customer �personality analysis

Sagung Ratu | W2W3 Amsterdam

Backgrounds

A supermarket has customer data, total purchases of each product, and purchase channels with more than 2000 rows of data.

• Supermarkets want to know how the characteristics of their consumers.
• The company also wants to know how influential each purchase channel (Web, Catalog, and Store) is on the 6 product categories (Wine, Fruit, Meat, Fish, Sweet, and Gold).

What we do?

EDA

To find customer characteristic

Regresion

To find the relationship between each purchase channels and each products

Methodology

01

02

03

04

05

06

Data cleaning

• Missing And Null Values
• Check Typos
• Irrelevant Value
• Duplicates Data

​

Descriptive analysis

​

• Std Deviation, Median, Mean
• Skewness, Min And Max
• Q1,Q3, And IQR

Calculated Outlier

• Calculate And Highlight The Outlier
• Delete Some Outlier

Correlation and sampling

​

• Preparation Data For Correlation
• Do Correlation
• Sampling Using Systematic Sampling

​

analysis

• EDA
• Regression Models

Collect insight

The supermarket condition

Purchase channels

3 purchase channels

• Store
• Web
• Catalog

Total Customers

6 product category

product category

We have +2K customers

• Wine
• Fruit
• Meat
• Fish
• Sweet
• Gold

Store Shopping Is The Most Popular Shopping Method (Sales Channels)

Around 46%

Sales Come From Store Sales

Wine is most popular product �50% sales come from wine category

Education background �of customers

Graduation

PHD

basic

1126 customer

(50%)

483 customer (22%)

master

572 customer

(26%)

54 customers (2%)

Most of our customer is married and around age 36-45

strategy

Mercury

It’s the closest planet to the Sun

Jupiter

It’s the biggest planet of them all

Saturn

It’s a gas giant and also has rings

Since very strong correlation (0.7 to 1.0 and -0.7 to -1.0) indicates multicollinearity, to make regression modelling, we must avoid those variables especially when they are independent variables, it can be seen in the table in the grey box that the relationship between the independent variables doesn't have a value that is too high, although there is 0.5 but it is still tolerable. Means that there is no or little possibility of multicollinearity

Correlation Table

Example of regression analysis

Result of Regression

• Of the 6 products (Wine, Fruit, Meat, Fish, Sweet, and Gold) and 3 purchase channels (Web, Catalog, and Store ), It turns out that only wine and meat are affected by the purchase channel. Fruit, fish, sweet and gold are not affected by the purchase channel.
• Only wine and meat products can and are eligible to obtain a multiple regression model, because only these two products are affected by the purchase channel.
• The model of multiple regression are:
• Wine: = 20.91 (web) + 52.45 (catalog) + 28.08 (store) -106.72
• Meat: = 49.61 (catalog) + 12.9 (store)-47.34
• For fruit, fish, sweet, and gold products, have a small r square or small adjusted r square value, which means that the sales channel still has not significantly affected to the sales.

​

Thank You