“The importance of individual product return
policy elements and their effect on online store
choice decision”
P.C. Bos June 23, 2016
2
“The importance of individual product return policy elements and their effect on online store choice decision” Department of Marketing Master Thesis P.C.Bos Plutolaan 535, Groningen 06-83388966 Keesbos2002@hotmail.com
Supervisor: prof. dr. T.H.A.Bijmolt University of Groningen
Preface
This thesis is the final part of the master Marketing Intelligence at the University of Groningen. The process, planning and execution of the work in this thesis was coordinated by my supervisor prof. dr. T.H.A. Bijmolt (Faculty of Economics and Business). I was engaged in this thesis from February 2016 to June 2016.
3
I would like to thank my supervisor for assisting and coordinating this project. Many thanks also go to dr. Felix Eggers (Faculty of Economics and Business) for assisting in the research design of the choice based conjoint and being available for questions during the process. I would also like to thank the respondents for filling in the questionnaire of this thesis. Without them I would have been unable to perform the analyses and test the hypotheses.
4
Management summary
The increasing popularity of buying and ordering products online comes with many
advantages for both the online retailers and the customers. Retailers and customers nowadays are able to search for a product and order a product 24/7. However, just buying and ordering the products online without feeling, tasting or actually seeing might lead to disappointments and feelings of regret. Returning products to the stores is a phenomenon which is becoming more and more common. The possibility to return a product sounds as the perfect solution for customers but is a major problem for online retailers. Processing and restocking these returns comes with high costs and online retailers need to optimize their product return policies to deal with this problem.
The main cause of this research was to analyze the effect of multiple elements of a product return policy on online store choice decisions by customers. The results of this research are relevant to online retailers since they need to optimize their return policy in able to avoid high costs but still attract customers to their store.
A return policy often consists of five elements: Time, scope, effort, exchange and monetary. These five elements all have their own level of strictness. However, the five elements needed to be independent (for the sake of the conjoint analysis) from each other which led to a merge of the monetary and the exchange element of the return policy. To test if there are any
differences between the five elements of a return policy and which elements are most important to customers when it comes to selecting an online clothing store, a choice-based conjoint analysis was conducted. A choice-based conjoint analysis gives the opportunity to measure the preferences of the respondents by offering them multiple choice-sets. By
selecting their favorite and most preferred option, the utility of every level of a product return policy element can be measured.
5
attribute, but also gives the opportunity to divide the respondents into segments based on their preferences. These segments have their own socio-demographic characteristics which makes the segmentation very useful for retailers with different target groups.
The choice-based conjoint analysis showed that the monetary/exchange element of a product return policy is relatively the most important to customers when choosing an online clothing store. Customers highly value the possibility of a full money-back guarantee. All of the five elements differed a lot in relative attribute importance. Time and effort are relatively the least important aspects of a return policy.
From these results, the management of an online retailer can decide to balance their policy for example in a way that customers have a full money-back guarantee, but they need to fill in additional forms and return the product in the original packaging.
6
Contents
Chapter 1: Introduction ... 8 1.1 Research question ... 9 1.2 Relevance ... 9 1.3 Structure of thesis ... 10Chapter 2: Theoretical framework ... 10
2.1 Product return ... 11
2.2 Product return policy leniency and elements ... 12
2.3 Hypotheses ... 14
Chapter 3: Research design ... 19
8
Chapter 1: Introduction
The increasing use of different devices by customers and multichannel strategies by retailers creates the tempting and accessible atmosphere to buy lots of products online. E-commerce is becoming more and more popular to retailers because it gives them the opportunity to sell their products 24 hours a day and 7 days a week. For customers the trend towards e-commerce brings some advantages because of the accessibility of products and the reduction of the search cost (Samar and Setoputro 2004).
This trend towards online shopping goes hand in hand with the ever trending topic of product returns. Customers like to see, feel and try out a product first before they eventually decide to keep the product. According to Peterson and Anderson (2013) “a product return occurs when a customer brings back a previously purchased product with the intent to receive a repair, refund, or replacement’’. The main reasons for returning a product are that customers do not understand how the product works, not like the product after all or just regret an impulsive purchase (Lawton 2008). Sometimes consumers even return a product because of “opportunistic behavior”. In this case, a consumer buys a product, uses the product for a while, and then returns the product to the store to collect its refund (Shulman et al 2009). Product returns are often free of charge for customers, which makes it very attractive to order and try large quantities of products. Some firms in Europe experience return rates which are close to 75% because of the trend towards a lenient product return policy (Mostard et al 2005).
9
A product return policy consists of different elements which can vary in terms of strictness. Petersen, Andrew and Kumar (2009) found in their research that strict policies impact long-term customer value because of reductions in future purchases. Combining different product return policies in such a way that customers will choose one store over another is a challenge for retailers.
1.1 Research question
The research question of this thesis is:
What is the effect of the different elements of a product return policy on the online store choice decision of customers?
Another question addressed in this research is whether the strictness of the return policy has an effect on store choice at all. Customers might just look at marketing mix variables to base their store choice decision on and not even look at the return policy at all. The general price level of the stores will be dissimilar across the alternative choices for customers to test the effect of this marketing mix variable.
The effect of the return policy on store choice might be influenced by previous internet purchases and return behavior. Experience of past actions might influence the process of gaining information in future decisions (Simonsohn et al 2008). Previous internet purchases and returning behavior will be treated as moderating variables in this research.
A final question of this topic is to analyze if there are any differences between groups of customers in their preferences towards return policies. A segmentation will be conducted to find the differences between groups of respondents based on socio-demographic characteristics.
1.2 Relevance
10
coordination and might be very expensive. Therefore, product return elements with little importance can be partly neglected by managers while at the same time the more important elements can be exploited. The segmentation of the respondents in multiple segments can also be very useful for managers when creating or adjusting their return policy. Different target groups might require different return policies to minimize returns and maximize profits.
1.3 Structure of thesis
The first part of this thesis will focus on previous literature and studies carried out in the topic of product return. The various elements of a product return policy will be identified and explained in this part. Furthermore the previous findings about the effect of strict and lenient return policies will be discussed. Hypotheses are formed and explained in this section of the thesis. Thereafter, an overview of previous literature is compiled to give some interesting insights and a clear overview of previous research. The results from the literature review are used in the research design of this study .
The next part of this thesis will explain the research design and the plan of analysis of the study. The respondents who participated in this study will be described in this section as well. It will also elaborate on the advantages and disadvantages of the choice-based conjoint analysis. Subsequently the method of data collection will be addressed.
Thereafter, the results of the analysis and the answers to the hypotheses will be presented. Finally the discussion, main conclusions, recommendations, implications and advice for further research will be given.
Chapter 2: Theoretical framework
11
2.1 Product return
Having a product return policy has some advantages as well as disadvantages. A big advantage of having a product return policy is to satisfy customers and trying to reduce the uncertainty that customers have towards the product. Retailers have to control and keep improving their quality standards since the return of products by customers depends for a large part on product quality (Li and Xu 2013). If the product is broken or does not meet the quality expectations a customer has, product return will be stimulated. A return policy can also take the role of signaling, as a lenient return policy is often an indicator for high product quality (Li and Xu 2013). A strict policy is then again conceived by customers as a signal of low product quality (Wood 2001; Anderson et al 2009).
Besides the role of quality signaling there is another advantage for retailers in adopting a return policy. Che (1996) studied the return policy for experience goods and describes the price setting for retailers of those goods. He states that retailers (especially monopoly sellers) can raise their prices for protecting and insuring customers for ex-post loss. This is the loss as perceived by customers after they purchased the product.
The major disadvantage of product return for e-retailers is obviously the cost it brings. This holds especially for online stores since return rates of those stores are typically higher than offline stores (Ofek et al 2011; Peterson and Anderson 2013). This can be explained by the fact that customers who buy products online cannot sense, feel and taste the product before buying the product. Firms tend to see product return as a drain on their profits and as cost of doing business (Petersen and Anderson 2013). Examples of costs that occur for retailers are inspection costs, repackaging costs and shipping costs (Su 2009). Heiman et al (2001) divide the cost of returning products in three groups:
• Loss in value of a returned product. The returned product might contain traces of use which might influence the value of a product in a negative way.
• Cost of handling the returned product. This includes the cost of checking the returned product and the whole process of packaging and preparing the returned product for a new customer.
12
Making the return policy very strict might reduce this cost but this might have a negative influence on the purchase intentions of customers (Bechwati et al 2005).
Another disadvantage of product return is the chance of customer abuse (Davis et all 1998). Customers might use the return policy to their advantage and might order products they do not need and will return for sure.
This thesis focusses on the effect of different product return policy elements on the store choice of consumers. Note that product return policy not only exists in the business to consumer environment. A topic in previous literature concerning product returns which cannot be ignored (since a lot of research is done in this topic) is the topic of return policies for retailers in a business to business environment (Pasternack 2008; Cai et al 2009; Su 2009). Pasternack (2008) researches the different return policies between manufacturer and retailer. The findings were not unambiguous for every retailer and manufacturer. Whether there should be a full refund policy or a partial refund policy depends on a function of retailers demand. It was proven by Pasternack that partial credit refunds for unsold products lead to a high channel coordination. Manufacturers may also charge the retailer a return penalty such as a restocking fee. This can bridge the gap in extra costs from product returns but can also lead to a shift in the demand curve potentially affecting the quantity sold by the manufacturer (Shulman et al 2009).
2.2 Product return policy leniency and elements
A distinction can be made between strict and lenient return policies. There has been quite some research done on both strict and lenient return policies and their effect on return behavior and sales (Wood 2001; Mukhopadhyay and Setoputro 2004; Bechwati et al 2005). Overall, customers value a lenient return policy, but the research on the effect of leniency on purchase and return behavior is inconclusive (Janakiraman et al 2015).
13
When it comes to return rates, there were some interesting findings about the effect of the leniency of a return policy on those return rates. Because of the popularity of trying out products and the possibility to return all of those products, customers get tempted to order lots and lots of products. Bonifield, Cole and Schultz (2010) showed in their study that there is a positive relationship between the number of products ordered and the number of returned products. This suggests that a lenient policy might lead to higher return rates. However, Wood (2001) contradicts this theory since he found that there was no effect at all of a lenient product return policy on return rates.
A return policy normally consists of many different product return policy elements. There are only a few studies that specifically focus on the effect of specific elements of the product return policy. Those studies either test the effect of policy elements on purchase, or on returning behavior ( Janakiraman et al 2015; Wood 2001; Samar and Seputro 2004; Li and Xu 2013). There is little research about the influence of specific elements of a return policy on the store choice of the customer.
The product return policies of an online retailer accounts for an important part of the purchase process of customers. Su (2009) found that over 70 percent of the customers consider the product return policies of a retailer before they decide to order the product. Previous research tried to capture all of the different individual elements of the product return policy. Davis et al (1998) and Heiman et al (2001) had some of the most influential and explicit distinctions in the various elements of a product return policy.
In more recent research, Janakiraman et al (2015) captures the different elements of a product return policy in 5 categories: Time leniency, monetary leniency, effort leniency, scope leniency and exchange leniency. These 5 categories are overlapping with the categorizations of Davis et al (1998) and Heiman et al (2001) as it captures all of their elements. Because of this complete specification of policy elements, this is the categorization that will be used in this research as well. The 5 categories of the product return policy can be explained as followed:
14
• Customers may receive full monetary refunds (100% refunds) or partial monetary refunds (f.e.85% or 50%) because of shipping or restocking costs. This is defined as Monetary Leniency. Partial monetary refunds are often seen as unfair by customers because the retailer gets to keep a portion of the refund (Davis et al 1998).
• Retailers can ask for receipts, original packaging, additional forms and all kinds of other requirements for returning the product. This comes under the Effort Leniency category.
• Scope Leniency is about the range of products accepted by the retailer for return. For example, in a lenient policy both a defect and a non-defect product can be returned to the retailer while in a strict policy only a defective product qualify for returning (Petersen and Kumar 2010). Also, while some retailers accept returning products which were on sale, others do not accept product return for items on sale.
• Customers might receive cash refunds or they may be entitled to non-cash refunds (discount or replacement for example).This is all captured in the Exchange Leniency category.
2.3 Hypotheses
So far, in this chapter the advantages and disadvantages of a product return policy have been discussed as well as the different elements of the return policy. Hypotheses will be formed partly based on this previous distinctions between the different policy elements.
The five different return policy elements as discussed in chapter 2.2 may differ in importance to the respondents and in their effect on online store choice decisions. This will be analyzed and discussed in the results section of this report. But since the elements differ a lot form each other at first sight, the first hypothesis that will be tested is:
H1: The different elements of the return policy differ in importance to customers when it comes to store choice
15
the perception of being treated fairly or unfairly when it comes to the requested price. Since price is that important, the following hypothesis is constructed:
H2: Monetary Leniency has the highest relative attribute importance of the 5 return policy elements
When considering respondents preferences, it is reasonable to believe that respondents will always choose those conditions which are most favorable to them. This leads to the third hypothesis:
H3: For every individual return policy element, more lenient attribute levels are always preferred
Not only managing the return policy, but also other factors might affect the store choice. Yu and Wang (2008) found that different customer classes should be targeted by different return policy elements. This means that different customers might react differently to varying prices and products. Marketing mix variables such as price and product also impact the way a customer values the firm and its products. This in turn might influence return behavior and the firms profitability (Petersen and Anderson 2013).
The effect of the different elements of the return policy on store choice might differ across various types of products. Previous research used multiple types of products, for example music, seasonal goods, electronics or books (Davis et al 1998; Padmanabhan and Png 1997) On average a product is returned in 6% of all sales but this varies widely across industries (Ketzenberg and Zuidwijk 2009).
16
product type for their research (Padmanabhan and Png 1997 ; Kang and Johnson 2009 ;Li and Xu 2013) because it is known for its high return rate (Hess and Mayhew 1997). Ofek et al (2011) found that product types where consumers can benefit from store assistance to find the right fit, typically have a larger likelihood of return when sold online.
Ofek et al (2011) found that some sellers (especially in the fashion industry) do not even sell some of their product categories online anymore because of the extremely high risk of return. The high return rates in the fashion industry make clothing a very interesting product type when doing research in return policies. Therefore, this product type will be used during the data collection further on in this thesis.
The main topic is to test the effect of the individual product return policy elements on store choice. However, this effect might be influenced by different variables as described in earlier paragraphs. The relevance of the individual return policy elements might differ for different general price levels of the store. The price level variable also has a direct link to store choice since it might be reasonable to think that some respondents just ignore the return policy and base their choice on the price level of the store.
A previous study showed that lower priced products had lower return rates than higher priced products when all else being equal (Anderson et al 2009). Managing price carefully in combination with encouraging return policies might stimulate sales because of the price risk. This is about concerns customers have about prices that might rise in the future. This might lead to customers accelerating on their purchase decision and buying the product as soon as possible (Su 2009).
Another moderating variable to consider are the socio-demographic variables like age, education, income and gender. These might have a moderating effect on the relationship between the marketing mix and the store choice. The socio-demographic variables may also have a moderating effect on the relationship between return policy and store choice. For example, age might influence the importance of a single return policy attribute on store choice.
17
purchasing and returning products) respondents might show different results in store choice than less experienced respondents. Simonsohn et al (2008) studied the impact of experience on processing information. They found that information gained from experience might lead to an influential change in behavior. It is reasonable to expect that customers with experience in returning products might have a higher return rate than customers without experience and will not be discouraged by the extra effort of returning a product. This will be the fourth hypothesis of this report:
H4: Respondents with experience in returning products are less susceptible towards the effort element of the return policy
In this research, the respondents have to choose a store when taking into account two different contexts. One in which the respondents have to shop for an expensive piece of clothing for a special occasion and one in which the respondents have to shop for a piece of clothing for daily use (lower priced). Shopping for expensive clothing might trigger peoples need for protecting themselves of wasting their money. It is therefore reasonable to think that the monetary element of the return policy (e.g. 100% money back) will be more important in the context of people buying an expensive piece of clothing for a special occasion. This reasoning allows for the fifth hypothesis:
18
The conceptual model as appears in figure 1 shows the aim of the research in this paper.
Figure 1: Conceptual model
19
Chapter 3: Research design
3.1 Research
The subject of this research is to test if there are any significant differences between the different elements of a product return policy. To test if there are any significant differences between the elements and to gain insight into the relative importance of every single element on store choice, a choice-based conjoint analysis was created. The following part will discuss the design of the research and provide a plan of analysis.
An important advantage of using a choice-based conjoint analysis is that it provides lots of information. Including the relevant attributes in multiple choice sets, it shows the preferences of the respondents and offers the information needed to analyze the differences of importance in the attribute levels. Another advantage of a choice-based conjoint analysis is the possibility to include a no-choice option. This adds extra realism (Haaijer et al 2001) since respondents might choose to make no decision in different options at all in real life and just walk out of the stores without buying a single product.
An important disadvantage of a choice-based conjoint analysis can be the length of the questionnaire. Even one choice set is a lot to read for the respondents. Important is to keep the number of choice sets somewhere between 8 and 16 (Eggers et al 2015) to make sure the respondents keep their attention on the questionnaire and to keep the analysis valid. Fatigue effects by respondents might lead to errors in the choice decision process and this could bias the parameters.
3.2 Questions
20
As discussed in the literature review chapter, there are five types of individual policy elements that can be distinguished: Time Leniency, Effort Leniency, Monetary Leniency, Exchange Leniency and Scope Leniency. However, for a choice-based conjoint to be valid, the different attributes should be independent from each other. When using all the five different elements, this criteria cannot be met because the Monetary Leniency and the Exchange Leniency are not completely independent. The Monetary Leniency can be either a full refund or a partial refund. The Exchange Leniency is about what a customer receives in return (for example an alternative product). Obviously, an alternative product cannot be combined with a partial refund. To solve this problem, the Monetary Leniency and the Exchange Leniency will be treated as a combined attribute. The other individual elements of the product return policy will remain unchanged which means there will be four return policy attributes in the choice-based conjoint analysis.
Besides the return policy attributes there is a price level attribute included in the conjoint. The attribute levels of the price variable in the choice set cannot be too different from each other since price might have a major influence on store choice. A choice was made for a minor difference of 10% above the average price level and 10% below the average price level. In this way the price attribute has been taken into account without having an influence that might be too heavy-weighted for the conjoint analysis.
21
Table 1: General product return policies of five large online retailers.
A European law which must be taken into account is the right of withdrawal. Since June 2014, consumers have the right to cancel their online purchase within 14 days. Before 2014 this was just 7 days so lots of web stores had to change their return policy. Within these 14 days, the consumer has the right to return the product without the need for a good reason. All the different attributes have the same number of levels, three in this case. This is important because when attributes differ in the number of attribute levels, the number-of-levels effect might occur (Verlegh et al 2002). This is a problem because this effect might lead to a higher relevance of attributes which have more levels. Important is that each of the attribute levels is shown once in the choice set. This, in combination with the random selection of stimuli, keeps the choice sets balanced and as orthogonal as possible.
Bol.com Wehkamp Amazon Otto Ebay Time Leniency 14 or 30
days
14 days 14, 30 or 90
days
14 days 14 days
Effort Leniency Original packaging, return label
Return label Return label Original
packaging, additional forms Return label Monetary/exchange Leniency 100% refund Shipping costs Replacement, restocking fee Shipping costs Shipping costs
Scope Leniency Any product Not customer specific
Any product Any
product
22
The different attributes and the attribute levels to measure the importance of the individual return policy elements can be found in table 2 below:
Time Leniency
Within 30 days after acceptance
Within 21 days after acceptance
Within 14 days after acceptance
Scope Leniency
Every product can be returned
Every product except for the products which are on sale
Every product except for personalized products
Effort Leniency
Product and additional forms
Product and the original packaging
Just the product
Monetary/exchange Leniency Alternative product or 100% money back Alternative product or 85% money back (because of shipping costs) Alternative product, no money back
Respondents were instructed with the different scenarios as offered in the questionnaire to make sure they keep in mind the two different scenarios. Furthermore, respondents were secured that the information they provide is confidential and that the whole questionnaire will remain anonymous. The respondents were given a short introduction to the questionnaire without influencing them in a way which might bias the results. The topic of the questionnaire was not known beforehand to the respondents to make sure the validity is as high as possible. The respondents were just instructed with the fact that they had to make some choices in the questionnaire.
The questionnaire was send online (after pretesting the questionnaire) to the respondents. The respondents were selected according to the principles of convenience sampling. The main
23
advantage of convenience sampling is the high accessibility of the subjects. Other advantages of the convenience sampling technique is that it is relatively fast and inexpensive. An important disadvantage of convenience sampling to consider is that the sample might not be representative for the entire Dutch population that shops for online clothing.
The choice sets were randomly generated to secure a great variety of choice sets among the respondents and all of the possible interaction effect can be measured. However, there were also two holdout choice sets included (one for each scenario). Including the holdout choice sets will allow for direct comparison between the two scenarios (appendix 1). There was also a dual response no-choice option included in the choice set. Respondents had to first pick one of the alternatives to make sure they give their preferences. After they picked one of the three alternatives the respondents had to reveal if they would actually buy something in the store of their choice (yes or no).
The results of the questionnaire were analyzed with LatentGold, which is a program designed to estimate choice models. This program estimates the relative attribute importance of the different product return policy elements. It shows which of the individual attributes is most important to the respondents. Furthermore it reveals the most preferred attribute levels for each product return element. Another interesting option of LatentGold is the segmentation function. The results of the questionnaire can be ordered by the differences in for example age, gender or education. Some of the product return elements can differ in importance related to socio-demographic variables. Segmentation can offer valuable insights in the differences within those socio-demographic variables.
3.3 Respondents
24
The 10 randomly selected choice sets were used for estimating the model. The holdout choice sets were left out in these models because they were created by the researcher which might lead to an experimenter bias. The holdout choice sets were compared to see the effect of different scenarios on attribute importance. The no-choice option was chosen to left out of the estimation to measure purely the preferences of the respondents.
Table 3 shows the distribution of income and education among the respondents. The
questionnaire was handed out to Dutch respondents so the education and income divisions might be different in other countries. Note that not all of the respondents were willing to answer the question about their income level.
Education level Number of respondents
Monthly household income
Number of respondents
Primary education 1 (0,83%) 0-1000 Euro’s 71 (59,17%)
Secondary education 41 (34,17%) 1000-2000 Euro’s 22 (18,33%)
Intermediate vocational education 7 (5,83%) 2000-1000 Euro’s 7 (5,83%) Higher vocational education bachelor 18 (15,00%) 3000-4000 Euro’s 0 (0,00%)
University bachelor 35 (29,17%) 4000-5000 Euro’s 3 (2,50%)
Master’s degree 18 (15,00%) Over 5000 Euro’s 3 (2,50%)
Unknown 14 (11,66%) Table 3: Distribution of levels of income and education
25
26
Chapter 4: Results
4.1 Model fit
When trying to find and estimate the best model, multiple models were created and compared. In the first model, all of the attributes in the choice sets are treated as nominal variables and thus the first model is a full part-worth model. This first model consists of 10 parameters, the null model consists of 0 parameters. The null model has a log likelihood (LL) of -966,24 and a LL² of 1932,48. This value needs to be compared with the range of likelihood values and therefore the minimum log likelihood (LL(0)) needs to be calculated. The LL(0) is calculated by the number of respondents multiplied by the number of choice sets and the natural logarithm of 1 divided by the number of alternatives per choice set. The LL(0) in this case is -1318,33. A likelihood ratio test was done to test if the first estimated model is significantly better than the null model. The chi-square of the likelihood ratio test is 704,18. The chi-square table gives a p-value of (df=10, α=5%, critical value=18,31) less than 0,01, which means that the first estimated model differs significantly from the null model. This tells that the respondents did not answer the questionnaire at random.
Another test to obtain the model fit is the Pseudo R², also known as McFaddens R². The Pseudo R² adjusted was calculated to correct for the increasing number of parameters. This is calculated by dividing the LL(1) (after correcting for the number of parameters, which is still 10) with the LL(0) and subtract this from 1. The Pseudo R² adjusted for the first model is 0,27. A Pseudo R² adjusted between 0,2 and 0,4 is widely accepted as acceptable (Eggers et al 2015).
27
Next, the first estimated part-worth model is compared with other models in which not all of the attributes are treated as nominal variables. The results of this comparison are shown in table 4 below. The log likelihood test of the created models after the first part-worth model were compared with the best model fit at that time (see also table 5 about information criteria) and not with the null model anymore. Effort and Scope Leniency have nominal attribute levels so these attributes are not treated as linear variables at all.
Table 4: Goodness of fit comparison between models
Another way to select the best model is to look at some of the information criteria. In table 5, the AIC, BIC and CAIC of the different models are compared.
Model Numeric variables AIC BIC CAIC
1 All nominal 1952,48 1980,35 1990,35
2 Price 1997,31 2022,40 2031,40
3 Time 1951,98 1977,06 1986,06
4 Exchange+monetary 2576,10 2601,19 2610,19
Table 5: Information criteria based on the log likelihood Model Numeric variables Number of
28
29
4.2 Model parameters
After the selection of the model with the best fit, the beta’s for the different attribute levels can be examined. These attribute parameters can be found in table 6 below.
Attribute/level ß P-value Price <0,01 Average 0,073 10% less expensive 0,327 10% more expensive -0,399 Time 0,27 Within 14 days 0,073 Within 21 days -0,060 Within 30 days -0,013 Scope <0,01 All products 0,231
All, except personalized 0,037
All, except products on sale -0,269
Effort
Just the product 0,034 0,40
Product and additional forms -0,067
30
Monetary/Exchange <0,01
Alternative product -0,603
Alternative product or 100% money back 1,031
Alternative product or 85% money back (shipping cost and restocking cost)
-0,428
Table 6: Attribute parameters
Table 6 shows that 3 out of the 5 attributes actually have a significant effect (α=5% ) on store choice. Price, Scope and Monetary/exchange all have a p-value lower than 0,01. Time and Effort Leniency do not have a significant effect on store choice according to this research.
Arguably the most important output given by LatentGold is the one of relative attribute importance. The relative attribute importance is a measure of how much influence every single attribute has on the choices (in this case it is about the influence of the attributes on the store choice) made by the respondents. Table 7 below shows the relative attribute importance of the attributes which were included in the questionnaire:
Attribute Relative attribute importance
Price 23,45%
Time Leniency 4,31%
Scope Leniency 16,15%
Effort Leniency 3,28%
31
The relative attribute importance of the Monetary/exchange Leniency is by far the highest (52,81%). This might also be an indicator of large differences in the part-worth utilities within the attribute. Scope Leniency and Price Leniency are relatively the second and third most influential attributes. According to this research, Time Leniency and Effort Leniency are relatively not that influential on the store choice decision. Relative attribute importance is ratio-scaled and thus has an meaningful null point. The first hypothesis questioned the differences between the attributes in relative attribute importance:
H1: The different elements of the return policy differ in importance to customers when it comes to store choice
The differences between the different product return policy elements are substantial. Time Leniency only has a relative attribute importance of 4,31% while Monetary/exchange Leniency has a relative attribute importance of 52,81%. Since relative attribute importance is ratio scaled, in this case, Monetary/exchange Leniency is over 12 times as important as Time Leniency. This hypothesis will not be rejected.
The next hypothesis discussed is the second hypothesis which also considers the relative attribute importance:
H2: Monetary Leniency has the highest relative attribute importance of the 4 return policy elements
32
The third hypothesis can be answered with the help of table 6:
H3: For every individual return policy element, more lenient attribute levels are always preferred
This hypothesis is hard to answer for the Scope and the Effort Leniency because these attribute levels are nominal. Price level and Monetary/exchange Leniency both show the highest beta’s for the more lenient attribute levels. More surprisingly are the betas for the Time Leniency since the shortest time period shows the highest beta (0,073). This attribute however does not have a significant effect on the store choice so the preference for the attribute levels of Time Leniency are negligible. This hypothesis is true for the significant attribute levels so this hypothesis will not be rejected.
4.3 Market Simulation
The parameters (see table 6) can also be used for market predictions and market simulation. Imagine for example that there are 3 stores, store A,B and C. The price level and the return policies of the store are shown in table 8 below:
Store A
* 10% cheaper price * 21 days to return * No products on sale can be returned
* Send product and original packaging *Receive alternative product Store B * 10% more exp.price * 30 days to return * All products can be returned
* Send product and additional forms *Receive alternative product or 100% money back Store C * Average price * 14 days to return * No personalized products can be returned
* Just return the product *Receive alternative product or 85% money back
33
Store A has the lowest market share at the moment and wants to improve its market share by changing the Monetary/exchange element of their policy since they found out this element is very important to customers. If store A offers a 100% money back guarantee to their
customers, their market share (based on the beta’s found in table 6) will rise to 49% which will make store A the market leader (market shares of store B and C are respectively 37% and 14%). These market simulations and predictions can be very useful to managers to measure the effects of their ideas and decisions.
4.4 Segmentation
A priori segmentation was used to search for differences in preferences between groups of respondents based on different characteristics. First, gender was used as a segmentation variable. Interestingly, men and woman showed significant differences (p<0,05) in Price level and Monetary/exchange Leniency. The largest difference in the Monetary/exchange attribute seems to be that woman dislike alternative products a lot more than men do (ß= -0,916 for female, ß=-0,364 for male). Another very interesting finding when looking at the differences in men and woman is visible in the price levels of the stores. The range in betas for men is higher than the ones for woman. Men have a higher beta than woman when it comes to cheaper stores and a lower beta at the more expensive stores. This might be an indication that the male respondents in this research are more price sensitive than the female respondents are. Similarly, the relative attribute importance for the price level is way higher for men than for woman (30,70% against 13,93%) while the relative importance for the Monetary/exchange attribute is higher for woman than for men (49,51% against 58,06%).
34
linear pattern in the differences in the Monetary/exchange Leniency between the income groups.
The following a priori segmentation considers the experience of respondents in returning products and is connected with hypothesis four.
H4: Respondents with experience in returning products are less susceptible towards the effort element of the return policy
When segmenting on return behavior, there is no significant difference in the effort attribute between the respondents which are more experienced with returning products and the respondents which have less experience with returning the products (p>0,05). This hypothesis can therefore be rejected. There is a significant difference in price level and Monetary/exchange Leniency between the different groups of product returners. Interesting is the trend the respondents show in the popularity of selecting an alternative product: The more experienced a respondent is with returning a product, the lower the beta for choosing an alternative product becomes. This might be because of bad experiences in the past.
Finally an a priori segment was conducted for experience internet users and less experienced internet users. For the experience in internet purchases there were no significant differences between groups at all (p>0,05)
35
classification error is required between the models with a different number of segments. This comparison is summarized in table 9:
Number of segments
LL df AIC BIC CAIC Classification error 1 -966,24 230 1952,48 1987,29 1997,29 0,00 2 -879,88 199 1841,77 1984,47 2025,47 0,08 3 -840,77 168 1825,54 2076,15 2148,16 0,10 4 -787,51 137 1781,02 2139,53 2242,53 0,02 5 -765,42 106 1798,85 2265,25 2399,25 0,06
Table 9: Latent class comparison
The BIC is the lowest for 2 segments, the CAIC is lowest for 1 segment. The CAIC of 2 segments does not differ a lot from the CAIC of 1 segment and is increasing a lot from 3 segments onwards. The classification error for 2 segments is relatively high compared to the other models (0,08). However, because of the low BIC and CAIC, and despite the higher classification error, 2 segments will be used for further segmentation analysis. A description of an average respondent in both of the segments is summarized in table 10:
Table 10: The two segments
The relative attribute importance for the 2 segments can be found in table 11 below:
Segment Size Age Gender Education Income (average)
Return behavior
Scenario
1 58,80% 22,8 Female High school 1050 2x per year Expensive
36
Segment Price Time Scope Effort Monetary/exchange
1 10,68% 3,60% 17,86% 7,32% 60,53%
2 42,63% 13,90% 15,42% 4,13% 23,92%
Table 11: relative attribute importance per segment
The first segment can be described as a somewhat younger segment with a relatively high income. The first segment consists of especially people who shop for expensive clothing online. The second segment especially shops for cheaper clothing and this is in line with the relatively lower income of the second segment. This first segment is strongly influenced by the Monetary/exchange return policy attribute when it comes to choosing a store. This is in line with the findings from the a priori segmentation on gender. The second segment is especially characterized by their lower income and are looking for cheap clothing for daily use. In line with the low income of the second segment is the high relative attribute importance for the price level of a store (42,63%). The two segments do not differ in education. The first segment is more experienced in returning products bought online compared to the second segment. The relevance of segmenting the respondents will be further refined in the discussion section of this thesis.
The final hypothesis to discuss is the following:
H5: When buying clothes online, the monetary policy element is relatively more important in the scenario of clothes for a special occasion, when compared to the scenario of clothes for daily use.
37
we indeed found a difference in relative importance for the Monetary/exchange attribute (42,23% for the daily use scenario and 57,08% for the special occasion scenario). Because of this difference of over 15% in attribute importance, H5 will not be rejected. The other differences in relative attribute importance can be found in table 12:
Scenario Price Time Scope Effort Monetary/exchange Clothing for daily use 32,68% 7,19% 15,48% 2,42% 42,23%
Clothing for a special occasion
12,26% 10,06% 15,71% 4,89% 57,08%
Table 12: relative attribute importance of two scenarios
The differences in relative attribute importance in the price level and monetary/exchange attribute make sense. When respondents get a scenario where they do not have to worry about spending a lot of money (they already decided on buying an expensive product), price level will be less important. Similarly, when buying an expensive piece of clothing for a special occasion, people might strongly demand for the guarantee of 100% money back when they want to return the product.
Another way to compare the two different scenarios is by direct comparison of the two holdout sets. The result of the picks for the holdout sets is summarized in table 13. The holdout set itself can be found in appendix 1.
Scenario Option 1 Option 2 Option 3
Clothing for daily use
39,2% 25,8% 35,0%
Clothing for a special occasion
63,3% 21,7% 15,0%
38
The first scenario, in which the respondent had to imagine him/herself to buy a piece of clothing for daily use, barely has any differences. Not one of the attributes had a significant effect (p>0,05) on the store choice in the clothing for daily use scenario. The range in the second scenario, the one where respondents have to buy a piece of clothing for a special occasion, is way larger. The price level had an significant effect (p<0,01) on the store choice in the scenario where people had to buy an expensive piece of clothing for a special occasion. When products are imagined to be more expensive, respondents tend to give more importance to the price level of the product. This is contradicting with the findings from the comparison between the randomly selected choice sets. This might be because of experimenter bias since the holdout set was not randomly created but the attribute levels were selected by the experimenter. Interestingly, when looking at the parameters of the attribute levels, the option where respondents could pick the most expensive store was the most popular option (ß=1,04 for most expensive price level, ß= -0,76 for cheapest price level). The comparison of other choice sets showed the expensive Price level as least favorable in both scenarios.
Chapter 5: Conclusions
5.1 Conclusion
39 Table 14: Overview of results
The different elements of a product return policy appear to have a different attribute importance. As shown in the results section, the monetary/exchange attribute has the highest relative attribute importance to customers. This is very useful to know for the management of online retailers. Customers attach high value to the 100% money back guarantee. Whether this guarantee is feasible for the online stores is hard to say, since every single online retailer has to consider its own cost structure when offering a full money back return policy.
Costs and logistics can possibly be saved on Effort and Time Leniency. Respondents are barely influenced by the regulations regarding how and when the product needs to be returned, compared to other aspects of the return policy. Time Leniency is an interesting part of the product leniency since we found that a lenient time policy (e.g. many days to think about the purchase and to return the product) is not valued highly by consumers and even had a negative beta.
Different circumstances of buying also play a significant role in determining the influence of the elements of a return policy. When shopping online for an expensive piece of clothing for a special occasion, consumers tend to give a lot of importance to the monetary/exchange part of the policy. This means consumers prefer a full money back guarantee, instead of paying for shipping cost or receiving just an alternative product. Choosing an online store for shopping for cheaper clothing for daily use is highly influenced by the price level of the store. Consumers find it important that online stores offer cheap prices when shopping for clothing for daily use.
5.2 Discussion
40
choice-based conjoint analysis showed some similarities with previous studies carried out in product returns. There were some major differences in the importance of the individual product return elements as expected. Customers are influenced a lot by the Monetary/exchange Leniency of an online retailer. Probably because of uncertainty , as described in the theoretical framework, customers are hesitant with purchasing a product online without a full money back guarantee. Therefore, it was not surprising that the choice-based conjoint analysis pointed out the Monetary/exchange attribute to be the most important. A result that was quite unexpected was the effect of the Time Leniency on store choice. Logically, one would expect more time to be preferred by customers when returning the product. However, the effect of Time Leniency on store choice turned out to be negative. Customers do not prefer more days to overthink their purchase. An explanation for this might be that 14 days already is enough time for consumers to decide whether or not to keep the product, more days only come with more doubts.
Dividing the respondents in different segments might be useful for online retailers when making a product return policy. Retailers know their target group of customers and might adjust the return policy to this group. When for example the target group consists of young adults (especially females) with a relatively high income, it is important to focus on offering your customers a full money back guarantee (as can be seen in the specifications of segment 1 in the results section). When focusing on males with a relative low income, it can be very useful to keep in mind that this segment is very unexperienced with returning products and values a low price. One way to target this segment might be to offer a low price but to keep the return policy strict to restrain the costs.
41
The reliability of this research is high, the 120 respondents provided clear preferences in product return attributes levels. When conducting the same research again, the results would probably be very similar. An important remark is that the reliability is only high for respondents with similar characteristics as the ones in this sample. The reliability could be very low if this research would be carried out under elderly in a whole other continent. This also holds for the representativeness of this research. The representativeness of this research is not that high when looking at the entire Dutch online shopper population, but quite high when only considering the younger, higher educated adults.
To assess the validity of this research we make a distinction between internal and external validity. The internal validity considers the causal relationship between the variables and whether the results can really be explained by the variables and analysis done in this research. The internal validity of this research is high since the results of this research were realized following the choice of the appropriate attributes and asking the respondents to fill in the questionnaire based on these attributes. The external validity is about the generalizability of this research. The external validity of this research is relatively low since the respondents in the sample are very similar and the results of this research cannot be generalized to the Dutch population that buys products in online stores.
5.3 Limitations
42
5.4 Further research
The respondents in this research had to answer the questions for buying clothing online. Generalizing the results of this thesis to other types of products (e.g. electronics or consumables) might be difficult. Further research on the importance of individual product return elements needs to be carried out in other product areas to confirm the generalizability of the results in this thesis.
43
References
Anderson, E. T., Hansen, K., & Simester, D. (2009). The option value of returns: Theory and empirical evidence. Marketing Science, 28(3), 405-423
Bechwati, N. N., & Siegal, W. S. (2005). The Impact of the Pre-choice Process on Product Returns. Journal of Marketing Research, 42 (3), 358–367
Blanchard, D. (2007). Supply Chains Also Work in Reverse. Industry Week, 256 (5), 48 Bonifield, C., Cole, C., & Schultz, R. L. (2010). Product returns on the Internet: A case of mixed signals?. Journal of Business Research, 63(9), 1058-1065
Che, Y. K. (1996). Customer return policies for experience goods. The Journal of Industrial
Economics, 44 (1), 17-24
Davis, S., Hagerty, M., & Gerstner, E. (1998). Return policies and the optimal level of “hassle”. Journal of Economics and Business, 50(5), 445-460
Dekker, R., Inderfurth, K., Fleischmann, M., & Van Wassenhove, L. N. (2004). Reverse
logistics: a quantitative approach. Berlin: Springer
Eggers, F., Sattler., H. Voelckner., Teichert., T.(2015) Conjoint analysis and discrete
choice experiments. Working paper
Guide Jr, V. D. R., Souza, G. C., Van Wassenhove, L. N., & Blackburn, J. D. (2006). Time value of commercial product returns. Management Science, 52(8), 1200-1214
Haaijer, R., Kamakura, W. A., & Wedel, M. (2001). The'no-choice'alternative in conjoint choice experiments. International Journal of Market Research, 43(1), 93-106
Heiman, A., McWilliams, B., & Zilberman, D. (2001). Demonstrations and money-back guarantees: Market mechanisms to reduce uncertainty. Journal of Business Research, 54(1), 71-84
44
Janakiraman, N., Syrdal, H. A., & Freling, R. (2015). The Effect of Return Policy Leniency on Consumer Purchase and Return Decisions: A Meta-analytic Review. Journal of
Retailing
Kang, M., & Johnson, K. (2009). Identifying characteristics of consumers who frequently return apparel. Journal of Fashion Marketing and Management: An International
Journal, 13(1), 37-48
Lawton, C. (2008). The war on returns. Wall Street Journal, 8(1)
Li, Y., Xu, L., & Li, D. (2013). Examining relationships between the return policy, product quality, and pricing strategy in online direct selling. International Journal of Production
Economics, 144(2), 451-460
Mostard, J., & Teunter, R. (2006). The newsboy problem with resalable returns: A single period model and case study. European Journal of Operational Research, 169(1), 81-96 Mukhopadhyay, S. K., & Setoputro, R. (2004). Reverse logistics in e-business: optimal price and return policy. International Journal of Physical Distribution & Logistics
Management, 34(1), 70-89
Ofek, E., Katona, Z., & Sarvary, M. (2011). “Bricks and clicks”: The impact of product returns on the strategies of multichannel retailers. Marketing Science, 30(1), 42-60
Padmanabhan, V., & Png, I. P. (1997). Manufacturer's return policies and retail competition. Marketing Science, 16(1), 81-94
Pasternack, B. A. (2008). Optimal pricing and return policies for perishable commodities. Marketing science, 27(1), 133-140
Petersen, J., & Anderson, E. T. (2013). Leveraging Product Returns to Maximize Customer
45
Petersen, J. A., & Kumar, V. (2010). Can Product Returns Make You Money?. MIT Sloan
Management Review, 51(3), 84-89
Shulman, J. D., Coughlan, A. T., & Savaskan, R. C. (2009). Optimal restocking fees and information provision in an integrated demand-supply model of product returns. Manufacturing & Service Operations Management, 11(4), 577-594
Simonsohn, U., Karlsson, N., Loewenstein, G., & Ariely, D. (2008). The tree of experience in the forest of information: Overweighing experienced relative to observed information. Games and Economic Behavior, 62(1), 263-286
Su, X. (2009). Consumer returns policies and supply chain performance.Manufacturing &
Service Operations Management, 11(4), 595-612
Suri, R., Manchanda, R. V., & Kohli, C. S. (2002). Comparing fixed price and discounted price strategies: the role of affect on evaluations. Journal of Product & Brand
Management, 11(3), 160-173
Verlegh, P. W., Schifferstein, H. N., & Wittink, D. R. (2002). Range and number-of-levels effects in derived and stated measures of attribute importance. Marketing Letters, 13(1), 41-52 Wang, X. (2009). Retail return policy, endowment effect, and consumption propensity: an experimental study. The BE Journal of Economic Analysis & Policy, 9(1)
Wood, S. L. (2001). Remote purchase environments: The influence of return policy leniency on two-stage decision processes. Journal of Marketing Research, 38(2), 157-169
Xia, L., Monroe, K. B., & Cox, J. L. (2004). The price is unfair! A conceptual framework of price fairness perceptions. Journal of marketing, 68(4), 1-15
46
47
Heiman et al 2001 Durables and
Consumables Time Monetary Effort Content analysis
Lenient policy can be a sign in product quality thus providing extra information to the customer
Wood 2001 Durables and
consumables Time Monetary Effort Exchange Scope
Experiment Greater leniency in return
policy leads to increased probability of order
Samar & Setoputro
2004 Not specified Time
Exchange Terms Effort
Experiment More lenient return policy is
good for the firm cause it induces customers to go ahead and buy
Pasternack 2008 Consumables Scope
Monetary
Cross sectional study
‘’an optimal policy in the multi-retailer environment is only achievable if unlimited returns are permitted for partial credit’’
Li & Xu 2013 Durables and
consumables Monetary Exchange Time Direct sales model. Joint decision model
Lenient policy gives higher demand, but also higher returns and cost.
Su 2009 Durables and Consumables Monetary Effort Cross sectional study
Partial refunds are optimal
Wang 2009 Consumables Time
Monetary
Experiment Lenient policy positive effect on purchasing likelihood
Davis et al 1998 Durables and
consumables Time Effort Exchange Empirical study
Apply product specific return policies Ketzenberg & Zuidwijk 2009 Durables Exchange Time Effort
Experiment Intermediate return policy is optimal
Yu and Wang 2008 Durables Monetary Cross sectional
study:
48 Appendix 1: Previous literature
Time Decision tree those classifications.
Kang and Johnson 2009 Durables Time Monetary Effort Exchange Scope Cross sectional study: Pearson correlation and multiple linear regression
A strict return policy should not affect impulsive buying behavior.
Che 1996 Durables Monetary Exploratory
study
Return policies of free trails can promote the process of customer learning and reduce the risk for the customer Ofek, Katona
and Sarvary
2011 Durables and
consumables
Not specified Cross sectional
study
Product return rates are likely to be high in product categories that need a high level of assistance by specialists.
Shulman et al 2009 Not specified Monetary
Exchange
Cross sectional study
49 Appendix 2 : Holdout choice
