Unlocking Customer Insights:

Data-Driven Strategies for

Targeted Marketing in
Automotive Parts
An Analysis of Automobile Parts Transactions
Teresa Rose B
27th OCT 2024
 Agenda

Overview of
Problem Statement Data Overview
Analysis

Exploratory Data Customer

RFM Analysis
Analysis (EDA) Segmentation

Insights and
Recommendations
 Executive Summary
• Data: past 3 years.
• Objective: identify the underlying buying patterns of the
customers and recommend customized marketing
strategies for different segments of customers.
• Dataset: 20 columns and 2747 rows,
• Missing values and Duplicate values: None
• Outliers: some columns has few outliers
• The exploratory analysis and insights provide a clear
understanding of the data and highlight the key trends
and patterns in sales.
• RFM analysis has been performed to segment the
customers into four categories based on their buying
behavior, and customized marketing strategies have been
recommended for each segment.
• The presentation concludes with recommendations for
the company to enhance its customer relationships and
drive business growth.
 Problem
Statement
An automobile parts manufacturing company has collected data on
transactions for 3 years. They do not have any in-house data science team,
thus they have hired you as their consultant. Your job is to use your data
science skills to find the underlying buying patterns of the customers, provide
the company with suitable insights about their customers, and recommend
customized marketing strategies for different segments of customers.

• Dataset:
• Auto Sales Data: Sales_Data.xlsx
 Column Name Description
ORDERNUMBER This column represents the unique identification number assigned to each order.
QUANTITYORDERED It indicates the number of items ordered in each order.
PRICEEACH This column specifies the price of each item in the order.
ORDERLINENUMBER It represents the line number of each item within an order.
SALES This column denotes the total sales amount for each order, which is calculated by multiplying
the quantity ordered by the price of each item.
ORDERDATE It denotes the date on which the order was placed.
DAYS_SINCE_LASTORDER This column represents the number of days that have passed since the last order for each
customer. It can be used to analyze customer purchasing patterns.

Data STATUS It indicates the status of the order, such as "Shipped," "In Process," "Cancelled," "Disputed,"
"On Hold," or "Resolved"

Dictionary PRODUCTLINE
MSRP
This column specifies the product line categories to which each item belongs.
It stands for Manufacturer's Suggested Retail Price and represents the suggested selling price
for each item.
PRODUCTCODE This column represents the unique code assigned to each product.
CUSTOMERNAME It denotes the name of the customer who placed the order.
PHONE This column contains the contact phone number for the customer.
ADDRESSLINE1 It represents the first line of the customer's address.
CITY This column specifies the city where the customer is located.
POSTALCODE It denotes the postal code or ZIP code associated with the customer's address.
COUNTRY This column indicates the country where the customer is located.
CONTACTLASTNAME It represents the last name of the contact person associated with the customer.
CONTACTFIRSTNAME This column denotes the first name of the contact person associated with the customer.
DEALSIZE It indicates the size of the deal or order, which are the categories "Small," "Medium," or
"Large."
 Data Overview
Shape:
Dataset: • Total Rows: 2747
• Total Columns: 20
Sales_Data.xlsx

Data Types:
• float64
• datetime64[ns]
• int64
• object
 Data Type of each column
QUANTITYORDE ORDERLINENUM
ORDERNUMBER PRICEEACH
Total Data Types: RED BER
int64 float64
int64 int64

DAYS_SINCE_LAS
SALES ORDERDATE STATUS PRODUCTLINE
TORDER
float64 datetime64[ns] object object
int64

CUSTOMERNAM
MSRP PRODUCTCODE PHONE ADDRESSLINE1
E
int64 object object object
object

CONTACTLASTN CONTACTFIRSTN
CITY POSTALCODE COUNTRY
AME AME
object object object
object object

DEALSIZE
.
object
 Summary
• Numeric columns: ORDERNUMBER, QUANTITYORDERED, PRICEEACH, ORDERLINENUMBER
SALES, DAYS_SINCE_LASTORDER, MSRP
DAYS_SINCE_LASTO
Index ORDERNUMBER QUANTITYORDERED PRICEEACH ORDERLINENUMBER SALES ORDERDATE RDER MSRP
count 2747 2747 2747 2747 2747 2747 2747 2747

mean 10259.76156 35.10302148 101.0989511 6.491081179 3553.047583 56:17.2 1757.085912 100.6916636

min 10100 6 26.88 1 482.13 1/6/2018 0:00 42 33

25% 10181 27 68.745 3 2204.35 11/8/2018 0:00 1077 68
50% 10264 35 95.55 6 3184.8 6/24/2019 0:00 1761 99
75% 10334.5 43 127.1 9 4503.095 11/17/2019 0:00 2436.5 124
max 10425 97 252.87 18 14082.8 5/31/2020 0:00 3562 214

Inference :
• The average number of items ordered per sales order is 35, with a standard deviation of 9.76.
• The average price of each item is 101.09, with a standard deviation of 42.04.
• The average sales amount per order is 3553.05, with a standard deviation of 1838.95.
• The average time since the last order is 1757.09 days, with a standard deviation of 819.28.
• The summary statistics do not indicate any red flags or abnormalities that could potentially
indicate issues with the data.
 Exploratory Analysis & Insights

Univariate, Bivariate,
and multivariate
analysis using data
visualization Summarize the
inferences from
• Weekly, Monthly, Quarterly,
Yearly Trends in Sales
the above
• Sales Across different analysis
Categories of different
features in the given data
Product Line
Analysis
Deal Size
Analysis
Multivariate
Analysis
 Heat Map
➢ Sales & MSRP (0.63):
Higher MSRP may enhance perceived product
value, positively impacting sales.

➢ Sales & Quantity Ordered (0.55):

More sales correlate with larger order
quantities, indicating bulk purchasing trends.
➢ Days Since Last Order & MSRP (-0.57):
Longer shelf time might lead to price
reductions, signaling the need for inventory
management.

➢ Price Each & MSRP (0.78):

Strong link suggests that unit prices align
closely with MSRP, supporting pricing
consistency.
 Yearly Trends

• Inference:

* Sales showed a positive trend

and grew significantly from 2018
(3,353,014) to 2019 (4,669,925),
suggesting that the business
YEAR Num_of_Orders Sales Num_of_Products performed well and experienced an
2018 99 3353014.06 953
increase in customer demand.
2019 142 4669924.56 1329
* However, since we only have
2020 57 1737283.09 465 sales data until May for 2020
(1,737,283), it is difficult to make
conclusive statements about the
entire year.
 Monthly
Trends
• * The sales data exhibits
fluctuations throughout the
year, with peak sales recorded
in November (2,088,537) and
October (1,001,377),
indicating heightened
customer activity likely
influenced by seasonal factor.

* In contrast, June records

the lowest sales (454,757),
suggesting a potential
decrease in customer demand
during that period.

MONTH Num_of_Orders Sales Num_of_Products

2018-01 5 129753.6 39
2018-02 3 140836.19 41
2018-03 5 155809.32 44
2018-04 7 201609.55 58
2018-05 6 192673.11 58
 Quarterly
Trends
• Inference:

* The quarterly sales data over the

three-year period reveals a consistent
pattern of higher sales in Q4
(3,699,161) compared to other
quarters. This indicates a potential
seasonal trend where the end of the
year witnesses increased customer
spending.

* Q1 (2,254,029) and Q2 (2,048,120)

show relatively stable sales figures,
while Q3 (1,758,911) records the
lowest sales among the quarters.

QUARTER Num_of_Orders Sales Num_of_Products

2018Q1 13 426399.11 124

2018Q2 19 562365.22 162

2018Q3 19 649514.54 184

2018Q4 48 1714735.19 483

2019Q1 24 809841.36 225

 Sales by
Week
• Inference:

* Sundays have consistently

shown the highest sales
(2,175,121) over the three-year
period, indicating that
customers tend to be more
active and likely prefer shopping
on weekends.

* Thursdays, on the other hand,

have the lowest sales (151979
among the weekdays,
suggesting that customers may
be less engaged or influenced by
factors that affect their buying
behaviour on this specific day.
• Summary:

• Product Line Analysis: Classic cars and vintage cars are the most popular product lines, indicating a strong demand for these
vehicle models. Conversely, ships, trucks, buses, and trains have relatively lower counts, suggesting a niche market or specialized
customer base for these product lines.

• Deal Size Analysis: The majority of the deals fall into the small and medium categories, with counts of 1,246 and 1,349,
respectively. This indicates that the company primarily engages in transactions of moderate to smaller deal sizes.

• Order Status Analysis: The significant count of 2,541 successfully shipped orders reflects the company's outstanding order
fulfilment capabilities and commitment to delivering orders promptly.

• Quality Ordered Analysis: The bins with quality orders of 21, 28, 35, and 42 have the highest counts, indicating their popularity
among customers. With counts ranging from 576 to 631, these quantities meet the demands of a significant portion of customers.

• Sales Analysis: Sales showed a positive trend and grew significantly from 2018 (3,353,014) to 2019 (4,669,925), suggesting that
the business performed well and experienced an increase in customer demand. However, since we only have sales data until May
for 2020 (1,737,283), it is difficult to make conclusive statements about the entire year.
• Customer Segmentation using RFM
analysis

• 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

• RFM analysis is a customer

segmentation technique
commonly used in marketing
and retail to analyse and
categorize customers based
on their purchasing
behaviour. RFM stands for
Recency, Frequency, and
Monetary Value, which are
three key metrics used to
evaluate customer
engagement and profitability.
 Parameters and Assumptions
PARAMETERS
• Monetary: This is calculated by multiplying the price of each item by the quantity ordered, representing the revenue generated by each
customer.

• Recency: This is determined by subtracting the order date from a fixed reference date (e.g., 01-06-2020), indicating how recently a customer
made a purchase.

• Frequency: This is calculated as the count of purchases made by a customer, reflecting how often a customer makes purchases.

• Auto-binning: Customers are segmented into four categories based on their RFM scores: High, Moderate, Low, and Very Low. This
segmentation aids in categorizing customers according to their value and behavior.

• These parameters are utilized in RFM analysis to evaluate customer behavior, identify customer segments, and inform data-driven marketing
and sales decisions.

ASSUMPTIONS
• Higher Monetary Value Indicates Higher Spending: This assumption suggests that customers who generate higher monetary values through
their purchases are likely to be more valuable and potentially more profitable for the business.

• Recent Purchases Reflect Customer Engagement: It is assumed that customers who have made purchases more recently are likely to be more
engaged with the company and its offerings, presenting higher potential for repeat purchases or upselling/cross-selling opportunities.

• Higher Frequency of Purchases Reflects Customer Loyalty: This assumption posits that customers who make purchases more frequently
demonstrate greater loyalty to the company. They may have a stronger connection to the brand, higher customer satisfaction, and an
increased likelihood of recommending the company to others.
 KNIME
WORKFLOW
Output table head
Inferences from RFM Analysis
and identified segments:

• Who are your best customers?

• Which customers are on the verge
of churning?
• Who are your lost customers?
• Who are your loyal customers?
CUSTOMERNAME RECENCY_HML MONETARY_HML FREQUENCY_HML

Euro Shopping Channel High High High

Mini Gifts Distributors Ltd. High High High
La Rochelle Gifts High High High
Souveniers And Things Co. High High High
Reims Collectables High High High

• The top 5 best customers are

TOP 5 Best determined by calculating the RFM
score, which is the sum of scores for

Customers Monetary, Recency, and Frequency.

Higher RFM scores indicate more
valuable customers
 • The customers with very low to low
Customers at Risk recency, high to moderate monetary
value, and high to moderate
of Churning: purchase frequency are at an
increased risk of churning.
Identifying
Potential Churners
CUSTOMERNAME RECENCY_HML MONETARY_HML FREQUENCY_HML

Double Decker Gift Stores, Ltd Very Low Very Low Very Low

West Coast Collectables Co. Very Low Very Low Very Low
Signal Collectibles Ltd. Very Low Very Low Very Low
Daedalus Designs Imports Very Low Very Low Very Low
CAF Imports Very Low Very Low Very Low

• Customers with low frequency of

Top 5 Customers purchases, indicating reduced
engagement.
at Increased Risk • Customers with low recency,
implying prolonged inactivity.
of Churning • Customers with low monetary value,
suggesting decreased spending.
 CUSTOMERNAME Recency- MONETARY- FREQUENCY-

Saveley & Henriot, Co. High High High

Vida Sport, Ltd High Moderate Moderate

Herkku Gifts High Moderate Moderate
Amica Models & Co. High Moderate Low
Marta's Replicas Co. High Moderate Moderate
• Loyal customer refers to customers
Top 5 Loyal who are consistent and repeat
purchases over time. They may not

Customer necessarily have the highest

monetary value, but they
demonstrate a strong commitment
to the brand by repeatedly choosing
to do business with the company.
 •
RECOMMENDATIONS:
Customer Retention Strategies :Engage and retain customers who are
at risk of churning by offering personalized incentives, implementing
loyalty programs, and providing exclusive promotions that encourage
their continued loyalty.
• Continuous Monitoring and Analysis: Regularly monitor customer
behavior, sales trends, and key performance indicators to identify
shifting patterns, enabling proactive measures to address any
potential issues.
• Upselling and Cross-Selling Opportunities :Identify opportunities for
upselling and cross-selling to existing customers, particularly those
with high monetary value. Recommend relevant products or services
based on their purchase history to increase their average order value
and strengthen customer loyalty.
Thank you