Data Science Analysis For Ecommerce Company

Dataset Development and Attributes

Research Methodology

 In this paper we are using liner regression mode in order to develop a prediction mode for the sales of the products in a specific region for the e-commerce organization. Liners regression analysis can be extended to incorporate more than one independent variable in order create an accurate model.  Using the linear regression, the organization can be able to determine and exploit future opportunities as well as mitigate the risks is the most noticeable utilization of linear regression technique in the business organization. Product demand analysis for example is helpful in the prediction of the quantity of the goods which a customer will presumably buy from the e-commerce organization. However, the demand of the goods is not only the variable that are considered for the predictions in business. With the use of the linear regression technique the business organization can go a long way past estimating the sales of the products in order to improve the revenue.  

 In addition to that, Regression analysis can be helpful for a business to find specific relationship among the different variables by exposing the patterns between the different variables of a dataset that were previously not realised. For example, analysis of monthly sales and customer accounts are helpful in highlighting the market patterns. This patterns may include increase in demand for certain products or in a certain geographic region of the country.

Use of this techniques helps the organizations to reduce the tremendous amount unstructured raw data in actionable as well as usable information for its business. Therefore, it can be stated that linear regression analysis leads to the more accurate and smarter decisions to improve the business performance in the market.

For instance, in the forecasting model, it may be trusted that the quantity of products sold can be dependent on the number of customers in that region as well as on the total monthly sales of the product in the specific region. Recorded information on these three factors must be considered and the best fit for the should be evaluated. Different stages of the analysis are provided in the different sections of this report.  

Analytical Findings

 For the acquired data set, at first we collected the statistical details of it using python language.   The overall statistics of the dataset generates the following result,

	product price	monthly sales ($)	customers count
count	1303	1303	1303
mean	13.046	150.11	19.98
std	3.180	57.926	8.81
min	8.00	50.00	5.00
25%	10.00	98.00	12.00
50%	13.00	153.00	20.00
75%	16.00	198.00	27.00
max	18.00	250.00	35.00

                                                                                 Table 1: Statistical data about the dataset

Researching Data Mining Techniques

The above table can be interpreted as, the dataset includes total 1303 rows of data where the maximum mean and minimum value for product price are $ 8, $ 13.04, and $18 respectively.  For the monthly sales of the product we found that, the minimum value for the monthly sales is $50, mean value is 150.114 and maximum value is $250.

In further analysis of the data we found that there are, 7 unique product names, 2 different shipping types. In addition to that, mainly the customers are form 5 geographic regions of Australia (TAS, QLD, NSW, VIC, ACT) and 2 customer types (namely existing and new)

Most likely geographic region to target new customers

From the statistical analysis of the monthly sales and customer count from the different geographic regions are given by, the following table,

	product_price	monthly_sales($)	customers_count
Gegraphic_region
NSW	3377	37605	5065
QLD	3737	43752	5618
TAS	3541	39383	5302
ACT	3322	39951	5203
VIC	3023	34908	4850

                                                                             Table 2: variation of sales in different regions

 From the above table, it is evident that, comparatively there are lesser number of customers from the VIC region. Which is 4850 and the total revenue from this region is also minimum among all the regions.  Therefore, it can be said that, it is important to target the VIC area to find and attract new customers in order to increase the revenue.

Products to be prioritized for increase in sale

When the sales of the products are grouped, according to the number of sales provided in the dataset, the following table is generated,

	product_price	monthly_sales($)	customers_count
Product_Name
ARM WATCHES	2673	30160	3906
BINOCULARS	2403	28548	3905
BUGSHIELD SPRAY	2386	27287	3636
EDGY PUTTERS	2273	27185	3576
JILO IRON	2288	26059	3576
SOLAR COOKING GEAR	2509	29132	3670
TRS EYEWEAR	2468	27228	3769

                                                                                            Table 3: sale of products

 From the above table it is evident that, the sales of JILO IRON are minimum in the terms of the total monthly sales and total number of customers who bought the product. The total amount of monthly sales is provided by $26059 and the total number of customers who bought the product is given by 3576. Even though the number of customer are same for the JILO IRON and EDGY PUTTERS are same but the total revenues for JILO IRON is lesser than the EDGY PUTTERS.   Therefore, the organization should prioritize the sale of the JILO IRON in order to improve the overall profit from the business.

Developing Predictive Models

Impact of changing the shipping type of the products

 In the analysis of the dataset, we found that, the revenue and the total number of customers for free and customer paid delivery is given by,

 

shipping_type	monthly_sales($)	customers_count
FREE	99329	13714
CUSTOMER PAID	96270	12324

As it can be observed in the above table that the total monthly sales and the customer count is lesser for the customer paid shipping compared to the free shipping products from the e-commerce organization.

Prediction model using liner regression technique

For the given data set we have used the liner regression model in order to predict the sales of the products in the ‘VIC’ geographic region. After implementing the liner regression model we found that in this region trend of the sales of the product is increasing and with proper strategies the present business scenario can be improved.  The regression plot is like the following figure,

                                                           

                                                             Figure 1: prediction models with the linear regression technique

Recommendations to the company

Upselling of the products: From the above analysis of the dataset it can be observed that there is relation between the sales of the different products in the dataset. Therefore, it is suggested to use upsell technique by the e-commerce organization. Upselling is the point at which the business organization urges or promotes the client or the customer to spend more than they had initially planned.

Personized product recommendation:   The organization can try to offer relevant products recommendations to the potential customers at the different phases of the shopping on the website. The useful and Intuitive recommendations can encourage the users to buy the add on products.

Providing free shipping of the products: As observed in the previous analysis it can be said that, there are opportunities of improvement of the sales of the products in the “VIC” region.  

Implementation plan based on the recommendations

 The organization should try to close out the products that are slow selling or slow moving from the shelf. Items that stays on rack for longer period of time and sales rate is slow can, in a few occasions, reduce the revenue and decelerate a business improvement. Therefore, it is important to have closeout deals.

Analytical Findings: Geographic Regions and Products

 More over the organization should provide free shipping for the products in the different regions listed in the data se as depicted in the above sections of the data analysis there are number of sales that are lesser than the freely shipped products.

In addition to that, the e-commerce organization can help in improving the business, sales of the product as well as revenue from the sales of the product. Positive feedback of the   clients can enable other users to pick up the trust of potential clients. Negative feedbacks can enable the business organization to pick up understanding and work on the focuses that clients communicated an issue with the administrations.

Conclusion:

Sales forecasting is a necessary tool for any business to exploit a business opportunity due to the trend of business market and historical data. Precise sales forecast encourages powerful and effective allotment of the scarce business resources. Over-estimation of the sales lead to business issues.  This can be listed as excessive inventory usage leads to un necessary use of profitable rack space and prompts out increased operational cost.

With the ever increasing competition in the market e-commerce organization need to know, what are the sources that is generating revenue historically or the customer base is loyal and the sources from which sales are coming from in order to acquire them.   In this scenario, it becomes important to forecast the product sales which may help in the lesser wastage of the inventory by maintaining a stricter inventory control system at place. Next, due to the inefficient use of rare working capital obstructed stock clearing. Third, product storage charges are acquired to store abundance stock out in the open or private ware houses of the organization.

Baardman, L., Levin, I., Perakis, G. and Singhvi, D., 2017. Leveraging Comparables for New Product Sales Forecasting.

Exterkate, P., Groenen, P.J., Heij, C. and van Dijk, D., 2016. Nonlinear forecasting with many predictors using kernel ridge regression. International Journal of Forecasting, 32(3), pp.736-753.

Montgomery, D.C., Jennings, C.L. and Kulahci, M., 2015. Introduction to time series analysis and forecasting. John Wiley & Sons.

Omar, H., Hoang, V.H. and Liu, D.R., 2016. A Hybrid Neural Network Model for Sales Forecasting Based on ARIMA and Search Popularity of Article Titles. Computational intelligence and neuroscience, 2016.

Trapero, J.R., Kourentzes, N. and Fildes, R., 2015. On the identification of sales forecasting models in the presence of promotions. Journal of the operational Research Society, 66(2), pp.299-307.

Van Donselaar, K.H., Peters, J., De Jong, A. and Broekmeulen, R.A.C.M., 2016. Analysis and forecasting of demand during promotions for perishable items. International Journal of Production Economics, 172, pp.65-75.

 Å½ylius, G.G., Simutis, R. and Vaitkus, V., 2015. Evaluation of computational intelligence techniques for daily product sales forecasting. International Journal of Computing, 14(3), pp.157-164.