Modelling the Relationship Between Product Variety and the Product Return Fraction

(1)

Modelling the Relationship Between Product

Variety and the Product Return Fraction

Master’s Thesis, MSc Supply Chain Management

University of Groningen, Faculty of Economics and Business

Name: Mengqiu Ge

Student Number: S2002469

Supervisor: Mr. van Foreest

(2)

2

Acknowledgements

(3)

3

Abstract

There is pressure to seek out causes of why products are returned because understanding this may lead to increased profits. This paper investigates the relationship between product variety and the product return fraction in order to help fast-fashion retailers make trade-offs between product variety benefits and product return costs. This research proposed seven definitions of product variety (style*colour*size; style*colour; style*size; colour*size; style; colour; size) with the support of literature and data given by the company. The product return fraction was measured by the number of products returned divided by the number of products sold within each product type. Consequently, we examined the relationship between each product variety and the product return fraction separately by inserting scatterplot. To observe the trend of dots on graphs, both polynomial and linear trend lines were added, along with R-squared values to test reliability. Findings show that style*colour and colour are negatively related to return fraction, while style*size exerts a positive influence on the product return fraction. The differences of return behaviours between women and men were also tested. Style*colour*size and colour*size play more important roles in women’s return behaviour. However, in the tests regarding the relationship between product variety and the product return fraction as well as the relationship between men and women, we found that results are not sufficient as R-squared values of all trend lines on graphs are lower than 0.6, which is far from the most reliable rate 1. In other words, product variety and the product return fraction are not correlated. E-retailers are therefore advised to provide plenty of apparel products to attract customers.

(4)

4

1. Introduction

In order to stay competitive and dominate apparel industry markets, e-retailers provide various products to attract customers (Mehrojoo & Pasek, 2014). This selling strategy drives sales up significantly and profits increases as a result (Kekre & Srinivasan, 1990). However, increased product variety also leads to more returned products (Petersen & Kumar, 2009). Once the product is returned, e-retailers not only need to return money to customers, but also bear the refurnish and restock costs, which can impact the financial success of the company (Aberdeen, 2009).

Although existing literature found that product variety is positively related to product return amount (Petersen & Kumar, 2009), this does not imply that product variety and the product return fraction are related. Mathematically speaking, the change of absolute product return number has not been proven to be associated with the change of relative return fraction. There are three possibilities regarding the change of the product return fraction. The ideal situation is that the relative return fraction does not change even though returns nominally increase. As an example, the product return fraction is 10% when 200 products are returned out of 2,000 products sold; when 5,000 products are on offer and 500 are returned, the return fraction is also 10%. When the product return fraction is constant and insensitive to product variety, it becomes easier for e-retailers to decide on the number of products to offer and manage the product return disposition. The second possibility is that if product variety correlated to the product return fraction, profits may diminish significantly due to increased return costs. The third possibility is the opposite of the second one. When product variety is uncorrelated to the product return fraction, e-retailers are able to earn more profits in accordance to increased sales and decreased return costs. Since there is little knowledge regarding the change of the product return fraction resulting from increased product variety in the apparel industry, e-retailers offer a multitude of products in an attempt to retain customers and sales (Vaagen & Wallace, 2008). However, the profit margin they may be gaining is not clear as return costs are not considered. This lack of consideration is the main reason for this research.

(6)

6

affecting customers’ selection of products in the product-seeking stage and determines the buying decision (Eckman et al., 1990). Size is the key factor driving customers to make the buying decision in the clothes-trial phase in physical stores (Eckman et al., 1990). Due to the limitations of online shopping, customers can only inspect colour and size after receiving their product. Once customers find colour distortion and size not to their satisfaction, they are more likely to return it. Hourigan et al. (2012) and Shephard et al. (2014) also found men and women have different shopping behaviours when facing product variety. However, there is scarce knowledge regarding the relationship between product variety and the product return fraction between women and men.

A fast-fashion company in the apparel industry offers almost 60 products to female and male customers. It is suffering from a product return fraction as high as 27% (given by company). Each returned product requires €11 for handling and restock, which is a heavy burden for the company (given by company). Therefore, the company is looking for an effective strategy to reduce the product return fraction. Current research regarding return management emphasizes operational and tactical perspectives, such as inventory management and logistics planning (Mollenkopf et al., 2011). Few academic studies have researched customer reasons for returns. Existing literature has found that product variety positively affects product return quantities, and number of products returned by each customer highly differs (Petersen & Kumar, 2009; Hess & Mayhew, 1997). However, insight into the relationship between product variety and the product return fraction is lacking. This presents a challenge to e-retailers because they must still provide various products types to retain customers, yet this may lead to high operational and return costs. Moreover, women and men are found to have different attitudes towards online shopping (Hourigan et al., 2012; Shephard et al., 2014). However, female and male return behaviours are given less attention.

(7)

7

consideration, it aids this fast-fashion apparel e-retailer in providing the most appropriate number of product types to maintain the highest level of profits without sacrificing customer service (Mehrojoo & Pasek, 2014).

Based on the concerns above, the research question is:

What is the relationship between product variety and the product return fraction in the apparel industry?

The rest of the paper is organized as follows: the second section is a theoretical review which describes each variable separately and also describes their interrelations. The third section discusses the methodology including the research setting, variable measurements, data collection and data analysis. In the fourth section, results are generated based on the data analysis and this is followed by a discussion. Finally, a conclusion and further research possibilities are proposed.

2. Literature Review

In order to give insights to the research question, we explain the definition of product return and product variety, and why product return management is important. Differences between women and men in their purchasing and return behaviour are also explained.

2.1 Product return

Product return consists of three types, environmental return, marketing return and consumer return (Rogers, Rogers & Lembke, 2010). The present paper focuses on consumer return, which is the phenomenon when “the product moves from end consumers to upstream suppliers” (Rogers et al, 2010). Customer returns normally occur in the post-purchase consumer buying behaviour process, and are considered to be a customer complaint behaviour (Blodgett et al., 1997).

(8)

8

business because product return is costly. Once the product is returned, the retailer has to transfer money back to customers and spend €5.5-€16.3 on average for handling and restocking per return, and totals a 3.8% loss of profit per year (The Economist, 2013; Ofek et al, 2011; Petersen and Kumar, 2010). High product return rates (25%-40%) within the apparel industry make the situation even worse. However, retailers have to accept these costs to retain customers (Chen and Bell, 2011).

Product returns result from product defects, different product performance compared to customer expectations, not fulfilling customer needs, and incomplete knowledge of product attributes perceived by customers (Ferguson et al., 2006; Guide et al., 2006; Blanchard, 2012, Rao et al., 2014, Lee, 2015). Lawton (2008) conducted an empirical study in online marketing and confirmed that the main reason for product returns is products not meeting customer expectations. Due to the limitations of online shopping, customers are not able to examine the physical product before purchasing; online product information therefore plays a key role in providing product attributes (Ofek et al; 2011). Although fast-fashion retailers are attempting to present detailed product information online, it is still difficult for customers to evaluate whether or not the product meets their preferences. Hong & Pavlou (2014) named the situation as product fit uncertainty, which is associated with experiential product characteristics (e.g. fit, size, colour). Once customers feel product attributes mismatch with their preference after receiving it, they will be dissatisfied and return the product (Hong & Pavlou, 2014).

2.2 Product variety

(9)

9

2.2.1 Product variety and product returns

Increasing product variety is important for the firm. According to Kekre & Srinivasan (1990), more product types result in a higher market share. Park et al. (2005) confirmed this finding and proposed that product variety not only raises market share but also increases customer satisfaction and enhances competitiveness. Product variety also affects customer buying behaviour. Kahn (1998) found that consumers are more satisfied when they face high product variety. Various product options allow customers to enjoy variety-seeking behaviour, which satisfies their intellectual curiosity (Kahn, 1998). When they are satisfied with their shopping process, they are willing to purchase more products (Collier & Bienstock, 2006). Adversely, reductions in product variety decrease both shopping frequency and order quantity (Borle et al., 2005). In other words, the sales decrease and the profits are significantly affected (Vaagen & Wallace, 2008).

Product variety motivates customers to purchase items, although there is a big difference between female and male customers in buying behaviour. Women are more fashion conscious, more addicted to fashion and new styles, and more involved in high fashion (Solomon and Schopler, 1982; Davis, 1994; O’Cass, 2004). Compared to women, male customers typically begin online shopping when they have a need. They are considered to be goal-focused shoppers because their shopping objective is clear (Phang et al., 2010). Male customers usually collect information of the product in advance and make quick purchase actions (Wolfinbarger & Gilly, 2010). Because women are fashion oriented whereas men are goal-focused, women tend to be more willing to make more purchases when faced with a greater variety of products (Summers, 1970; Phang, 2010).

(10)

10

2004). Consequently, women are more likely to return products. Since men collect information before they make online purchases, their sufficient knowledge and clear shopping objectives increase the probability of product satisfaction. Consequently, the possibility of returns is reduced for men (Maity & Arnold, 2013).

2.2.2 Style and product returns

Style reflects the fashion of apparel products and exerts an expressive function within the apparel industry from customers’ perspectives (Francis & Dickey, 1984). Offering various styles allows customers to meet their variety seeking behaviour (Kahn, 1998). Kim and Lennon (2002) also confirmed that the number of apparel products on offer is positively related to purchase intention. Styles are especially important to women because they are willing to seek information about popular colours and styles before purchasing. Since women are more involved in fashion and willing to try new styles, they are more likely to buy fashionable clothes when more options are available (Davis, 1994; O’Cass, 2004).

Apparel products are defined as high-risk items in online shopping “due to the sensory and interactive nature of the apparel purchase process”, unmet expectations with apparel products always occur (Hawes and Lumpkin, 1986; Bhatnagar et al., 2000). Customers are willing to try new styles, they do not know if the new style fits their body however (Hong & Pavlou, 2014). When they find the product does not fit, they are more likely to return products (Hong & Pavlou, 2014). In other words, the number of style on offer is positively related to the product return fraction. Women are more involved in fashion and try new styles and perceive a higher risk in online shopping than men, they are more likely to return product types due to a mismatch with the new style (Goldsmith et al., 1987; Stith & Goldsmith, 1989; Lee et al, 2002; Bakewell et al., 2006; Cho & Workman, 2011).

2.2.3 Colour and product return

(11)

11

Female and male customers show different attitudes on colour. Women are more sensitive to the colours of apparel products compared to men. The reason for this is female customers are more fashion conscious, have greater involvement in fashion and want to be fashion leaders (Goldsmith et al., 1987; Stith & Goldsmith, 1989; Bakewell et al., 2006; Cho & Workman, 2011). As a result, they are more likely to follow colour trends in fashion.

Colour is also related to product returns. Apparel products are considered as high-risky products to purchase due to the uncertain information online such as colour distortion (Bhatnagar et al., 2000; Sebastianelli et al., 2008). Once customers find colours mismatch with what they desired, they will return it. However, there is no literature found stating the correlation between the number of colours and the product return fraction. Since women are more sensitive to colour than men (Shim & Bickle, 1993), they are more likely to return products due to colour distortion.

2.2.4 Size and product return

Size is one of the main product attributes in the apparel industry. It is vital for customers as it reflects the fit of the clothes (Davis, 1985; Forsythe, 1991; Gipson & Francis, 1986; Hatch & Roberts, 1985; McLean, Roper & Smothers, 1986). It also has an impact on consumers’ buying behaviour. As Eckman et al. (1990) studied, when customers try on clothes in a physical store, size plays an important role in their purchasing decisions.

Female and male customers have different attitudes toward apparel size for different product types. Anderson et al. (1997) claimed that size is generally more crucial for female customers because they are more willing to show their body shape compared to male customers.

(12)

12

Bickle, 1993). Compared to women, men do not give as much consideration to fit when shopping for apparel products. Because of these qualities, women are more likely to return products once they think a product does not fit ((Garbarino & Strahilevitze, 2004).

2.3 Conceptual model:

Based on the above, a conceptual model is proposed in Figure 1. As indicated in the figure, the relationship between product variety and sales has been shown to be positively related (Kekre & Srinivasan, 1990). Sales are found to generate profits, which is common sense. Regarding the product return fraction, it has a negative impact on profits (Aberdeen, 2009). Hereby, we focus on the exploration of the relationship between product variety and the product return fraction as it has been scarcely investigated. This relationship is highlighted by a blue rectangle.

Figure 1 – Conceptual Model

Taking into account all aspects described above, three sub-questions are derived.

1. How does product variety affect the profit margin of the company?

The first sub-question shows why product variety is important for the company financially, which is proved mathematically in Chapter 4.

2. How are product variety and the product return fraction related? 3. How does gender play a role in return behaviour ?

(13)

13

3. Methodology

3.1 Research setting and data collection

To test the relationship between product variety and product return fraction, we employed exploratory mathematical modelling. Although previous studies have researched product variety in surveys (e.g. Shim & Bickle, 1993; Abraham-Murali & Littrell, 1995), those surveys are in relation to either brand variety or the fashion level of apparel products, which is not applicable in our research. Adopting an exploratory analysis allows us to analyse the data with great speed by inserting all known information into the analytical system (Bankes, 1993; Hodges, 1991). Exploratory mathematical analysis also provides visualized descriptive evidence to generate conclusions (TUDelft Documentation). However, mathematical analysis also has its limitations. The output of mathematical analysis highly depends on the input data. If the input data is not accurate, neither is the output.

The research was conducted in three phases. In the first phase, we presented how product variety affects profits quantitatively in Chapter 4. The formula includes price per item, return fraction relative to product variety, number of products sold relative to product variety and return cost per item. The second phase is associated with the relationship between product variety and the product return fraction. The third phase is on the return-behaviour differences between women and men. The second and third phases are based on the dataset given by the company. The dataset includes more than 95,000 data rows on style, colour and size. This large data sample ensures the significance of the results. This group of data was sorted and analysed. Thereafter, the filtered data were inserted into graphs with trend lines and R-squared value to test the trend line’s reliability.

3.2 Variables and measurements 3.2.1 Profit margin

(14)

14

product variety and the product return fraction may also be impacted by product variety (Kekre & Srinivasan, 1990). Thus, we took product variety into consideration.

Measurement:

Profit margin= Sales - Total expenses

3.2.2 Product return

Past studies have stated that product return rates should be measured by the relative frequency of products returned instead of total amount of return times and the individual return rate. Kang and Johnson (2009) showed that the total amount of returned products was incorrect. For instance, customers might return only two pieces of an eight-piece purchase. Based on “total amount” calculations, the return would include all items which is far from the actual case. Therefore, it is prudent to record the product return fraction by the number of returned products divided by total number of return products sold. Consequently, we used the percentage of products returned mathematically.

Measurement:

The product return fraction was measured by the percentage of products returned, which was achieved by dividing the number of products returned by product sold within each product type from September, 2014 to April, 2015. The percentage of returns within each product type was recorded.

3.2.3 Product variety

(15)

15

decisions when customers try on clothes in physical stores. Thus, we believe colour, style and size might have a different effect on the product return fractions. Moreover, we also examine how combinations of these elements affect the product return fraction. There are four combinations which we consider 1) style and colour (style*colour); 2) style and size (style*size); 3) colour and size (colour*size); 4) style, colour and size (style*colour*size).

Measurement:

Product variety could be measured as: 1) style*colour*size; 2) style*colour; 3) style*size; 4) colour*size; 5) style; 6) colour; 7) size. Measuring for each possibility is unique. Style*colour*size records the number of all products provided within each product type. Style*colour is calculated according to how many colours were offered regarding different styles within each product type. Style*size is calculated according to the number of sizes provided for style in each product type. Colour*size is calculated according to the number of available sizes for each colour in each product type. When taking style, colour and size separately, we recorded the number of styles, colours and sizes available in each product type, respectively.

3.3 Data analysis

Analysing the data is not necessary for the first sub-question as it presents a mathematical model. Thus, our data analysis starts with the data extraction from the dataset provided by the company for sub-questions two and three.

(16)

16

what there really is. Therefore, each style was calculated manually, and then added up. Regarding style*colour, style*size and colour*size, we multiplied style with colour, style with size and colour and size separately within each product type. Regarding style, colour and size, we recorded the number of style, colour and size provided within each product type. Afterwards, we placed each product variety of each product type in Excel with their corresponding product return fraction. Finally, seven scatterplot graphs were made with trend lines to examine the tendency presented on each graph. Trend lines present trends by linking several dots on the graph. In total, there are six types of trend lines: exponential, linear, logarithmic, polynomial, power and moving average, respectively. As patterns in the data were unknown, linear and polynomial trend lines were finally adopted based on their definitions. Linear trend lines are able to explore the straight trend of data points and polynomials are the most suitable for data with fluctuating values (TechAdvisory). In addition, with polynomial trend lines, there is a sub-option called “order” which shows the degree of the curve of trend line. For instance, order two has a relatively flat curve, while order six presents several obvious concave and convex areas. Gelman & Imbens (2014) proposed that high-order polynomials (orders greater than two) should be avoided. Mckenzie (2014) reviewed the paper written by Gelman & Imbens (2014) and repeated reasons why high-order polynomials are problematic. First, “they can give huge weight to points that are far away from the discontinuity”. Second, “Estimates can be highly sensitive to the degree of the polynomial fitted.” Third, high-order polynomials make confidence intervals too narrow, which results in more rejections of the null than usual (Mckenzie, 2014). Thus, polynomial regressions with low orders are much safer. We considered orders zero, one and two. However, orders zero and one were not possible—only order two worked. Furthermore, while making trend lines we also inserted the R-squared value of each trend line to examine its reliability. The trend lines with higher R-squared values are believed to be more convincing and reliable than the ones on the same graph.

(17)

17

style*colour*size. Then we multiplied style and colour, style and size, and colour and size. Concerning style, colour and size, we recorded the number of styles, colours and sizes offered in each product type. After placing each product variety with the its corresponding product return fraction, graphs for women’s and men’s return behaviours were made. We also adopted polynomial trend lines and linear trend lines to observe the trends, assuming the trend line has a reliable R-square. If the R-squared value of one trend line is higher than the other one on the same graph, the higher one is more trustful.

4. Results and discussion

4.1 Results

4.1.1 Product variety and profit margin

The objective of e-retailers is maximise profit levels resulting from product variety. Consequently, we employed the profit margin formula to show how product variety influences profit margin. In this formula, two components are involved, revenues and expenses. Revenues are defined as price per product times the number of products sold, and expenses included total return costs as other costs were unknown.

Profit Margin =Sales − Total Expenses_Sales

(Formula 1)

Formula 1 could be divided into two separate equations, which are:

Sales = 𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝 𝑝𝑝𝑝𝑝𝑝𝑝 𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝 ∗ 𝑛𝑛𝑝𝑝𝑛𝑛𝑛𝑛𝑝𝑝𝑝𝑝 𝑝𝑝𝑜𝑜 𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝 𝑝𝑝𝑝𝑝𝑠𝑠𝑝𝑝 (Formula 2) Total Expenses = 𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑛𝑛 𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝 𝑝𝑝𝑝𝑝𝑝𝑝 𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝 ∗ 𝑛𝑛𝑝𝑝𝑛𝑛𝑛𝑛𝑝𝑝𝑝𝑝 𝑝𝑝𝑜𝑜 𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝 𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑛𝑛𝑝𝑝𝑝𝑝 (Formula 3) We assume: - v is product variety

- s (v) is the number of products sold

- f (v) is the return fraction relative to product variety - α is the return cost per item

- p is the price per product

(18)

18

Sales = 𝑝𝑝 ∗ 𝑝𝑝 (𝑣𝑣)

(Transferred Formula 2)

Total Expenses = 𝛼𝛼 ∗ 𝑜𝑜 (𝑣𝑣) ∗ 𝑝𝑝 (𝑣𝑣)

(Transferred Formula 3)

Based on Transferred Formulas 2 and 3, Formula 1 can be expressed as: Profit Margin =𝑝𝑝 ∗ 𝑝𝑝 (𝑣𝑣) − 𝛼𝛼 ∗ 𝑜𝑜 (𝑣𝑣) ∗ 𝑝𝑝 (𝑣𝑣)_{𝑝𝑝 ∗ 𝑝𝑝 (𝑣𝑣)}

Since s(v) appears in both the numerator and the nominator, it is possible to eliminate it in the formula. Therefore,

Profit Margin =𝑝𝑝 − 𝛼𝛼 ∗ 𝑜𝑜 (𝑣𝑣) 𝑝𝑝

Price per product and return cost per product are fixed. Consequently, profit margin was only determined by return fraction resulted from product variety.

4.1.2 The relationship between product variety and the product return fraction

This section examines the relationship between product variety and the product return fraction, which gives solutions to the second sub-question. Based on the results, it is possible to point out which elements are the most critical factors influencing the product return fraction.

1) style*colour*size & the product return fraction

Graph 1 (in Appendix A) shows that both polynomial and linear trend lines in the chart are quite flat, which means product variety (considering style, colour and size) and return fraction are not correlated.

2) style*colour & the product return fraction

Graph 2 (in Appendix A) reveals an obvious downward trend and the highest point is almost 10% above the lowest point. Thus, we conclude that, when considering style and colour as the elements of product variety, it was negatively related to the product return fraction.

3) style*size & the product return fraction

(19)

19

4) colour*size & the product return fraction

Graph 4 (in Appendix A) reveals a flat trend which means this product variety and the product return fraction are not related.

5) style & the product return fraction

As we can see in Graph 5 (in Appendix A), style also had no effect on the product return fraction.

6) colour & the product return fraction

The trend line regarding colour (Graph 6 in Appendix A) shows a dramatic drop from 32% to 15%, which means that if more colour are offered, the product return fraction is lower.

7) size & the product return fraction

Graph 7

Graph 7 is associated with size and shows a concave polynomial trend line while the linear trend line has an upward tendency. Because the polynomial trend line has a higher R-squared value compared to the linear trend line, we believe the polynomial trend line is more accurate in representing the relationship. From the polynomial trend line, it is obvious that the return fraction rate is the lowest when only 1 size is available.

4.1.3 The relationship between product variety and the product return fraction between men and women

(20)

20

This section answers the third sub-question, whether women and men have different return behaviours when they face various apparel products. This test provides ways in which e-retailers could change their operational decisions.

1) style*colour*size & the product return fraction between men and women

Graph 8.1 and 8.2 (in Appendix B) demonstrate downward trends when considering style, colour and size together. However, Graph 8.1 shows slight differences from Graph 8.2, that the trend line on the men’s graph (Graph 8.1) is relatively flatter than women’s (Graph 8.2). Therefore, it could be concluded that men are less likely to return products regarding style, colour and size, compared to female customers.

2) style*colour & the product return fraction between men and women

When taking a look at style*colour, men and women have similar downward trends. The range between the highest point and the lowest point is around 14% on both graphs (Graph 9.1 and Graph 9.2 in Appendix B), which means there is no big differences between men and women in their return behaviour regarding style*colour.

3) style*size & the product return fraction between men and women

(21)

21

Graph 10.2

Product variety regarding style*size reveals a contrast on the polynomial trend lines for women and (linear trend lines were ignored due to lower R-squared value). As we can see from the polynomial trend lines on both graphs, it is clear that the graph for men is convex while female’s graph is concave. We conclude that men are less willing to return products when offered 250 products and women are less likely to return when 15 products are available with respects to style*size.

4) colour*size & the product return fraction between men and women

Graph 11.1 and 11.2 (in Appendix B) demonstrate slight downward trends on both polynomial trend lines. However, the trend line on the women’s graph is steeper meaning that the number of colour*size plays a relatively more significant role in women’s return behaviour compared to men.

(22)

22

5) style & the product return fraction between men and women

Graph 12.1

Graph 12.2

Concerning style, the men’s graph presents a gradual drop on both the linear and polynomial trend lines. Thus, we can summarize that the number of styles available negatively relates to the product return fraction for men. When it comes to the women’s graph, the polynomial trend line is concave, and women return the least amount of products when 10 product styles are on offer.

6) colour & the product return fraction between men and women

Graph 13.1 and 13.2 (in Appendix B) show that colour exerts a negative influence on both men’s and women’s return fraction. As we stated above, colour plays a key role in driving

(23)

23

customers to select and buy apparel products; over 60% of customers claim that satisfaction from apparel products is in part from colour (Eckman et al., 1990). Both graphs give evidence to this statement. However, there is a difference between women and men. As we can see from these graphs, both linear and polynomial trend lines on the women’s graph are slightly more steep than the male’s. This means the number of colours provided affects women more significantly than it does for men regarding return behaviour. However, both women and men show they are inclined to return fewer products when e-retailers offer 20 colours.

7) size & the product return fraction between men and women

(24)

24

Graph 14.1 and 14.2 demonstrates the how the product variety of size affects the product return fraction. The horizontal ranges between women and men highly differ, despite similar trends being shown on their polynomial trend lines. As we can see from graphs, men are offered more sizes than women and they are less likely to return products when there is only one size available. When it comes to the female’s graph, they are inclined to return fewer products when six sizes are on offer.

4.2 Discussion

As indicated in the results, the product return fraction of style remains constant. However, this finding is not consistent with the literature. Hong & Pavlou (2014) found out that customers love trying new styles, they are more likely to return products when they find “product fit uncertainty” after receiving products in online shopping. This contradictory result might be resulted from: 1) Different methodologies employed in both researches. Hong & Pavlou (2014) adopted self-reported assessment (survey) to measure the variable “product fit uncertainty” with the scale developed by Churchill (1979). However, our research was directly based on the company given by the company, which might result in different results. 2) Different products-oriented. Hong & Pavlou (2014) examined the “product fit uncertainty” of experienced goods in their research. Experienced good is defined as “whose utility cannot be ascertained before purchase” such as wine, cosmetics and apparel products. (Nelson 1970, 1974). However, we tested apparel products only in our research. This might be one reason that results are different. 3) Different research regions. In Hong & Pavlou’s (2014) study, data were obtained from two online markets, which were Taobao China (the biggest online-shopping website in China) and eBay U.S.A. However, our research was Dutch-market oriented. Due to the cultural differences, respondents might report their shopping habits differently. Thus, this issue is also considered as one of the reasons why results differ.

(25)

25

catalog shopping, which is quite different from our research group which is customers at all ages. Customers at different ages care about size in different ways, this might explain why findings differ.

5. Conclusion and further research

E-retailers provide various product types to attract customers and generate sales (Mehrojoo & Pasek, 2014; Kekre & Srinivasan, 1990). Although more products are returned in this type of business, they believe increased sales could cover return costs (Vaagen & Wallace, 2008; Petersen & Kumar, 2009). Research on the relationship between product variety and the product return fraction evaluates if there are greater profits from providing a larger number of products, and helps them make trade-offs between the benefits of product variety and product return costs.

In this research we proposed seven definitions of product variety by combining elements of style, colour and size; and found their impact on the product return fraction for each definition. We found that style*colour*size, colour*size and style have no effect on the product return fraction. Thus companies are advised to offer as many as styles to generate sales (Kekre & Srinivasan, 1990). Our research also reveals that style*colour and colour are negatively related to the product return fraction. It is reasonable for companies to offer more colours to attract customers, which is twenty colours in our case (Graph 6) (Mehrojoo & Pasek, 2014; Kekre & Srinivasan, 1990). Style*size is found the one that exerts a positive influence on the product return fraction. Additionally, size is also found to affect the product return fraction — customers are less willing to return products when there is only one size available. However, offering only one size is not practical in real business. It is necessary to extend the research to find the sub-optimal solution. For instance, companies can check out the second least-return point and the corresponding point of the number of colours. If companies do want to attract customers, it is also possible to find the equilibrium point between profits and return costs generated from colour, so that companies ensure their profits from style and size while make profit sacrifices from colour.

(26)

26

less willing to return products when regarding style*colour*size. Style*colour affects men more than women, while colour has the opposite effect. However, both women and men are less likely to return products when there are twenty colours are on offer. Thus, companies are advised to offer twenty colours to remain the lowest return fraction. Style*size and style on men and women have different effects on their respective return fractions. We found that style is negatively related to male’s return fraction and style exerts a concave tendency on women’s return fraction. Thus, companies could provide eighty styles to attract male customer and either ten or 100 styles for female customers. Size is found to influence men and women in the same way as well. The lowest return points are different, however. Female customers could be offered six sizes to keep the lowest return point. The trend of men’s return fraction is similar to the trend of total sizes analysis (Graph 7), thus companies could adopt the same advises stated in the previous paragraph to find the best solution. Apart from these advices, the detailed presentation of product information such as the fabric, colour and information of size helps to reduce the product return fraction, (Park et al., 2005). This eliminates the return possibilities of female customers in particular as women feel more uncertain about online shopping (Garbarino & Strahilevitz, 2004).

Before drawing the final conclusion, an important factor should be taken into consideration. Existing literature has stated that “a trend line is most reliable when its R-squared value is at or near 1”. However, R-squared values in our graphs are quite low with the highest being around 0.5. Consequently, the relationship between product variety and the product return fraction should be rejected. In other words, product variety is not related to the product return

fraction. The company is advised to offer as many products as they can to attract consumers as the product return fraction is constant.

(27)

27

purchase intention or the relationship between the quality of colour presentation online and the product return fraction (e.g. Bhatnagar et al., 2000; Ha & Stoel, 2004).

(28)

28

companies could adjust the number of products on offer seasonally in terms of style, colour and size.

6. References

Aberdeen, 2009, Reverse logistics: driving improved returns directly to the bottomline.

Available online at

http://www.aberdeen.com/aberdeen-library/6323/RA-reverse-logistics-return-refurbishment.aspx (accessed 11 March 2016).

Abraham - Murali, L., Littrell, M.A., 1995, Consumers’ Conceptualization of Apparel Attributes, Clothing and Textiles Research Journal, Vol 13(2), pp 65-74.

Bankes, S., 1993, Exploratory modeling for policy analysis, Operations Research, Vol 41(3), pp. 435-449.

Bhatnagar, A., Misra, S. and Rao, H.R. (2000), “On risk, convenience, and Internet shopping behavior”, Communications of the ACM, Vol. 43(11), pp. 98-105.

Bernon, M., Rossi, S., & Cullen, J. 2011, Retail reverse logistics: a call and grounding framework for research. International Journal of Physical Distribution &Logistics

Management, Vol 41(5), pp. 484-510.

Black, D. W. 2007. A review of compulsive buying disorder. World Psychiatry, Vol 6(1), pp. 14-18.

Blodgett, J., Hill, D., Tax, S., 1997, The effect of Distributive, Procedural, and Interactional Justice on Postcomplaint Behavior, Journal of Retailing, Vol 73(2), pp. 185-210.

Borle S, Boatwright P, Kadane JB, Nunes JC, Galit S. The Effect of Product Assortment

Changes on Customer Retention. Marketing Science; 2005. Vol 24(4), pp. 616-622.

(29)

29

Chen, J., and P. C. Bell., 2011. “Coordinating a Decentralized Supply Chain with Customer Returns and price-dependent Stochastic: Demand Using a Buyback Policy.” European

Journal of Operational Research, Vol 212 (2), pp. 293-300.

Chen, K., and Yen, DC., 2004, Improving the quality of online presence through interactivity,

Information & Management, Vol 42, pp.217-226.

Cho, S. and Workman, J., 2011, Gender, fashion innovativeness and opinion leadership, and need for touch: Effects on multi-channel choice and touch/non-touch preference in clothing shopping, Journal of Fashion Marketing and Management, Vol 15(3), pp. 363-382.

Churchill, G.A. Jr, 1979, A paradigm for developing better measures of marketing constructs.

J. Marketing Res. Vol 16(1), pp. 64–73.

Collier, J.E., Bienstock, C.C., Measuring Service Quality in E-Retailing, Journal of Service

Research, Vol 8(3), pp. 260-275.

Curves in all the wrong places: Gelman and Imbens on why not to use higher-order polynomials in RD, Submitted by David McKenzie On Mon, 09/08/2014

http://blogs.worldbank.org/impactevaluations/curves-all-wrong-places-gelman-and-imbens-why-not-use-higher-order-polynomials-rd

Davenport, K., Houston, J. E., & Griffiths, M. D. 2012, Excessive eating and compulsive buying behaviours in women: An empirical pilot study examining reward sensitivity, anxiety, impulsivity, self-esteem and social desirability. International Journal of Mental Health and

Addiction, Vol 10(4), pp. 474-489.

Dittmar, H., Long, K. and Meek, R., 2004, Buying on the Internet: Gender Differences in On-line and Conventional Buying Motivations, Sex Roles, Vol 50(5/6), pp. 423-444.

Eckman, M., Damhorst, M.L. and Kadolph, S.J., 1990, Toward a Model of the In-Store Purchase Decision Process: Consumer Use of Criteria for Evaluating Women’s Apparel.

(30)

30

Edmonds, T., McNair, F.M., Olds, P.R., Milam, E.E., Fundamental financial accounting concepts, 8th Edition.

Ferguson, M., Guide Jr, V. D. R., & Souza, G. C. 2006. Supply chain coordination for false failure returns. Manufacturing & Service Operations Management, Vol 8(4), pp. 376-393.

Francis, S. K., & Dickey, L. E., 1984, Dimensions of satisfaction with purchases of women’s dresses: Before and after garment care. Journal of Consumer Studies and Home Economics, Vol 8(2), pp. 153-168.

Garbarino, E., Strahilevitz, M., 2004, Gender differences in the perceived risk of buying online and the effects of receiving a site recommendation. Journal of Business Research, Vol 57(7), pp. 768–775

Gelman, A., Imbens, G., 2014, Why High-Order Polynomials Should Not be Used in Regression Discontinuity Designs. Link:

http://www.stat.columbia.edu/~gelman/research/unpublished/regression_discontinuity_14aug 02.pdf

Guide Jr, V. D. R., Souza, G. C., Van Wassenhove, L. N., & Blackburn, J. D. 2006. Time value of commercial product returns, Management Science, Vol 52(8), pp. 1200-1214.

Ha, Y. and Stoel, L., 2004, Internet apparel shopping behaviors: the influence of general innovativeness, International Journal of Retail & Distribution Management, Vol 32(8), pp. 377-385.

Hatch, K., Roberts, J., 1985, Use of intrinsic and extrinsic cues to assess textile product quality, Journal of Consumer Studies & Home Economics, Vol 9(4), pp. 341-357.

(31)

31

Hess, J.D., Chu, W., Gerstner, E., 1996, Control ling Product Returns in Direct Marketing,

Marketing Letters, Vol 7(4), pp. 307-317.

Hodges, J.S., 1991, Six (or so) things you can do with a bad model, Operations Research, Vol 39(3), pp. 355-365.

Hong, Y., Pavlou, P. A., 2014, Product fit uncertainty in online markets: Nature, effects, and antecedents. Information Systems Research, Vol 25(2), pp. 328–344.

Hourigan, S., Bougoure, U., 2012, Towards a better understanding of fashion clothing involvement, Australasian Marketing Journal, Vol 20, pp. 127–135.

Inman, J. J., Winer, R. S., Ferraro, R., 2009, The interplay among category characteristics, customer characteristics, and customer activities on in-store decision making. Journal of

Marketing, Vol 73(5), pp. 19-29.

Kahn, B. E., 1998, Dynamic Relationships with customers: High-Variety Strategies. Journal

of the Academy of Marketing Science, Vol 26(1), pp. 45-53.

Karlsson, C. (Ed.), 2009, Researching Operations Management, Routledge.

Kekre, S., Srinivasan, K., 1990, Broader product line: A necessity to achieve success?

Management Science, Vol 36 (10), pp. 1216–1231.

Kim, M. and Lennon, S.J. (2000), “Television shopping for apparel in the United States: effects of perceived amount of information on perceived risks and purchase intention”, Family

and Consumer Sciences Research Journal, Vol. 28(3), pp. 301-30.

Lawton, C. 2008. The war on returns. Wall Street Journal, Vol 8(1).

(32)

32

Lee, J., & Lund, M., 2003, Strategies to optimize return on investment (ROI) through effective reverse supply chain programs. In Electronics and the Environment, 2003. IEEE International Symposium (pp. 335-340). IEEE.

Lee, S-E, Kunz, G. I., Fiore, A. M., Campbell, J. R., 2002, Acceptance of mass customization of apparel: Merchandising issues associated with preference for product, process, and place,

Clothing and Textiles Research Journal, Vol 20(3), pp. 138-146.

Maity, D., & Arnold, T. 2013. Search: An Expense or an Experience? Exploring the Influence of Search on Product Return Intentions, Psychology & Marketing, Vol 30(7), pp. 576–587.

Mayhew, G.E., 1997, Modeling Merchandise Returns in Direct Marketing, Journal of Direct

Marketing, Vol 11(2), pp. 20-35.

McLean, F. P., Roper, L. L. and Smothers, R. (1986), Imported Versus Domestic Blouses: Women's Preferences and Purchase Motives, Home Economics Research Journal, Vol 4(3), pp.306-313.

Mehrjoo, M., Pasek, Z.J., 2014, Impact of Product Variety on Supply Chain in Fast Fashion Apparel Industry, Variety Management in Manufacturing. Proceedings of the 47th CIRP Conference on Manufacturing Systems, pp. 296 – 301.

Mollenkopf, D., Frankel, R., Russo, I., 2011. Creating value through returns management: Exploring the Marketing-Operations interface. Journal of Operations Management, Vol 29(5), pp. 391–403.

Nelson P (1970) Information and consumer behavior. J. Political Econom. Vol 78(2), pp. 311–329.

(33)

33

Ofek, E., Katona,Z., Sarvary, M., 2011, “Bricks and Clicks”: The Impact of Product Returns on the Strategies of Multichannel Retailers, Marketing Science, Vol. 30(1), pp. 42-60.

Park, T., Velicheti, KK., Kim, Y., 2005,The Impact Of Product Variety On Retailing Operations In The Supply Chain. California Journal of Operations Management III.

Petersen, A. J., & Kumar, V., 2009, Are Product Return a Necessary Evil? Antecedents and Consequences, Journal of Marketing, Vol 73, pp. 35-51.

Phang, C. W., Kankanhalli, A., Ramakrishnan, K., Raman, K. S., 2010, Customers’ preference of online store visit strategies: an investigation of demographic variables.

European Journal of Information Systems, Vol 19(3), pp. 344–358.

Rao, S., Rabinovich, E., Raju, D., 2014, The role of physical distribution services as determinants of product returns in Internet retailing. Journal of Operations Management, Vol 32(6), pp. 295–312.

Rogers, D. S., Rogers, Z. S., Lembke, R., 2010, Creating value through product stewardship and take-back. Sustainability Accounting, Management and Policy Journal, Vol 1(2), pp. 133-160.

Rogers, D., Tibben‐Lembke, R., 2001, An examination of reverse logistics practices. Journal

of Business Logistics, Vol 22(2), pp. 129–148.

Saleh, M. A. E.H., 2012, An Investigation of the Relationship between Unplanned Buying and Post-purchase Regret. International Journal of Marketing Studies, Vol4(4), pp.106.

Sebastianelli, R., Tamimi, N., and Rajan, M., 2008, Perceived Quality of Online Shopping: Does Gender Make a Difference? Journal of Internet Commerce, Vol 7(4), pp. 445-469.

Shephard, A., Kinley, T., Josiam, B., 2014, Fashion leadership, shopping enjoyment,and gender: Hispanic versus, Caucasian consumers' shopping preferences, Journal of Retailing

(34)

34

Shim. S., Bickle,M.C., 1993, Women 55 Years and Older As Catalog Shoppers: Satisfaction with Apparel Fit and Catalog Attributes., Clothing and Textiles Research Journal, Vol 11(4), pp. 53-64.

Shulman, J. D., Coughlan, A. T., & Savaskan, R. C. 2011. Managing consumer returns in a competitive environment, Management Science, Vol 57(2), pp. 347-362.

Slama, M. E., and Tashchian, A., 1985, Selected socioeconomic and demographic characteristics associated with purchasing involvement, Journal of Marketing, Vol 49(1), pp. 72-82.

Summers, J., 1970, The identity of women’s clothing fashion opinion leaders. Journal of

Marketing Research, Vol 7(2), pp. 178–185.

The Economist. 2013. Return to Santa - E-commerce firms have a hard core of costly, impossible-to-please customers. Available at:

http://www.economist.com/news/business/21591874-e-commerce-firms-have-hard-core-costly-impossible-please-customers-return-santa (Dec 21, 2013)

TUDelft documentation, Exploratory Modelling and Analysis (EMA) Workbench, Available at: http://simulation.tbm.tudelft.nl/ema-workbench/contents.html

Ulrich K, and Randall T., 2001, Product Variety, Supply Chain Structure, and Firm Performance: Analysis of the US Bicycle Industry, Management Science, Vol 47(12), pp. 1588-1604.

Vaagen, H., Wallace, S., 2008, Product variety arising from hedging in the fashion supply chains, Int. J. Production Economics, Vol 114, pp. 431–455.

Wolfinbarger, M., and Gilly, M. C., 2010, Shopping Online for Freedom, Control, and Fun.

(35)

35

7. Appendices

7.1 Appendix A- The relationship between product variety and the product return fraction

Graph 1 - style*colour*size & the product return fraction

Graph 2 - style*colour & the product return fraction

(36)

36

Graph 3 - style*size & the product return fraction

Graph 4 - colour*size & the product return fraction

(37)

37

Graph 5 - style & the product return fraction

Graph 6 - colour & the product return fraction

(38)

38

7.2 Appendix B- The relationship between product variety and the product return fraction between men and women

Graph 8.1 - style*colour*size & the product return fraction of men

Graph 8.2 - style*colour*size & the product return fraction of women

(39)

39

Graph 9.1 - style*colour & the product return fraction of men

Graph 9.2 - style*colour & the product return fraction of women

(40)

40

Graph 11.1 - colour*size & the product return fraction of men

Graph 11.2 - colour*size & the product return fraction of women

(41)

41

Graph 13.1- colour & the product return fraction of men

Graph 13.2 - colour & the product return fraction of women

Modelling the Relationship Between Product Variety and the Product Return Fraction