complete general summary of information. This is due to their lack in ability to process attribute information of products as efficiently (Brucks, 1985). Based on these arguments one would suggest that as customers buy more often online, they gain experience and therefore have less need to return products. This leads to the following hypothesis.

Hypothesis 1. Customers with a higher order frequency return fewer products than first time customers.

4.2. The effect of age on product returns

On age there is some debate on what customer groups shop most often online. Some studies have found online shoppers to be older than non-online shoppers (Donthu & Carcia, 1999; Tan, 1999). The study of Swinyard and Smith (2003), in contrast found younger customers to shop more online. However, a more recent publication indicates that it is the older generation that makes up the majority of online shoppers, while younger generations use the internet to a greater extent for social activities. They define older customers here as customers of age 33 and above (Jones & Fox, 2009). This suggest that older customers have more experience in online shopping than younger customers have and may therefore be more effective decision makers in the process of shopping online.

(9)

9

Within e-retailing, the older customers become, the more they prefer detailed product information. This increased preference starts at around the age of 30. During online shopping these customers have the greatest interest in product specifications, usage instructions, warranty information, product histories, and countries of origin (Burke, 2002). This preference for detailed information suggests that older customers think more carefully before actually ordering a product, which may decrease their need to return products.

These arguments hint into the direction that older customers are better able to make effective choices in product selection on the internet; therefore the hypothesis below is defined.

Hypothesis 2. Younger customers return products less frequently than older customers.

Continuing on the differentiation made by Jones and Fox (2009) in an effort to find smaller groups which differ in a greater extent to other age groups, a second age based hypothesis is developed. In their study they differentiate on different generations of internet- users and found users of ages 33 through 44, as the most experienced group with 80% using the internet as a shopping channel. To investigate how this group relates to other internet users on product return rates, the hypothesis below is stated.

Hypothesis 3. Customers of ages 33 through 44 return significantly less products than customers of other age groups.

4.3. The effect of gender on product returns

Customers of a different gender may also differ in their return behaviour. Some profound differences in male and female behaviour and attitudes in e-commerce have been found.

Rodgers and Harris (2003) found that females are less satisfied with online shopping than males. Furthermore, females also had a more negative attitude towards e-commerce than males. Other research has shown that female online shoppers find it far more important than males that products ordered have a return label and that the postage is subtracted from the refund (Ulbrich et al., 2011). Next to that, females also express a greater need for alternative means of returning products (Burke, 2002). These preferences indicate that females are more likely to return a product than males are. Besides that, just as with older customers, males find it significantly more important than females that the website contains accurate and detailed product information (Ulbrich et al., 2011). This gives an indication that they will use this information to a greater extent decreasing the probability that they make faulty decisions and therefore have to return more products. Furthermore, females have been found to find online shopping less practical and convenient than males (Rodgers & Harris, 2003).

These arguments indicate that female customers make product decisions online, using less information than their male counterparts and with the assumption that they can return the product anyhow. Therefore they may be more prone to make mistakes while shopping online and thus return products relatively more often. These arguments bring forth the following hypothesis.

Hypothesis 4. In e-retailing male customers return products less frequently than female customers.

(10)

10 5. Method

5.1. Data collection

The hypotheses constructed in the previous section were tested on the basis of transactional data of product purchases and returns tight to individual customer Ids, gender, and birthdates. Data of two e-retailers was used. This was done to be able to compare findings for different kinds of e-retailers. Therefore data was collected from two companies that are quite different from each other.

The data was therefore also treated separately for each e-retailer. Doing so is also consistent with prior empirical research on this subject (Olson & Boyer, 2003; Rabinovich et al., 2011). This will increase the reliability of the results, because for a single e-retailer all exogenous parameters that might affect the purchasing and return behaviour of customers are identical. These parameters are the time customers are given to reflect on their purchase, whether customers incur costs when they return a product, and the way products are displayed to customers in the online shopping environment (Esper et al., 2003).

The transactional data that was acquired was data of all 12 months of the year 2011, for both companies. This period was selected to account for any demand and return variation attributable to seasonal fluctuations.

5.2. Sample e-retailers

The first e-retailer is a company called Fonq. They sell furniture, kitchen appliances, toys, wellness products, and gardening products. For this e-retailer all data on product sales and returns of 2011 was acquired. At this e-retailer, there are slightly more female customers, constituting 54% of the population.. The average customer’s age is around 44 and on average 4.6% of the products are returned.

The second company exclusively sells ink and will for confidentiality reasons in this study be given the pseudonym Ink Retailer. Most customers will only need ink for the particular brand and type of printer they got. The results of this study will make it clear if certain customers are better at selecting the right type. For this company a sample of 15.000 orders was used. Their customer base consists primarily of male customers accounting for 72%. The average age of their customers is 48. They incur quite a small product return rate of just 1.4%.

5.3. Data transformation

For Fonq, there are a broad set of return reasons specified in the dataset. Amongst them are also returns for repair, pick errors and a failure to deliver by the transporter, and random defects. The results are filtered for these return reasons, because it is not the objective of this study to find out what group sends most products back for repair. The other set of reasons are filtered because they are equally likely to occur for different groups, this has also been verified in the data.

In case of the data for the Ink Retailer, returns are categorized into a few groups. There are general returns with no reason, returns due to a defect or cartridge leakage, not delivered or accepted at the door, and returns due a failure by the customer to retrieve the product at a pickup point. Results are computed only for returns with no reason, because defects and

(11)

11

delivery failures are often beyond the control of the customer and each customer is equally likely to receive such a defective product, this has again been verified in the dataset.

Further, the data needs to be transformed to make separate customer groups. For the order frequency hypothesis, the data is divided into a group which only ordered once in the year 2011, and a group that ordered multiple times. In case of the gender hypothesis groups were naturally divided into gender groups. For the first age hypothesis groups were split at the age of 33. So everyone at or above the age of 33 was put in the old group and everyone below 33 in the young group. For the second age hypothesis groups of age 33-44 were compared with all customers of an age outside this group to see if larger differences can be found. This split was based on the study of Jones and Fox (2009) as explained in the age hypotheses section this study found that starting at this age people use the internet significantly more for online shopping.

5.4. Data analysis

The complete data transformation resulted in contingency tables such as the one put as an example below.

For the statistical analysis, this research thus deals with binomial data as the customers either do or do not return an item. The two groups that are compared for each hypothesis consists of different individual e.g. customer cannot both be male or female or frequent and less frequent buyers at the same time in this study. Therefore an unpaired test needs to be selected. The two most appropriate statistical tests for binomial unpaired data are Fisher’s exact test and Pearson’s chi-squared test. This study deals with large samples and possibly well-balanced tables, therefore Pearson’s chi-squared test is most appropriate. This is also consistent with prior research in which chi-squared tests are also used to investigate differences in age and gender (Chung-Herrera et al., 2010; Yuksel, 2004).

So a chi-squared goodness of fit analysis was conducted. In this analysis the theoretical distribution with which the sample is compared, is that where the two opposing groups of each hypothesis are just as likely to return a product, which are the null hypotheses. The values that go with this are the expected counts in the example table above. Based on this analysis the four research hypotheses are either rejected or not rejected.

Table 1 example gender * returned cross tabulation

Not returned Returned Total

Female Count

Expected Count Male Count

Expected Count

Total Count

Expected Count

(12)

12

The formula that was used to determine the value of chi squared ( ) is stated below. In which O (observed) is the collected data and both Es (expected) are the values that come with the null hypothesis in which no difference would occur between the different groups under investigation.

∑

The degree(s) of freedom (DF) is equal to the number of levels (k) of the categorical variable minus 1 so: . In this case k is 2 for each of the comparisons so:

. The significance level that is used to either reject or confirm the three hypotheses is the recommended 0.05. With this value of DF and a significance level of 0.05, the critical value is 3.841. This means that each null hypothesis is rejected when the value of is larger than this value.

6. Results

6.1. The order frequency hypothesis

The order frequency hypothesis will only be tested for Fonq, because of data related reasons. It appeared that on the contrary to what the hypothesis suggested, customers with a higher order frequency return products far more often. Customers that shopped for the first time at this e-retailer returned 1.62% of the products they ordered, while returns coming from customers with an order frequency higher than 1 generated 2.53% returns, which means 55.7% more often.

Table 2 shows the value 486.772 of , which is larger than the critical value of 3.841, meaning that customers with an order frequency higher than 1 return significantly more products. The asymptotic significance value in the table above is also extremely low at 0.000 at three decimals, indicating that with a much higher significance level for instance 0.001 the result would still be significant.

Table 2 Chi-squared test order frequency Fonq

Value DF Asymp. Sig. (2-sided) Exact Sig. (2-sided) Exact Sig. (1-sided)

Pearson Chi-Squared 486.772a 1 0.000

Continuity Correction^b 486.275 1 0.000

Likelihood Ratio 455.427 1 0.000

Fisher's Exact Test 0.000 0.000

N of Valid Cases 571696

(13)

13 6.2. The age hypotheses

In the paragraphs below the results for the age hypotheses are discussed separately for each e-retailer. If the first age hypothesis holds, customers over 32 are less likely to return products than younger customers. If the second age hypothesis holds customers of ages 33 through 44 return significantly less than customers in other age groups.

For Fonq young customers return 2.43%, while old customers return 2.26%, so young customers return products 7.73% more often.

The value of is 8.693, this is larger than the critical value, so for Fonq younger customers do indeed return significantly less.

For the second age related hypothesis there are three age groups, customers under 33 return 3.43% and the oldest group of customers over 44 return 2.37%. So these two groups have about equal return rates. These groups combined return 2.39% of their products. The 33- 44 group in contrast returns only 2.10%. So the combined group returns products 13.68%

more often; so there is a larger difference than for the previous hypothesis. The statistical results are presented below.

For the second age related hypothesis at Fonq the results are significant with a value of 12.416. Customers of ages 33-44 return significantly less products than customers of other ages.

Table 3 Chi-squared test age spilt 33 for Fonq

Table 4 Chi-squared test age 33-44 versus other for Fonq

(14)

14

For the first age related hypothesis at the Ink Retailer the opposite was found, here customer over 32 returned 25.09% less than their opposing group. However, this difference was not significant for this sample. This can be seen in the table below as the value is .448 which is lower than the critical value.

For the second age hypotheses at the Ink Retailer, customers over 44 returned 1.49% and customers under 33 returned 1.18%. When they are combined they return 1.43%. The middle group of customers of ages 33-44 return 1.03%. This means that the combined group of young and old customers return 38.32% more than the 33-44 group.

In the table above it can be seen that the results are significant and that indeed also for the Ink Retailer customers of ages 33 through 44 returns significantly less than customers of other ages.

However, it needs to be noted that the groups under age 33 and above 44 varied much more in their return amounts than was the case at Fonq, as customer under 33 returned much fewer products. This also makes clear why in the sample for the first age hypothesis the opposite was found than for Fonq. At the Ink Retailer customers over 44 returned relatively more and under 33 relatively less than for Fonq. This suggests that a different age split, in order to identify customer groups that return fewer products is more appropriate for the Ink Retailer. Additional analysis shows that when ages are split into two groups at age 44, the older group returns 0.79% versus 0.45% for the younger group, this is a difference of 75.29%, which is significant with a value of 5.485.

Table 5 Chi-squared test age split 33 for Ink Retailer

Table 6 Chi-squared test age 33-44 versus other for Ink Retailer

(15)

15 6.3.The gender hypothesis

In terms of gender, the research hypothesis states that females are more likely to return a product than males. For Fonq this is certainly the case. Males return 1.72%, while females return 2.06%, which is 20.19% more.

The value of is 84.019, which is larger than the critical value of 3.841 and thus hypothesis 2 is not rejected for Fonq. Male customers return products relatively less frequently than female customers.

For the Ink Retailer, in terms of percentages over the year 2011, 0.70% of males returned products versus 0.75% for females. This means in the sample females return products 6.78%

more often than males. The results of the chi-squared test are depicted below.

The value of of 0.095 is below the critical value of 3.841. So for this retailer the null hypothesis cannot be rejected in favour of the research hypothesis. The percentage difference in the sample in combination with the results from Fonq only gives an indication that the relation for this e-retailer may be the same, but this cannot be determined at a significant level.

6.4. Summary of the results

For the hypotheses this means that the results from the previous subsections have the following consequences:

 Hypothesis 1 is rejected; empirical results show that customers with a higher order frequency do not send fewer products return than first time customers, the results instead show that it is the other way around at a significant level.

 Hypothesis 2 is not rejected; At Fonq customers under age 33 return significantly more products. For the Ink Retailer no significant difference was detected at this age split.

Table 7 Chi-squared test gender for Fonq

Table 8 Chi-squared test gender for Ink Retailer

(16)

16

 Hypothesis 3 is not rejected; for both e-retailers customers of ages 33 through 44 returned significantly less products than the combined groups of customers of other ages.

 Hypothesis 4 is not rejected; on the internet females return significantly more products than males. However, this relation was not found at a significant level for the Ink Retailer.

Another interesting additional result is that at the Ink Retailer customers over 44 return significantly more products that customers at or under this age. This difference is also quite large in comparison to other differences. This distinction was not found at Fonq, indicating to a difference in product return behaviour based on age for the different e-retailers.

7. Discussion

7.1. The order frequency difference

The results show that customers buying more often online tend to return relatively more in contrast to what the literature arguments for the first hypothesis suggested. The results indicated a huge difference between customers ordering once versus customers ordering multiple times. The first explanation for this difference is that customers that make an error in selecting a product online, will often place another new order to get the desired product. This will increase their order frequency, which causes them to fall into the high order frequency group.

However, the fact that the difference between the two groups is so large receives additional explanation. Petersen and Kumar (2009) suggested that customers shopping in a new channel tend to return more often due to ineffective product selections. The results of the current study suggest that the internet may be an exception to this, as customers may need time to get used to the return process of this channel. Once customers gain trust in the effectiveness of the return process they may also order products even if they are not entirely sure if the product they will order is the right one.

Literature also points to another explanation. Initially customers are more likely to purchase low-priced items due to a high perceived financial risk. Once they are more accustomed to the new channel or e-retailer, their perceived risk decreases. This makes them less reluctant to buy higher-priced items (McCorkle, 1990). This can increase return rates at higher order frequencies, because multiple studies found high-priced items to be returned more often (Anderson et al., 2009; Rabinovich et al., 2011).

7.2. The age difference

On age there are some profound differences between the two e-retailers. For Fonq the results clearly show that the age group 33-44 returned the least, customers under and above that age returned an about equal percentage of the products. For the Ink Retailer this was different, 33-44 group returned less products, but it appeared that customer over 44 returned far more than customers under 33. The additional analyses showed that at the Ink Retailer customers over 44 returned significantly more products than customers below this age.

This means that for Fonq the group that made the fewest errors in online product selection is the group identified in the literature as having most shopping experience with ages 33 through 44 (Jones & Fox, 2009). For the Ink Retailer this group is somewhat broader and also includes customers below 33.

(17)

17

This indicates that indeed, increased preference for detailed product specifications in virtual stores helps customers over 32 to make fewer errors in product selection (Burke, 2002). Customers above 44, however, have an even higher preference for this information, but were still found to return more often especially for the Ink Retailer. This may be, because they are less accustomed to online shopping. They shop much less online than adults under 44 (Jones & Fox, 2009).

Still the question remains why at the Ink Retailer customers above age 44 returned far more than customers below this age. Also, customers under age 33 returned about equal to the 33-44 group at this e-retailer. At Fonq, in contrast, customers under age 33 returned just as often as customers over 44. So, customers under age 33 appeared to be better decision makers at the Ink Retailer while customers above 44 appeared to have more problems in selecting the right product.

An explanation may lay in the type of products sold by the Ink Retailer. These are products that are somewhat technical in nature and computer-related. The 45+ group constitutes of many customers, including seniors, who did not grow up with these types of products. This may be the reason why specifically this age group is less effective at selecting the right type of product at this e-retailer. This can also explain why at this e-retailer customer under age 33 return so few products. Customer under age 33 did in contrast to the previous group grow up with computers and computer-related products; they may therefore make fewer errors in selecting products in this category.

7.3.The gender difference

There is a significant difference between males and females in the amount of products returned for one of the e-retailers. This can be explained through the arguments made for the hypothesis, suggesting that the preference of males for detailed product information and females’ preference for alternative and lenient return policies, do make males return fewer products (Ulbrich et al., 2011; Burke, 2002).

At the Ink Retailer this difference was not detected at a significant level. There is only an indication that this difference does exist at this e-retailer. This is because in the sample females do return slightly more than males. However, if this difference does in fact exist, it is expected that this difference is smaller than for Fonq.

The difference between the e-retailers can be caused by a difference in return policy leniency. As specified earlier, females prefer more lenient return policies than males (Ulbrich et al., 2011; Burke, 2002). Between the two e-retailers there are some differences in leniency, such as alternative means of returning a product and postal costs. This may make females who would normally return relatively often more reluctant to shop at the Ink Retailer. This may also be one of the reasons why 72% of its customers are male.

8. Conclusions

8.1. Summary

This paper investigated what customer groups make the most effective decisions in the online environment so that e-retailers can further inspect the products that these groups return and make necessary changes in order to reduce returns. Based on literature, relations were

(18)

18

hypothesized, which suggested differences in decision making accuracy between customers with a different order frequency, age, and gender. To test these potential relationships return behaviour of these groups was tested at two e-retailers. A chi-squared test was conducted to identify any significant differences in return amounts for these groups. The results indicated that indeed there are differences. For Fonq there was a difference for order frequency as more returns came from customers that ordered more than once. Furthermore, there was a relationship for gender as females returned significantly more often than males at this e- retailer, also different age groups varied significantly in the amount of products they returned.

In the first section of this paper, the objective “Identify customer groups that make the most accurate online purchasing decisions” was stated. This study shows that males are more effective decision makers in the online environment as females return significantly more at Fonq for reasons within the control of customers such as: “wrong product ordered”, or

“product ordered double”. The strongest difference between customer groups was based on age, here significant differences were found for both e-retailers. It appeared that the most effective decision makers for Fonq were customers in the age group 33-44, while at the Ink Retailer these were customers under age 45. This means that for e-retailers selling a broad variety of products the most effective decision makers will be customers of ages 33 through 44. While the results from the Ink Retailer give an indication that e-retailers selling products with which a specific age group is more accustomed, other age groups may be the better decision makers. Order frequency was not found to be a good indicator of decision making accuracy, because most likely frequent buyers also buy more expensive products, which are returned more often and customer that make an error in product selection often re-order the product, which increases their order frequency.

As shown in the results, the difference between the high and low returning group can be extremely large (up to 75 per cent for the Ink Retailer), so differentiating between customer groups is certainly expected to pay off.

8.2. Managerial implications

The results of this study are relevant for mangers of e-retailing companies in various ways. These are:

 Males and customers of ages 33 through 44 make the fewest errors during product selection.

 The age group that is most effective at decision making can vary somewhat per e-retailer.

 The percentage difference in product returns between customer groups can be very large.

 E-retailers should focus on improving products returned by low returning groups.

 Customer groups returning products often should receive extra help in online shopping.

In terms of gender, e-retailers should focus on what products are returned by males. By doing this they might detect production errors or errors in the representation of products on their website. By adapting these accordingly or possibly completely removing these products, returns are expected to decline for all customers. This approach is in accordance with the study of Byrne (2007), which states that product continuation decisions should be made from the consumers’ perspective instead of exclusively looking at the amount sold.

(19)

19

E-retailers selling a broad variety of products like Fonq should look into products returned by the age group 33-44. The results from the Ink Retailer show that this distinction may somewhat differ per e-retailer. It could be that for products were young customers are very familiar with, these younger customers may also be very effective decision makers. This indicates that e-retailer can benefit from assessing their customers’ knowledge on the specific products that they are selling to determine the right age group to look at.

The other way around, e-retailers can use the results of this study to aide customers that return excessive amounts in online shopping so that they return less. The internet offers a wide variety of opportunities for this; they can for instance supply these customers with additional product information or give them a warning when they order a product that does not match with their profile. For example, some returns are a result of a double order or a product received twice. This happens when a customer accidently presses twice on a product and then goes to the register, being unaware that the product is now twice in its basket. If this customer is part of the high returning group, the customer could receive a warning when trying to order the same product twice.

8.3. Limitations and future research

First, it needs to be noted that the sample of 15.000 orders for the Ink Retailer was quite small considering their low return rate. Because the largest share of its customers are also relatively old, customers under age 33 constituted of only 14% of the population. Therefore this group contained only 22 returns, which is quite small for a statistical analyses. However, for the hypothesis that goes with this, the same relation was found for the other e-retailer in which all orders of 2011 were used.

Second, for order frequency there are some important limitations. As specified earlier, customers that make an error in online product selection often need to re-order to get the desired product, causing them to fall into the high order frequency group. This makes it quite elaborate to distinguish between customers that regularly order new products and customer that have to re-order often because they had to return a product. Future studies could try to make a distinction by applying a filter that excludes customers re-ordering similar items.

However, the practical implications of doing this will be limited, because it is also hard for e- retailers to distinguish between these customer groups.

Third, the results of the Ink Retailer gives an indication that different age groups are most effective for different kinds of products. Future studies could investigate what product characteristics determines this. This could aid in better finding the most effective decision makers, especially for e-retailer that only sell products from one specific category like the Ink Retailer does.

Further, to the best of the author’s knowledge, this is the first study to investigate return behaviour based on specific customer characteristics in e-retailing. Therefore this study had a somewhat exploratory nature. The customer characteristics were chosen because of convenience and because they appeared most likely to generate positive results. Future studies could therefore further investigate other customer characteristics that may determine online decision making accuracy.

(20)

20 References

Anderson, E.T., Hansen, K. & Simester, D., 2009. The Option Value of Returns: Theory and Emperical Evidence. Marketing Science, 28(3), pp.1-19.

Bakos, J.Y., 1997. Reducing buyer search costs: implications for electronic marketplaces.

Management Science, 43(12), pp.1676-92.

Blackburn, J.D., Guide, V.D.R..J., Souza, G.C. & Van Wassenhove, L.N., 2004. Reverse Supply Chains for Commercial Returns. California Managmenet Review, 46(2), pp.6-22.

Bonifield, C., Cole, C. & Schultz, R.L., 2010. Product Returns on the Internet: A Case of Mixed Signals? Journal of Business Research, 63(9-10), p.1058–1065.

Brucks, M., 1985. The Effects of Product Class Knowledge on Information Search Behavior. Journal of Consumer Research, 12(1), pp.1-16.

Brynjolfsson, E., Hu, Y. & Smith, M.D., 2003. Consumer surplus in the digital economy:

estimating the value of increased product variety at online booksellers. Management Science, 49(11), pp.1580-96.

Burke, R.R., 2002. Technology and the Customer. Journal of the Academy of Marketing Science, 30(4), pp.411-32.

Burke, R.R., 2002. Technology and the Customer Interface: What Consumers Want in the Physical and Virtual Store. Journal of the Academy of Marketing Science, 30(4), pp.411-32.

Byrne, B., 2007. Finally, a strategic way to cut unnecessary SKUs. Strategy &

Leadership, 35(1), pp.30-36.

Cambell, J.R., 1991. The roots of gender inequity in technical areas. Journal of Research in Science Teaching, 28(3), pp.251-64.

Chung-Herrera, B.G., Gonzalez, G.R. & Hoffman, K.D., 2010. When demographic differences exist: an analysis of service failure and recovery among diverse participants.

Journal of Services Marketing, 24(2), pp.128-41.

Chung-Herrera, B.G., Gonzalez, G.R. & Hoffman, K.D., 2010. When demographic differences exist: an analysis of service failure and recovery among diverse participants.

Journal of Services Marketing, 24(2), pp.128-41.

Cooke, J.A., 2000. The physical challenges of the virtual sale. Logistics Management and Distribution Report, 39(10), pp.67-73.

Donthu, N. & Carcia, A., 1999. The Internet shopper. Journal of Advertising Research, 39(3), pp.52-58.

(21)

21

Esper, T.L., Jensen, T.D., Turnipseed, F.L. & Burton, S., 2003. The last mile: an examination of effects of online retail delivery strategies on consumers. Journal of Business Logistics, 24(2), pp.177-203.

Forbes, L.P., Kelley, S.W. & Hoffman, K.D., 2005. Typologies of e-commerce retail failures and recovery strategies. Journal of Services Marketing, 19(5), pp.280-92.

Guide, V.D.R.J., Harrison, T.P. & Van Wassenhove, L.N., 2003. The Challenge of Closed-Loop Supply Chains. Interfaces, 33(6), pp.3-6.

Jones, S. & Fox, S., 2009. Generations Online in 2009. Online Report. Washingtton, D.C.: PewResearchCenter Pew Internet & American Life Project.

Krikke, H., Le Blanc, L. & Van der Velde, S., 2004. Product Modularity and the Design of Closed-Loop Supply Chains. California Management Review, 46(2), pp.23-39.

McCorkle, D.E., 1990. The role of perceived risk in mail order catalog shopping. Journal of Direct Marketing, 4(4), pp.26-35.

Mollenkopf, D.A., Rabinovich, E., Laseter, T.M. & Boyer, K.K., 2007. Managing Internet product returns: a focus on effective service operations. Decision Sciences, 38(2), pp.215-50.

Olson, J.R. & Boyer, K.K., 2003. Factors influencing the utilization of Internet purchasing in small organizations. Journal of Operations Management, 21(2), pp.225-45.

Petersen, J.A. & Kumar, J., 2009. Are Product Returns a Necessary Evil? Antecedents and Consequences. Journal of Marketing, 73(3), pp.35-51.

Rabinovich, E., 2007. Linking e-service quality and markups: the role of imperfect information in the supply chain. Journal of Operations Management, 25(1), pp.14-41.

Rabinovich, E., Sinha, R. & Laseter, T., 2011. Unlimited shelf space in Internet supply chains: Treasure trove or wasteland? Journal of Operations Management, 29(4), pp.305-17.

Ramdas, K., 2003. Managing product variety: An integrative review and research directions. Production and Operations Management, 12(1), pp.79-101.

Rodgers, S. & Harris, A.N., 2003. Gender and E-Commerce: An Exploratory Study.

Journal of Advertising Research, 43(3), pp.322-29.

Swinyard, W.R. & Smith, S.M., 2003. Why people (don't) shop online: A lifestyle study of the internet consumer. Phychology of Marketing, 20(7), pp.567-97.

Tan, J.J., 1999. Strategies for reducing consumers’ riskaversion in Internet Shopping.

Journal of Consumer Marketing, 16(2), pp.163-80.

Tarn, J.M., Razi, M.A..W.H.J. & Perez Jr, A.A., 2003. E-fulfillment: The strategy and operational requirements. Logistics Information Management, 16(5), pp.350-62.

(22)

22

Tullis, T.S., 2007. Older Adults and the Web: Lessons Learned from Eye-Tracking.

Lecture Notes in Computer Science, 4554, pp.1030-39.

Ulbrich, F., Christensen, T. & Stankus, L., 2011. Gender-specific on-line shopping preferences. Electronic Commerce Research, 11(2), pp.181-99.

Wood, S.L., 2001. Remote purchase environments: The influence of return policy leniency on two-stage decision processes. Journal of Marketing Research, 83(2), p.157–169.

Xiao, B. & Benbasat, I., 2007. E-commerce Product Recommendation Agents: Use Characteristics, and Impact. MIS Quarterly, 31(1), pp.137-209.

Yuksel, A., 2004. Shopping experience evaluation: a case of domestic and. Tourism Management, 25(6), pp.751-59.

Reducing Product Returns in Internet Retailing through Customer Differentiation

Contents

Preface

Reducing Product Returns in Internet Retailing through Customer Differentiation

July 2012