ICT706-Data Analytics | Case Study

You are required to write a report for a business case study. The case study requires you to investigate solutions for key business problems. You have to assume that you are performing a data scientist role in a commercial technology innovation organisation. You are required to use classification techniques to provide decision support and recommendations to the executives in the organisation.

Introduction – What is the problem?

Today most of the world population preferred the online shopping which results into the exponential growth in eCommerce. eCommerce becoming very raising and popular business in every corner of the world. Online shopping is most popular from the all eCommerce business.  In Today’s market we can see the words like Amazon, eBay, etc. which very common in day to day life. Service provider attract the customer by giving offers to the customers.

Growth in the online shopping creates the opportunity and challenges to the service provider. The main challenges in the online shopping is competition and customer total satisfaction.

About Data:

We have data sets regarding to Cloths (Shirt, Trouser and Track Suit) for the 2000 products. We considered the following attributes

1. Product Name
2. Product Price (in $)

• Sale Price (in $)

1. Profit (in $)
2. Number of customers
3. Shipping Type (Free or Paid)

• Customer Type (New or Existing)
• Region (QLD, WA, VIC, TAS)

1. Product Brand
2. Product Material

We define following variables for our analysis from the above variables

Total Monthly sale amount (in $) = Sale Price (in $) × Number of customers

ign: justify;">Total monthly profit (in $) = Profit (in $) × Number of customers

Project Problem:

We are interested to know the following things

1. Profit analysis by shipping type, customer type, region, brand and material.
2. Whether there is any association between shipping type, customer type, region, brand and material.

• Whether the number of customers is significantly different for shipping type, customer type, region, brand and material.

1. Correlation analysis of variables

Research Methodology

Any data analysis is stronger if it is supported by statistical tools and techniques. We summarised the total sale amount and total profit in profit analysis. We carried the chi-squared test for association of attributes (shipping type, customer type, region, brand and material). We analyzed the mean of number of customers for shipping type, customer type, region, brand and material by the two sample t test and one way ANOVA. We also carried correlation analysis. We used Python and MS-Excel for data analysis. We used Grus (2015), McKinney (2012), Pedregosa et al. (2011) and Schutt and O'Neil (2013).

Analytical Findings

Profit Analysis:

In profit analysis, we have reported the total sales amount, total profit, profit percentage for the shipping type, customer type, region, brand and material.  We referred Berenson et al. (2012), Black (2009), Groebner et al. (2008), Kvanli et al. (2000) and Mendenhall and Sincich (1993).   Table 1 summarizes the profit analysis according to shipping type, customer type, region, brand and material.

 Table 1: Profit analysis according to different attributes

Attributes	Shipping Type	Sum of Total Sale	Sum of Total Profit	Profit %
Shipping Type	Free	661525	88664	13.40%
	Paid	1276094	171178	13.41%
Customer Type	Existing	879349	117241	13.33%
	New	1058270	142601	13.47%
Region	QLD	370742	49741	13.42%
	TAS	204055	27225	13.34%
	VIC	792802	106086	13.38%
	WA	570021	76790	13.47%
Brand	Adidas	387014	45529	11.76%
	Arrow	393305	60874	15.48%
	Nike	371811	46476	12.50%
	Polo	377941	52755	13.96%
	Woodland	407548	54209	13.30%
Material	Cotton	709179	94216	13.29%
	Other	601860	81071	13.47%
	Wool	626580	84555	13.49%
Total		1937620	259842	13.41%

Company getting 13.41% profit overall. There is significant changes in the profit by different attributes (expect brand). In the brand, company gets 15.48% profit in Arrow whereas earned only 11.76 % profit earned on the Adidas. Total sales amount is bigger for paid shipping type. Total sales revenue is comparatively low for existing customer and customers from TAS region.

Descriptive statistics:

We used the well-known books for this section such as Bickel  and Doksum (2015), Casella and Berger (2002), DeGroot and Schervish (2012), Hodges Jr and Lehmann (2005), Papoulis (1990), Pillers (2002) and Ross (2014).

Profit is mainly depend on the number of customers who bought the product. In this section, we reported the summary statistics for number of customer for the shipping type, customer type, region, brand and material. In Table 2, we have given sample size, mean, standard deviation, minimum and maximum for number of customers for shipping type, customer type, region, brand and material. We can observed that averagely there is 37.99 customer for each products with standard deviation 6.61.We can observed one interesting thing is that average number of customers for free shipping is more than paid shipping.

Table 2: Summary statistics for number of customers

Attributes	Level	Size	Mean	SD	Min	Max
Shipping Type	Free	636	41.09	6.69	24	59
	Paid	1364	36.54	6.06	15	56
Customer Type	Existing	908	37.97	6.47	15	57
	New	1092	38.00	6.73	19	59
Region	QLD	400	36.64	6.11	22	56
	TAS	216	37.32	6.63	19	55
	VIC	779	39.58	6.53	20	59
	WA	605	37.07	6.63	15	58
Brand	Adidas	402	37.96	6.50	19	55
	Arrow	401	38.01	6.88	20	59
	Nike	384	38.28	6.83	15	56
	Polo	391	37.81	6.45	21	56
	Woodland	422	37.90	6.43	22	56
Material	Cotton	665	38.08	6.72	15	59
	Other	662	37.91	6.33	21	56
	Wool	673	37.98	6.78	22	58
Total		2000	37.99	6.61	15	59

Chi-square test for association:

In this section we test whether there is any association between attributes or not. Following Table 3 shows the chi-square statistic and p-value for chi-square test of testing association. In this test we have null hypothesis that there is no significant association between two attributes and alternative hypothesis is there is significant association between two attributes. We test the association between following pair of attributes

1. shipping type and customer type.
2. shipping type and region.

• shipping type and brand.

1. shipping type and material.
2. customer type and region.
3. customer type and brand.

• customer type and material.
• region and brand.

1. region and material.
2. brand and material.

Table 3: Chi-squared test for association

Pairs of attributes	Chi-Square Statistic	P-Value
shipping type and customer type.	0.005	0.943
shipping type and region.	0.022	0.999
shipping type and brand.	2.417	0.660
shipping type and material.	8.489	0.014
customer type and region.	1.586	0.663
customer type and brand.	3.550	0.470
customer type and material.	2.125	0.346
region and brand.	5.803	0.926
region and material.	10.034	0.123
brand and material.	6.133	0.632

We observed that only sipping type and material have significant association at 5% level of significance whereas all other pairs are not associated.

Two Sample t-test:

We used independent two sample t test for testing the difference between mean numbers of customer for shipping type (free and paid) and customer type (new and existing) . For shipping type, we test the null hypothesis that the mean numbers of customer who got free shipping is same as mean number of customers who got paid shipping. For customer type, we test the null hypothesis that the mean numbers of customer who are new is same as existing customers. Following Table 4, shows the result of two sample t test for shipping type and customer type. Table 4 includes test statistics, degrees of freedom and p value.

Table 4: Two sample independent test for shipping type and customer type

Attributes	Levels	Test Statistic	P value
Shipping Type	Free and Paid	14.58	0.000
Customer Type	New and Existing	-0.13	0.899

From Table 4, we can see that P-value of shipping type is less than 5% suggest that there is significant difference between the mean numbers of customer who got free shipping and got paid shipping.  As if the shipping is free, customer got interested in buying the item than the items for which shipping is paid. There is no significant difference between mean numbers of new customer and existing customer.

One way ANOVA:

In this section we test whether there is any significant difference between means of different level of

1. Region (QLD, TAS, VIC and WA)
2. Brand (Adidas, Arrow, Nike, Polo and Woodland)

• Material (Cotton, Wool and other)

We used one way ANOVA for this purpose. Here our null hypothesis that the different levels has same mean and alternative hypothesis is that at least one level has different mean. Following Table 5 shows the output of one way ANOVA for region, brand and material

Table 5: Output of one way ANOVA for region, brand and material

Attributes	Level	F Statistic	P Value
Region	QLD, TAS, VIC and WA	26.13	0.000
Brand	Adidas, Arrow, Nike, Polo and Woodland	0.28	0.889
Material	Cotton, Wool and other	0.12	0.891

From Table 5, we conclude that there is significant difference between mean numbers of customer in different region. We can see that VIC has most numbers of customer compared to the other region. We claim that there is no significant difference between mean numbers of customer for brand and material.

Correlation Analysis:

In this section we calculate the correlation coefficient for studying the association between the variables like Product Price, Profit and Numbers of customer. Table 6 represents the Pearson’s correlation matrix for the Product Price, Profit and Numbers of customer.

Table 6: Pearson’s correlation coefficient a for Product Price, Profit and Numbers of customer

	Product Price	Profit	Number of customers
Product Price	1	0.43	0.001
Profit	0.43	1	-0.003
Number of customers	0.001	-0.003	1

From the correlation analysis, we can say that product price and numbers of customer moderate positive correlation where as there is negligible correlation (zero correlation) between product price and numbers of customer, profit and numbers of customer.

Regression Analysis:

We used simple linear regression model for predicting the monthly sale using number of customers as predictor variable. Table 7 represents the F Statistics, P value, R² and regression coefficients of simple linear regression.

Table 8: Output of Regression Analysis

F Statistic	7462.109
P Value	0.000
R2	0.789
Intercept	-3.535
Slope	25.596

We observed that P Value =0.000 suggests that there is significant relationship between total monthly sale and number of customers who bought the laptops. We also observed R² as 0.789 suggests that model fitting is good and adequate. We fitted the following straight line as

Total sale (in $) = -3.535 + 25.596  × Number of Customers

Recommendations to the company

15. In the brand, company gets 15.48% profit in Arrow whereas earned only 11.76 % profit earned on the Adidas. So company needs to marketing strategies for Arrow brand which gives you more profit.
16. We observed that there is significant difference between mean numbers of customer who got free delivery and who paid for delivery. Mean number of customers who got free delivery is more than who paid for delivery. So company should provide free shipping so that customer gets attracted to buy.

• We observed that there is significant difference in the mean of numbers of customer for different region. We observed that VIC has most numbers of customer compared to the other region. So company should concentrate on attracting the customers from other region.

An implementation plan based on the recommendations you have provided

1. Company should appoint more employee in shipping department so that many products got free shipping.
2. Company can adopt the marketing strategies which are in functioning in VIC region so that number of customers in other region increases.

• Company can attract the customer by offering them some offers like free delivery, cash back, etc.

Conclusions

Company getting 13.41% profit overall. There is no significant changes in the profit by different attributes (expect brand). In the brand, company gets 15.48% profit in Arrow whereas earned only 11.76 % profit earned on the Adidas. Total sales amount is bigger for paid shipping type. Total sales revenue is comparatively low for existing customers and customers from TAS region.

We can observed that averagely there is 37.99 customers for each products with standard deviation 6.61.We can observed one interesting thing is that average number of customers for free shipping is more than paid shipping.

We observed that only shipping type and material have significant association at 5% level of significance whereas all other pairs are not associated.

There is significant difference between the mean numbers of customer who got free shipping and got paid shipping.  As if the shipping is free, customer got interested in buying the item than the items for which shipping is paid. There is no significant difference between mean number of new customers and existing customers.

There is significant difference between mean numbers of customer in different region. We can see that VIC has most numbers of customer compared to the other region. We claim that there is no significant difference between mean number of customer for brand and material. We observed that there is significant relationship between number of customers and total sale. We have also provided recommendation and plan to the company.

List of References

Berenson, M., Levine, D., Szabat, K.A. and Krehbiel, T.C., (2012). Basic business statistics: Concepts and applications. Pearson higher education AU.

Bickel, P.J. and Doksum, K.A., (2015). Mathematical statistics: basic ideas and selected topics, volume I (Vol. 117). CRC Press.

Black, K., (2009). Business statistics: Contemporary decision making. John Wiley & Sons.

Casella, G. and Berger, R.L., (2002). Statistical inference (Vol. 2). Pacific Grove, CA: Duxbury.

DeGroot, M.H. and Schervish, M.J., (2012). Probability and statistics. Pearson Education.

Groebner, D.F., Shannon, P.W., Fry, P.C. and Smith, K.D., (2008). Business statistics. Pearson Education.

Grus, J., (2015). Data science from scratch: first principles with python. " O'Reilly Media, Inc.".

Hodges Jr, J.L. and Lehmann, E.L., (2005). Basic concepts of probability and statistics. Society for Industrial and Applied Mathematics.

Kvanli, A.H., Pavur, R.J. and Guynes, C.S., (2000). Introduction to business statistics. Cincinnati, OH: South-Western.

McKinney, W., (2012). Python for data analysis: Data wrangling with Pandas, NumPy, and IPython. " O'Reilly Media, Inc.".

Mendenhall, W. and Sincich, T., (1993). A second course in business statistics: Regression analysis. San Francisco: Dellen.

Papoulis, A., (1990). Probability & statistics (Vol. 2). Englewood Cliffs: Prentice-Hall.

Pedregosa, F., Varoquaux, G., Gramfort, A., Michel, V., Thirion, B., Grisel, O., Blondel, M., Prettenhofer, P., Weiss, R., Dubourg, V. and Vanderplas, J., (2011). Scikit-learn: Machine learning in Python. Journal of machine learning research, 12(Oct), pp.2825-2830.

Pillers Dobler, Carolyn. "Mathematical statistics: Basic ideas and selected topics." (2002): 332-332.

Ross, S.M., (2014). Introduction to probability and statistics for engineers and scientists. Academic Press.

Schutt, R. and O'Neil, C., (2013). Doing data science: Straight talk from the frontline. " O'Reilly Media, Inc."