Project Report on Marketing & Retail Analysis

(Data Science & Business Analytics Program)

Submitted by

Shehroz Arfeen Khan

Batch: PGPDSBA.O.SEP22.B

Great Lakes Institute of Management

Contents
Automobile (Part-A) ...................................................................................................................................... 3
Grocery Store (Part B) ................................................................................................................................. 11
Automobile (Part-A)
Kindly access the link for tableau work book

Shehroz Project MRA_Automobile | Tableau Public

1. Agenda & Executive Summary of the data -> Contents of the ppt -> Problem statement -> About
Data (Info, Shape, Summary Stats, your assumptions about data)

• We have received the 3 years data of automobile part manufacture. The data is about an
automobile parts manufacturing company.
• They have provided the data collected of transactions for 3 years.
• The data has 2747 entries (0 To 2746) of rows and 20 columns.
• The data has 1 datetime64 , 2 float64, 5 int64,and 12 Object data types.
• There is no missing values present in the data set.
• This data more or less reflects the purchasing behavior of customers in different categories.
• The company is into automobile part manufacture, and they have different product line like
Classic car , Motorcycle, plane, train, ship, Bus truck, vintage cars etc.
• The data maintained each transactions entry as order number and for each order number
maintained all required information like customer identity details , and product details like price
, quantity , product code, and sales for each customer.
• We noticed that one order number has many different entries with different product codes.
• Manufacturer's Suggested Retail Price(MSRP) for each product code is decided but we found that
this is not matching with Price of Each item & is inconsistent with MSRP.

2. PART A: Exploratory Analysis and Inferences -> Univariate, Bivariate, and multivariate analysis
using data visualization (Weekly, Monthly, Quarterly, Yearly Trends in Sales and Sales Across
different Categories of different features in the given data) -> Summarise the inferences

Univariate Analysis
 • Using boxplot on sales & quantity order variable we have plotted univariate analysis.
• We can clearly see that outlier is present there.
• Also using histogram on sales variable we did univariate analysis.
• For Categorical variable like product line we also did univariate analysis using bar plot.
• We have no noticed that the sales of classic cars products are high followed by vintage car product
sales

Bivariate Analysis
 • Using boxplot on sales & product line variables we have plotted bivariate analysis.
• We can clearly see that outlier is present in each product line category .
• Using boxplot on sales & deal size variables we have plotted bivariate analysis.
• We can clearly see that outlier is present in Large deal size. In Pie chart we can see the larger
portion of classic cars followed by vintage cars were as trains has the least demand.

Multivariate Analysis
 • MSRP, Price Each, status, sales & product line using these variables we did multivariate analysis.
• For this we used horizontal bar, tree map, stack bar, scatter plot respectively.
• As sales are high for classic cars the company has even sold below MSRP, there might be a chances
that the company has given more discounts to its customers. And vice versa for vintage cars were
the company has sold above MSRP.
• Ship, vintage car & train are been sold above the MSRP.
• By looking at the given data almost all the transactions are been shipped.

Timeseries & Trend sales

• Yearly, Quarterly, monthly, Weekly time series analysis & its trend are been shown.
• We observed that in Last quarter sales are high as compared to other quarters.
 • There is a seasonality seen.

Summary of Inferences

• Using histogram on sales variable we did univariate analysis.

• For Categorical variable like product line we also did univariate analysis using bar plot.
• Using boxplot on sales , product line, deal size variables we have plotted bivariate analysis.
• And using MSRP, Price Each, status, sales & product line variables we did multvariate analysis
• Ater deriving univariate, bivariate & Multivariate analysis we can see there is a high demand of
classic cars followed by vintage cars and least is for trains.
• The sale are high for the last quarter of the year & we can see seasonality in it.
• The demand for classic cars are so high that the company has also sold the products below MSRP
giving the customers a good discount. However, for vintage cars they have sold above the MSRP
too.
3. PART A: Customer Segmentation using RFM analysis (4 segments) -> What is RFM? -> What all
parameters used and assumptions made? -> Showcase the KNIME workflow image -> What
results are there in the output table head?
• RFM is recency, Frequency & monetary analysis
• Ignoring the column "Days Since last order" and create new column name Recency
• as "[Max(order date) - order date)]"We have assumed “01-06-2020“ as a reference date and
created recency column.
• If we can see the data there are same order number repeated for different product Code. So we
can assume count of each order number as frequency of an order number.
• In SALES column we get sales amount for each transaction.
• We can use SALES parameter and using an assumption of sum of aggregation we created a new
column as Monetary
• Then created four different bin for each Recency, frequency & Monetary using percentile range
(0,0.10,0.40,0.70,100).
• Based on above 4 bin assumption we have considered 4 segments like High, Medium , Low and
Churn.
KNIME Work flow Image
Output Table head

4. PART A: Inferences from RFM Analysis and identified segments -> Who are your best customers?
(give at least 5) -> Which customers are on the verge of churning? (give at least 5) -> Who are
your lost customers? (give at least 5) -> Who are your loyal customers? (give at least 5)
• On basis on Recency, frequency & monetary we have grouped our top customers.
• We have given the most signifcance to recency parameter as these customers has recently
purchased our products.
• Also according to RFM model the most importance is given to recency.
• Hence we have kept it as our parameter for selecting top customers.
• For eg. Customer name -Euro Shopping Channel, they have recently made a purchase, also has
high frequency with a high monetary transaction.

Our Loyal Customers

• On basis on Recency, frequency & monetary we have grouped our loyal customers.
• These customers have purchased multiple items with good monetary value.
• If we focus more on this segment of customers, we can easily turn them into our top best
customers too. Also, in this segment we can see the customers for productline - classic cars are
many.
Customers on verge of churning

• On basis on Recency, frequency & monetary we have grouped our Customers who are on verge
of churning.
• We should definitely focus on this group before we lose them and try to convert them into our
regular customers.
• For e.g. Customer name Saveley & Henriot,Co – Their frequency is good with good monetary
value, but low recency made them stand in this group. If the company pays more attention and
fulfil their requirement, then we can easily turn them into our regular customer and we can save
them from churning out.

Lost Customers

• On basis on Recency, frequency & monetary parameters we have grouped our Customers who
we’d lost. Their recency is very low and hasn’t made any purchase since long.
• So we can say these are our lost customers. If taken feedback from them and fulfill their demand
we might bring them back to been a good customer.

Recommendation

• Using Recency, frequency & monetary parameters we have grouped our top , loyal, on the verge
of churning and lost customers.
• Customers with good recency has been our top customers were as we also have lost customer
lists.
• Customers on verge of churning can be saved and can be converted into a good buyer.
• RFM model is used for deriving the customers types like Loyal, top or best, on verge of churning
& lost customers.
• Recency, frequency & monetary parameters were widely used to bifurcate the types of
customers.
• This model can be very helpful to the company to maintain its sales and customers and can focus
on how the company has lost the customers & can take various actions to bring back them.
• It is vital for the company to convert the customers who are on verge of churning into a regular
customer or atleast maintain them.
• And also how to increase the sales ratio can be identified.
Grocery Store (Part B)

Kindly access the Tableau work book below

shehroz_MRA_Grocery | Tableau Public

1. PART B: Exploratory Analysis --> Exploratory Analysis of data & an executive summary (in PPT)
of your top findings, supported by graphs. --> Are there trends across
months/years/quarters/days etc. that you are able to notice?

• Data: from 01-01-2018 to 26-02-2020

• Objective: project involves conducting a thorough analysis of Point of Sale (POS) Data for
providing recommendations through which a grocery store can increase its revenue by
popular combo offers & discounts for customers.
• Dataset: 20641 Rows, 3 columns,
• Missing values : None
• Duplicate values: 4730
• The exploratory analysis and insights provide a clear understanding of the data and highlight
the key trends and patterns in sales.
• Market Basket Analysis using association rules was performed to identify the relationships
between the products purchased by the customers.
• This analysis helped to identify the products that are frequently purchased together, which
can be used to create lucrative offers for the customers.
Duplicate values

• It is generally a good practice to drop duplicate rows in a dataset as they do not provide any
additional information and can skew the results of any analysis performed on the dataset.
• However, in this particular case, dropping duplicate rows may not be appropriate as there is no
unique identifier for each row.
• Each row consists of a date, a customer ID, and a product purchased, but the same product can
be purchased by multiple customers on the same date.
• Therefore, we drop duplicate rows, it may inadvertently remove valid information from the
dataset.
• So duplicate values are not removed from the dataset
Assumptions

• The data represents a list of items purchased at a grocery store on various dates.
• Each entry in the data represents a single item purchased.
• The first column in the data represents the date the item was purchased.
• The second column represents the customer who made the purchase.
• The third column represents the item purchased.
• The same item can be purchased by multiple customers on different dates.
• There is no information provided about the quantity or price of each item.
 • We have not dropped the duplicated values.

Exploratory Data Analysis

Count of Products Sold

Product Count
Count of products sold yearly

Top 3 products over the year

Eatables

Non-Eatables
Product sales vs Time

Observations

• As we have data till 26 feb 2020 that’s why the count of products sold in 2020 is low
• As we have data till 26 feb 2020 that’s why the count of products sold in Q1 is Hight.
• In 2019 Q1 sales was highest
 • In 2018 Q3 sales was highest
• Count of product sold in Q2 is aproxx same in 2019 and 2018.
• In 2018 most of the products were sold in January and least were sold in February.
• In 2019 most of the products were sold in March and least were sold in January.
• Most of the products were sold on Sundays.
• Least products were sold on Mondays.
• On other days sales in consistent.
• There are total 28 products in this category.
• Highest sold : poultry , Soda, cereals
• Least sold : pork, fruits, sandwich loaves
• There are 9 products in this categoary.
• Highest Sold Products : Soap, Toilet Paper.
• Lowest Sold Products: Hand soap.
• We can see poultry, cereals and soda are highly sold products over the years
• Poultry, Soda and Cereal are the top 3 Products sold in 2019
• Cereals, Poultry and Soda are the top 3 Products sold in 2018

Recommendations

• Focus on promoting and stocking up on poultry, soda, and cereals as they are consistently top-
selling products.
• Consider increasing the stock of soap and toilet paper as they are the highest sold non-eatable
products.
• Evaluate the reasons behind the low sales of hand soap and take measures to increase its sales.
• Schedule promotions and offers on Sundays to maximize sales on the day with the highest sales.
• Plan marketing campaigns and discounts during February to increase sales during the historically
low-sales month.
• Plan marketing campaigns and discounts during January and March to increase sales during the
historically high-sales months.
• Aim to replicate the sales patterns of Q1 2019 and Q3 2018.
• Keep the stock of products sold in Q2 consistent with the previous years to maintain sales levels.
• Keep in mind the limited data for 2020 while making sales and marketing decisions.
2. PART B: Use of Market Basket Analysis (Association Rules) -->Write Something about the
association rules and its relevance in this case -->Add KNIME workflow image -->Write about
threshold values of Support and Confidence

• Definition: Market Basket Analysis is a statistical technique that analyzes customer

purchase patterns to identify associations between different products. It helps businesses
understand which products are frequently purchased together and how customers'
buying habits affect sales.
• Data: To conduct market basket analysis, businesses need transactional data that includes
details such as customer ID, product ID, and transaction date. This data is then used to
create a matrix that represents the relationships between different products.
• Association Rules: Association rules are used to identify the strength of the relationship
between different products. These rules are expressed in terms of support, confidence,
 and lift. Support refers to the frequency of co-occurrence of items in a transaction, while
confidence measures the probability that if a customer buys one item, they will also buy
another. Lift measures the degree of correlation between two items.
• Applications: Market Basket Analysis is used in a variety of industries, including retail, e-
commerce, and marketing. Retailers use this technique to optimize product placement
and promotions. E-commerce companies use it to personalize product recommendations,
and marketers use it to develop targeted advertising campaigns.
• Benefits: Market Basket Analysis helps businesses increase revenue by identifying cross-
selling opportunities and developing targeted promotions. It also helps improve customer
satisfaction by providing personalized recommendations and improving the overall
shopping experience

Association Rule Parameters

• Support of Minimum: 0.05

• Maximum Item Set Length : 10

 • Minimum Confidence Level:0.6

Market basket analysis, support, confidence, and lift values

• In market basket analysis, support, confidence, and lift values are used to measure the strength
of association between items in a transaction dataset.

• Support: It is the probability of observing the items together in a transaction. It is calculated as

the number of transactions that contain both items divided by the total number of transactions.
It measures how frequent the itemset occurs in the dataset. High support indicates that the
itemset is popular and should be considered for promotion or placement together.

• Confidence: It is the conditional probability that a transaction containing one item also contains
another item. It is calculated as the number of transactions containing both items divided by the
number of transactions containing the first item. It measures the strength of the association
between two items. High confidence indicates that the items are likely to be bought together, and
can be used to recommend or suggest items to customers.

• Lift: It is the measure of how much more often two items occur together than expected if they
were independent of each other. It is calculated as the support of the itemset divided by the
product of the individual supports of the items. A lift value of 1 indicates that the items are
independent, while a value greater than 1 indicates a positive association between the items. A
lift value less than 1 indicates a negative association between the items. High lift indicates that
the items have a strong association and can be used for cross-selling or bundling.

Association rules are a technique used to find relationships or associations between items in a large
dataset. These rules are based on the concept of frequent itemsets, which are sets of items that appear
together frequently in a transactional dataset.

24 rules have been found with the dataset and set parameters
Recommendations

• Offer a "Buy Two Get One Free" promotion on yogurt, poultry, and aluminum foil to encourage
customers to purchase more items at once.
• Create a combo deal where customers can purchase cereals, bagels, and sandwich bags together
at a discounted price.
• Offer a discount on mixes when purchased with yogurt, poultry, or aluminum foil.
• Provide a discount on dinner rolls when purchased with spaghetti sauce or poultry.
• Create a "Paper Products Bundle" offer that includes paper towels, toilet paper, and/or tissues at
a discounted price.
• These discount offers and combos can help increase sales by providing customers with more value
for their money and encouraging them to purchase more items. It is important to promote these
offers through in-store signage, advertisements, and social media to ensure customers are aware
of the deals available.