Group 8: Cash On Delivery
Introduction
- COD has helped e-commerce grow, but it does have pitfalls.
- COD can be expensive for the seller if the buyer returns a product, as the company pays two-way courier charges.
- COD accounts for almost 50% to 80% of online transactions in India
- But the return rates in online shopping in COD transactions averages around 40 per cent
Hence, we are using data analysis to see impact of COD on consumer behaviour, customer preferences and profits.
Reading data into data frame
COD_data=read.csv("PromotionDataV4.csv",header = TRUE)
attach(COD_data)
x = read.csv("PromotionDataV4.csv")
table(x$COD)
0 1
17754 28144
Data Exploration
View(COD_data)
summary(COD_data)
OrderItemCode OrderID OrderDate
Min. :47856780 Min. :27856296 4/10/2015 15:39: 28
1st Qu.:48822578 1st Qu.:28418744 4/10/2015 20:48: 25
Median :49663205 Median :28918355 4/10/2015 12:38: 22
Mean :49722785 Mean :28953309 4/10/2015 12:52: 21
3rd Qu.:50629792 3rd Qu.:29484854 4/10/2015 11:48: 20
Max. :51238880 Max. :29857723 4/10/2015 12:29: 20
(Other) :45762
Category SubCategory Brand
Apparels:45898 TOPS :14605 HARPA :7271
DRESSES : 8869 GRITSTONES :6901
POLO-T-SHIRTS : 3793 FABALLEY :6899
ROUND NECK T-SHIRTS: 2937 MONTEIL & MUNERO:6539
SHIRTS : 2897 MEIRA :5422
CASUAL SHIRTS : 1882 THE VANCA :4804
(Other) :10915 (Other) :8062
ProductColor ProductSize MRP FinalTotalPrice
BLACK :11357 M :12959 Min. : 299 Min. : 0.0
BLUE : 6048 L :11087 1st Qu.: 649 1st Qu.: 525.0
RED : 3570 S : 9556 Median : 950 Median : 678.0
NAVY BLUE: 3304 XL : 7350 Mean : 1015 Mean : 754.8
WHITE : 2605 XXL : 2199 3rd Qu.: 1295 3rd Qu.: 882.0
PINK : 2461 XS : 670 Max. :10999 Max. :6977.9
(Other) :16553 (Other): 2077
VATPercent VAT CODCharge VendorDiscount
Min. :5 Min. : 0.00 Min. : 0.0 Min. : 0.0
1st Qu.:5 1st Qu.: 24.95 1st Qu.: 0.0 1st Qu.: 0.0
Median :5 Median : 29.95 Median : 0.0 Median : 110.0
Mean :5 Mean : 35.29 Mean :13.7 Mean : 225.4
3rd Qu.:5 3rd Qu.: 41.85 3rd Qu.:49.0 3rd Qu.: 400.0
Max. :5 Max. :329.95 Max. :49.0 Max. :4400.0
WebsiteDiscountCode WebsiteDiscount HasVendorDiscount
NIL :29871 Min. : 0.00 Min. :0.0000
APP05 : 2686 1st Qu.: 0.00 1st Qu.:0.0000
STYLE30: 1699 Median : 0.00 Median :1.0000
WOW30 : 1605 Mean : 83.81 Mean :0.5058
SALE30 : 1299 3rd Qu.: 44.95 3rd Qu.:1.0000
RUSH30 : 1144 Max. :3149.00 Max. :1.0000
(Other): 7594
HasWebsiteDiscount CustomerID COD
Min. :0.0000 Min. : 197 Min. :0.0000
1st Qu.:0.0000 1st Qu.: 4427436 1st Qu.:0.0000
Median :0.0000 Median : 8963196 Median :1.0000
Mean :0.3492 Mean : 9014567 Mean :0.6132
3rd Qu.:1.0000 3rd Qu.:14821276 3rd Qu.:1.0000
Max. :1.0000 Max. :16007333 Max. :1.0000
ShippingName ShippingCity ShippingState
Divya : 25 Bangalore: 4904 MH : 8669
Priyanka : 25 New Delhi: 4585 KA : 5762
jol shishir vyas: 23 Mumbai : 3845 DL : 5392
Pooja : 23 Pune : 2557 UP : 3959
Neha : 21 Gurgaon : 2004 HR : 3003
Shital Shegokar : 21 Hyderabad: 1999 AP : 2840
(Other) :45760 (Other) :26004 (Other):16273
ShippingPincode ShippingAddressType BillingCity BillingState
Min. :110001 Home :20537 Bangalore: 4829 MH : 8668
1st Qu.:201012 Null :20455 New Delhi: 4707 KA : 5750
Median :400706 Office: 4906 Mumbai : 3867 DL : 5504
Mean :400189 Pune : 2513 UP : 3962
3rd Qu.:560060 Hyderabad: 1983 HR : 2916
Max. :854301 Gurgaon : 1927 AP : 2864
(Other) :26072 (Other):16234
BillingPincode BillingAddressType
Min. :110001 Home :16443
1st Qu.:201012 Null :25571
Median :400706 Office: 3884
Mean :400234
3rd Qu.:560059
Max. :900000
Reading data into data frame
COD_data$COD.f <- factor(COD,
levels=c(0,1),
labels=c("Electronic","Cod"))
boxplot(FinalTotalPrice~COD.f,data=COD_data, main="Order data", xlab="Order Price", ylab="Payment Method", horizontal = TRUE)
Data Exploration
library(gplots)
plotmeans(FinalTotalPrice ~ COD.f, data = COD_data, frame = FALSE)
Data Exploration
mytable <- xtabs(~ COD.f+ShippingAddressType, data=COD_data)
addmargins(prop.table(mytable, 2), 1)*100 # row sum
ShippingAddressType
COD.f Home Null Office
Electronic 47.21722 25.25544 58.92784
Cod 52.78278 74.74456 41.07216
Sum 100.00000 100.00000 100.00000
chisq.test(mytable)
Pearson's Chi-squared test
data: mytable
X-squared = 3033.3, df = 2, p-value < 2.2e-16
- P<0.05, Hence, we reject the null hypothesis that Payment method and address type are independent.
- Here we can say that users with address type office can be inclined towards Electronic payment
- This insight can be used for targetted promotion based on payment methods
Price wise comparision
cod_percent = c()
non_cod_percent = c()
for(i in seq(500, max(x$FinalTotalPrice), by=500)) {
cod_percent = c(cod_percent, nrow(subset(x, x$COD == 1 & x$FinalTotalPrice > i))*100/nrow(subset(x, x$COD == 1)))
non_cod_percent = c(non_cod_percent, nrow(subset(x, x$COD == 0 & x$FinalTotalPrice > i))*100/nrow(subset(x, x$COD == 0)))
i = i + 500
}
We divided the purchases into bins of 500 and found out the percent of purchases.
Compare average order QTY for COD to that with Electronic
Null Hypothesis: Mean number of items for an order is similar for COD and Non COD transactions Alternate Hypothesis: Mean number of items for an order is not similar for COD and Non COD transactions
Test Statistics: T-test Result of T-test: t = 1.7493, df = 28164, p-value = 0.08025
Compare_QTY = aggregate(OrderItemCode, by=list(ID=OrderID, PAY=COD), length)
Compare_QTY$PAY <- as.factor(Compare_QTY$PAY)
t.test(Compare_QTY$x~ Compare_QTY$PAY,data = Compare_QTY)
Welch Two Sample t-test
data: Compare_QTY$x by Compare_QTY$PAY
t = 1.7493, df = 28164, p-value = 0.08025
alternative hypothesis: true difference in means is not equal to 0
95 percent confidence interval:
-0.001684438 0.029644012
sample estimates:
mean in group 0 mean in group 1
1.323247 1.309267
Compare average price per order for COD to that with Electronic
Null Hypothesis: Mean price for an order is similar for COD and Non COD transactions Alternate Hypothesis: Mean price for an order is not similar for COD and Non COD transactions
Test Statistics: T-test Result of T-test: t = -8.5188, df = 30373, p-value < 2.2e-16
Hence, we reject the null hypothesis.
Compare_price = aggregate(FinalTotalPrice, by=list(ID=OrderID, PAY=COD), sum)
Compare_QTY$PAY <- as.factor(Compare_QTY$PAY)
t.test(Compare_price$x~ Compare_price$PAY,data = Compare_price)
Welch Two Sample t-test
data: Compare_price$x by Compare_price$PAY
t = -8.5188, df = 30373, p-value < 2.2e-16
alternative hypothesis: true difference in means is not equal to 0
95 percent confidence interval:
-70.41672 -44.07408
sample estimates:
mean in group 0 mean in group 1
957.0355 1014.2809
COD dependence
Logistic Regression to find effect of shipping state, final total price and brand on the dependant variable COD/Electronic
model <- glm (COD ~ ShippingState + FinalTotalPrice+ Brand , data = COD_data)
summary(model)
Call:
glm(formula = COD ~ ShippingState + FinalTotalPrice + Brand,
data = COD_data)
Deviance Residuals:
Min 1Q Median 3Q Max
-1.1916 -0.5319 0.2473 0.4158 1.0273
Coefficients:
Estimate Std. Error t value Pr(>|t|)
(Intercept) 1.069e-01 1.114e-01 0.960 0.337157
ShippingStateAP 3.798e-01 1.114e-01 3.410 0.000649 ***
ShippingStateAR 7.769e-01 4.838e-01 1.606 0.108338
ShippingStateAS 6.309e-01 1.122e-01 5.624 1.88e-08 ***
ShippingStateBR 6.658e-01 1.134e-01 5.873 4.30e-09 ***
ShippingStateCH 4.818e-01 1.128e-01 4.270 1.96e-05 ***
ShippingStateCT 6.257e-01 1.138e-01 5.500 3.83e-08 ***
ShippingStateDD 8.219e-01 1.923e-01 4.275 1.92e-05 ***
ShippingStateDL 4.014e-01 1.112e-01 3.609 0.000308 ***
ShippingStateDN 6.104e-01 1.482e-01 4.118 3.83e-05 ***
ShippingStateGA 6.535e-01 1.129e-01 5.786 7.25e-09 ***
ShippingStateGJ 5.598e-01 1.115e-01 5.018 5.23e-07 ***
ShippingStateHP 5.337e-01 1.144e-01 4.665 3.10e-06 ***
ShippingStateHR 3.614e-01 1.114e-01 3.245 0.001174 **
ShippingStateJH 6.016e-01 1.128e-01 5.333 9.71e-08 ***
ShippingStateJK 7.177e-01 1.125e-01 6.378 1.81e-10 ***
ShippingStateKA 3.238e-01 1.112e-01 2.912 0.003596 **
ShippingStateKL -1.175e-01 1.137e-01 -1.033 0.301544
ShippingStateMH 3.913e-01 1.111e-01 3.521 0.000431 ***
ShippingStateML 6.775e-01 1.146e-01 5.911 3.42e-09 ***
ShippingStateMP 5.856e-01 1.119e-01 5.234 1.67e-07 ***
ShippingStateMZ 7.755e-01 1.190e-01 6.518 7.18e-11 ***
ShippingStateOR 5.335e-01 1.119e-01 4.768 1.87e-06 ***
ShippingStatePB 5.948e-01 1.120e-01 5.313 1.08e-07 ***
ShippingStatePY 5.800e-01 1.388e-01 4.180 2.92e-05 ***
ShippingStateRJ 5.529e-01 1.118e-01 4.944 7.67e-07 ***
ShippingStateSK -1.678e-01 1.497e-01 -1.121 0.262251
ShippingStateTN 3.524e-01 1.116e-01 3.158 0.001592 **
ShippingStateTR 7.062e-01 1.285e-01 5.496 3.91e-08 ***
ShippingStateUP 4.697e-01 1.113e-01 4.221 2.44e-05 ***
ShippingStateUT 5.219e-01 1.129e-01 4.621 3.82e-06 ***
ShippingStateWB 5.169e-01 1.114e-01 4.638 3.52e-06 ***
FinalTotalPrice 1.343e-04 7.697e-06 17.453 < 2e-16 ***
BrandFABALLEY -7.890e-02 9.278e-03 -8.504 < 2e-16 ***
BrandGRITSTONES 1.496e-03 9.689e-03 0.154 0.877312
BrandHARPA -7.323e-02 9.262e-03 -7.907 2.70e-15 ***
BrandMEIRA -1.538e-02 1.022e-02 -1.506 0.132117
BrandMISS CHASE 1.791e-02 1.366e-02 1.312 0.189569
BrandMONTEIL & MUNERO -5.718e-03 9.661e-03 -0.592 0.553956
BrandMR BUTTON -1.600e-01 1.880e-02 -8.507 < 2e-16 ***
BrandTHE VANCA -4.430e-02 1.015e-02 -4.364 1.28e-05 ***
BrandTSHIRT COMPANY -4.379e-02 1.465e-02 -2.990 0.002793 **
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
(Dispersion parameter for gaussian family taken to be 0.2216896)
Null deviance: 10886 on 45897 degrees of freedom
Residual deviance: 10166 on 45856 degrees of freedom
AIC: 61153
Number of Fisher Scoring iterations: 2
Plots (in %)
plot(cod_percent, type = "l")
Plots (in %)
plot(non_cod_percent, type = "l")
As seen from both the graphs, no significant change in purchase behaviour
City Wise Trend
city <- aggregate ( x$OrderID ~ x$BillingCity+x$COD, x, length)
head(city[order((city$`x$OrderID`), decreasing = TRUE),],15)
x$BillingCity x$COD x$OrderID
1143 New Delhi 1 2645
61 Bangalore 0 2638
1110 Mumbai 1 2229
712 Bangalore 1 2191
455 New Delhi 0 2062
422 Mumbai 0 1638
498 Pune 0 1348
1045 Kolkata 1 1247
1199 Pune 1 1165
253 Gurgaon 0 1027
284 Hyderabad 0 992
953 Hyderabad 1 991
924 Gurgaon 1 900
461 Noida 0 591
359 Kolkata 0 561
Here we see city wise distribution of top cities
State Wise Trend
state <- aggregate ( x$OrderID ~ x$BillingState+x$COD, x, length)
head(state[order((state$`x$OrderID`), decreasing = TRUE),],15)
x$BillingState x$COD x$OrderID
50 MH 1 4903
17 MH 0 3765
40 DL 1 3158
15 KA 0 2902
48 KA 1 2848
61 UP 1 2555
7 DL 0 2346
63 WB 1 1636
34 AP 1 1602
45 HR 1 1543
30 UP 0 1407
43 GJ 1 1379
12 HR 0 1373
2 AP 0 1262
59 TN 1 907
Here we see state wise distribution of top states
Vebdor Discount wise distribution
vendor_discount <- aggregate ( x$OrderID ~ x$HasVendorDiscount+x$COD, x, length)
vendor_discount[order((vendor_discount$`x$OrderID`), decreasing = TRUE),]
x$HasVendorDiscount x$COD x$OrderID
4 1 1 14227
3 0 1 13917
2 1 0 8989
1 0 0 8765
Website Discount wise distribution
website_discount <- aggregate ( x$OrderID ~ x$HasWebsiteDiscount+x$COD, x, length)
website_discount[order((website_discount$`x$OrderID`), decreasing = TRUE),]
x$HasWebsiteDiscount x$COD x$OrderID
3 0 1 19229
1 0 0 10642
4 1 1 8915
2 1 0 7112
Any Discount wise distribution
x$any_discount <- ifelse(x$HasVendorDiscount | x$HasWebsiteDiscount, 1, 0)
any_discount <- aggregate ( x$OrderID ~ x$any_discount+x$COD, x, length)
any_discount[order((any_discount$`x$OrderID`), decreasing = TRUE),]
x$any_discount x$COD x$OrderID
4 1 1 19900
2 1 0 13462
3 0 1 8244
1 0 0 4292
Based on COD Charges
cod_charge <- aggregate ( x$OrderID ~ x$CODCharge+x$COD, x, length)
cod_charge[order((cod_charge$`x$OrderID`), decreasing = TRUE),]
x$CODCharge x$COD x$OrderID
1 0 0 17754
2 0 1 15307
3 49 1 12837