The overall number of data in this analysis data set is 56,904 including 51,949 non-return orders and 4,955 returns. We define the target variable as 0 (non-return) and 1(return). The average return rate in this data set is 8.71% (4,955/56,904).
Table 4.3 Target Variable
Target Non-return Return Return Rate Total Total 51949
(100.0%)
4955
(100.0%) 8.71% 56904 (100.0%) 4.5.1 Categorical Data
Analysis of Cross Tabulation is conducted here to reveal general information of categorical data.
1. Gender
Female and male accounted for 30,535 (53.7%) and 26,369 (46.3%) of orders respectively. Orders from female customers are in the majority with slightly higher return rate in this data set.
Table 4.4 Gender / Return Cross Tabulation
Gender Non-return Return Return Rate Total Female 27,865
Taipei County (23.6%), Keelung City (15.6%), Tauyuan County (9.1%) and Taipei City (9.1%) are the main origins of orders (approximately 57%) in the whole data.
In addition, there is a higher return propensity of customers from Chiayi City (10.89%), Hsinchu County (9.96%), Tainan County (9.69%). By contrast, return propensity of customers in Off-shore Islands (Penghu County, Kinmen
30
County) is relatively lower.
Table 4.5 Location / Return Cross Tabulation
Code County Non-return Return Return Rate Total 01 Keelung City 8,049 09 Taichung County 2,118
(4.1%)
Table 4.5(Con.) Location / Return Cross Tabulation
Code County Non-return Return Return Rate Total 17 Kaohsiung City 2,882
(5.5%)
271
(5.5%) 8.59% 3,153 (5.5%) 18 Kaohsiung County 1,279
(2.5%) 3. Merchandise Category
There are 32 kinds of merchandise in our data set. The most popular merchandises in the data set are: “Maternal and Child”, “Beauty” and “Food and Speciality” with the order ratios of 10.8%, 10.1%, and 8.1% respectively.
“Female Clothing”, “Female Shoes”, “Watches and Clocks”, “Female Bags” and “Peripheral” have the highest return rate. It is obviously that merchandise related to female dressing needs a further observation.
32
Table 4.6 Merchandise Category / Return Cross Tabulation
Code Merchandise Category Non-return Return Return Rate Total FC Female Clothing 3,042
P Peripheral 3,576
(6.9%) MC Maternal and Children’s 5,472
(10.5%)
679
(13.7%) 11.04% 6,151 (10.8%)
AP Appliances 2,062
(4.0%)
CMC Communication 2,370
(4.6%) FBD Furniture and Bedding 166
(0.3%)
F Furnishing 4,027
(7.8%)
Table 4.6(Con.) Merchandise Category / Return Cross Tabulation
Code Merchandise Category Non-return Return Return Rate Total VT Video Games and Toys 468
AM Automotive 945
(1.8%) CF Collections and Fine Works 429
(0.8%) CE Computer Expendables 920
(1.8%)
34
(0.7%) 3.56% 954 (1.7%) HP Health and Personal Care 1,660
(3.2%)
34
4. Payment
Most of the transactions are paid by credit card (87.0%) with lowest return rate. There is a relatively higher return rate (12.27%) with transactions paid by installments while comparing to the others. Table 4.7 shows the difference of return rate for three kinds of payments.
Table 4.7 Payment / Return Cross Tabulation Payment Non-return Return Return Rate Total
DIV 3,760 5. Delivery Approach
Since the retail delivery is a new approach for customers to choose, there is an obviously unbalanced ratio between the two alternatives. However, a significant higher return rate belonged to retail delivery is worth for a further analysis.
Table 4.8 Delivery Approach / Return Cross Tabulation Delivery Approach Non-return Return Return Rate Total CVS-Retail Delivery 209
(0.4%)
32
(0.6%) 13.28% 241 (0.4%) TPS-Home Delivery 51,740
(99.6%)
Approximately 67.8% of the data are paid by customers, and the remaining are relatively distributed across free-carriage (14.6%) and conditional payment (17.7%).
There is the highest return rate while the carriage is paid by conditions (e.g., the retailer will afford the carriage if the amount of data is higher than NT
$1,000).
Table 4.9 Carriage / Return Cross Tabulation
Carriage Non-return Return Return Rate Total Conditional 8,908
(17.1%)
1,159
(23.4%) 11.51% 10,067 (17.7%) Pay by Customers 35,376
(68.1%)
3,178
(64.1%) 8.24% 38,554 (67.8%) Pay by Retailers 7,665
(14.8%) 7. Actual Delivery Days
The frequency of Actual Delivery Days spreads from “0” day to “268”
days. Moreover, the frequency more than “5 days” is much fewer than the first five levels (0, 1, 2, 3, and 4). Further, to separate the delivery fewer than and more than 5 workdays, we decide to discretize the variable to six levels-0, 1, 2, 3, 4, and more than 5.
In the Table 4.10, there is an obvious phenomenon that the return rate becomes higher as the Actual Delivery days increases. Actual Delivery days equals 2 days deserve almost the same return rate as average, while Actual Delivery more than 5 days has 1.5 times higher return rate compared to average.
Table 4.10 Actual Delivery Days / Return Cross Tabulation Actual Delivery
Days Non-return Return Return Rate Total
0 15084
36
Table 4.10(Con.) Actual Delivery Days / Return Cross Tabulation Actual Delivery
Days Non-return Return Return Rate Total
4 3583 8. Average Delivery Days
The frequency of Average Delivery Days spreads from “0” day to “14”
days. In the same way, the frequency more than “5 days” is much fewer than the first five levels (0, 1, 2, 3, and 4). To compare with “Actual Delivery Days”, therefore, we discretize the variable to six levels-0, 1, 2, 3, 4, and more than 5 as last variable.
Orders delivered in 1 or 2 days occupy the major percentage of Average Delivery Days (more than 70%), while only a few orders supplied by some specific retailers are delivered more than 5 days (1.86%).
Table 4.11 Average Delivery Days / Return Cross Tabulation Ave. Delivery
Days Non-return Return Return Rate Total
0 3,728
4.5.2 Continuous Data
Analysis of Descriptive Statistics is conducted here to reveal general information of continuous data.
1. Age
The average age of customers is 35 years old. The upper bound of age is 78 and the lower bound is 10.
Distribution of Age data shown in figure 4.4 reveals that people at 27 to 40 years old are the main customers who accounted for approximately 60% of orders.
The average price of each data is NT $1,724. The maximum is NT
$180,000 and the minimum is NT $1.
The lowest price among the items is NT $1, which is resulted from trail or promotion. Customers only need to pay for NT $1 plus the carriage.
3. Accumulated Number of Buyers
This variable represents the established reputation of each retailer. The accumulated number of buyers of each retailer is 1,590. The maximum is 13,611 and the minimum is 1.
38
4. Accumulated Number of Browsers
The average accumulated number of browsers of each retailer is 132,256.
The maximum is 12,656,670 and the minimum is 11.
5. Total Number of Merchandise
The average number of merchandise of each retailer is 713. The maximum is 15,344 and the minimum is 0.