Abstract

Customer segmentation is one of the key methods in marketing analytics and

has been used to segment customers on various criteria and drive business results.
One of the techniques used to segmenting the customers basis the behaviour they
have exhibited in the past is RFM Analysis. RFM is an acronym for Recency,
Frequency and Monetary value. With the increasing availability of past transac-
tional data, RFM analysis can be effectively used to segment customers and drive
subsequent business actions. This study performs customer segmentation on past
transactional data using K-Means clustering algorithm in Python and basis the
created segments, recommended course of actions is suggested.
Promotional codes are alphanumeric strings that online stores offer to encour-
age purchases on their website and are typically associated with an overarching
promotional marketing strategy. Generating an automatic coupon generator will
generate custom coupons for every group of customer according to the need. The
parameters considered will help gaining more profit for the company.
Contents

Abstract i

List of Tables ii

List of Figures iii

1 Introduction 1
1.1 Problem Statement . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.2 Existing System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.3 Proposed System . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3

2 Review of Literature 5

3 Methodology Planning 7
3.1 Hardware Software requirement . . . . . . . . . . . . . . . . . . . . 7
3.2 Algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
3.3 Code Implementation . . . . . . . . . . . . . . . . . . . . . . . . . . 12
3.4 Work Flow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16

4 Conclusion 18

5 Future scope 19

References 20

i
List of Tables

3.1 CSV file with all the orders . . . . . . . . . . . . . . . . . . . . . . 8

3.2 RFM Segmentation . . . . . . . . . . . . . . . . . . . . . . . . . . . 9

ii
List of Figures

3.1 Project Roadmap . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17

iii
Chapter 1

Introduction

Understanding people’s needs, behaviors, and expectations is key to communi-

cating and connecting with them, promotional strategies. Rather than viewing
potential and existing customers as a group, businesses should take the time to
get to know them as individual groups with distinct perspectives
The segmentation process begins by grouping customers and potential cus-
tomers into customer segments with similar characteristics, so that you can com-
municate to all of the individuals in that segment efficiently, effectively, and with a
sense of personal attention, without contacting each individual individually. As a
result, your marketing tactics will become more effective and efficient, saving you
time and money while also increasing the benefits.
Segmentation helps businesses customize their marketing strategies to provide
shoppers with the brand experience they’re looking for. This can range from send-
ing targeted email campaigns to designing in-store displays to serving personalized
digital ads. As a result, customers feel more connected and engaged. Personaliza-
tion can lead to customer loyalty and brand advocacy, especially when companies
nurture the business-consumer relationship in the right ways.
When a store issues promotional codes, they’re providing customers with an
incentive to buy, which benefits both the customer and the business. These in-
centives even have a direct impact on the overall shopping experience, according
to research conducted by the Center for Neuroeconomics Studies at Claremont
Graduate University. The discount can either be a percentage or a specific dollar

1
amount. Promo codes also provide customers with free shipping or gift-wrapping.
This marketing strategy essentially gives customers yet another reason to buy your
products.
Hence along with a successfull marketing strategy of generating coupons for
customer this system generates specialised custom coupons for customers. It anal-
yses the database containing customer’s purchase record and ans according to the
database after performing RFM analyses special coupons are generated for specific
group of customer.

1.1 Problem Statement

Any company in retail, no matter the industry, ends up collecting, creating, and
manipulating 1 data over the course of their lifespan. These data are produced
and recorded in a variety of contexts, for example no of orders, amount of prod-
ucts purchased by an individual and so on. The wider one’s access to data, the
clearer the picture that the data can portray. With a clear picture created from
data, previously unseen details begin to emerge, stimulating fresh perspectives and
innovations.
It is crucial for them to have a sense of who their customers are. Not just the
products they like to purchase, but when they like to purchase them, how often
they want to purchase them, and what their lifetime value may be to the company.
While some of these questions are more straight forward than others, it is clear
that they all require data munging, analysis, and presentation that involve skills
and techniques beyond what is required of a traditional analyst.
By integrating machine learning practices and conventional business under-
standings, the paths to answering these questions became more intertwined with
that of a similar question: What segments or groups of customers do we have?
After studying clustering and reading about it in numerous other contexts, it be-
came clear that segmenting retail customers became one way to investigate the
purchasing patterns and behaviors of its customers.
Companies need to advance their marketing strategies with time. Selling prod-
ucts is not sufficient, customer must be attracted with various methods. Customers
who are loyal to your brand should be rewarded for their good behaviour. Once the

2
customer engagement is lost re-engagement must be initiated. Further more the
current strategy of sending cold coupons, promo-code and offers to the customer
is not strategic.

1.2 Existing System

The collection of three values for each customer is called an RFM cell. In a simple
system, organizations average these values together, then sort customers from
highest to lowest to find the most valuable customers. Some businesses, instead of
simply averaging the three values, weigh the values differently.
Further in this project shows how we can segment customers into different
clusters using the K-means algorithm. But, we would be implementing RFM
analysis to get the desired values and those features will be used as an input in
K-means, to determine similarity.

1.3 Proposed System

In the Existing system, average of these values together are sorted into customers
from highest to lowest to find the most valuable customers. This data is further
utilized in generating an automatic coupom code generator which generals coupons
on the basis of analysis extracted form rfm of the database.
When a store issues promotional codes, they’re providing customers with an
incentive to buy, which benefits both the customer and the business. These in-
centives even have a direct impact on the overall shopping experience, according
to research conducted by the Center for Neuroeconomics Studies at Claremont
Graduate University. The discount can either be a percentage or a specific dollar
amount. Promo codes also provide customers with free shipping or gift-wrapping.
This marketing strategy essentially gives customers yet another reason to buy your
products.
Hence along with a successfull marketing strategy of generating coupons for
customer this system generates specialised custom coupons for customers. It anal-
yses the database containing customer’s purchase record and ans according to the

3
database after performing RFM analyses special coupons are generated for specific
group of customer.

4
Chapter 2

Review of Literature

According to Blattberg, R.C., Kim, BD., Neslin, S.A. (2008). RFM Analysis.
In: Database Marketing. International Series in Quantitative Marketing, vol 18.
Springer, New York, NY. https://doi.org/10.1007/978-0-387-72579-61 2
Abstract Recency (R), Frequency (F), and Monetary Value (M) are the most
popular database marketing metrics used to quantify customer trans- action his-
tory. Recency is how recently the customer has purchases; frequency is how often
the customer purchases, and monetary value is the dollar value of the purchases.
RFM analysis classifies customers into groups according to their RFM measures,
and relates these classifications to behaviors such as the likelihood of responding to
a catalog or other offer. RFM analysis was prob- ably the first “predictive model”
used in database marketing.
How do you select customers for target mailing? Or whom should you send
your catalogs or direct mail offers to? The need to mail smarter is always among
the top concerns of direct marketers. The direct mail promotion that results in
sales to 2 Percent of the mailed universe is considered a success. Identifying and
targeting the customers who are most likely to respond are therefore of prime
concern.
Because of the nature of their businesses, direct marketers including cat- alogers
have been collecting customer data, analyzing them, and developing models for
several decades to improve their business performance. One pop- ular approach
used to improve mailing efficiency is the RFM – Recency, Fre- quency, Monetary

5
amount – model. The primitive form of the RFM model was used about 50 years
ago by catalogers of general merchandise. For exam- ple, as early as 1961, George
Cullinan promoted the use and understanding of RFM customer data analysis.
Recognizing his contribution in advancing the direct marketing industry, the DMA
(Direct Marketing Association) inducted him into the DMA Hall of Fame in 1989.
The core concept of the RFM model is based on the empirical evidence. Direct
marketers have found that the response to a mailing offer is hetero- geneous across
customers. And they also found that customers who haveresponded well in the
past are likely to respond in the future. More specif- ically, direct marketers have
found that customers’ purchase response can be predicted using their previous
purchase histories. The three most impor- tant variables to summarize customers’
purchase histories are recency (R), frequency (F), and monetary amount (M). That
is, using RFM measures for each customer, one can predict his or her propensity
to respond. Once identifying who is going to respond, the direct marketer sends
catalogs to customers with high propensity.
It encounters several scoring challenges in the measure of frequency and is rela-
tively sensitive, which leads to pulling apart customers who have identical behavior
at the lower quintiles, but group customers together whose buying behaviors have
significant differences (Alam and Khalifa, 2009).
There is another scoring method (behavior quintile scoring method) developed
by John Wirth (The founder of Woodworker’s Supply of New Mexico), which
sorts customers based on their behavior and thus may have different number of
customers in each quintile. The scoring scheme of frequency covers five intervals,
including 0 to 3 months, 4 to 6 months, 7 to 12 months, 13 to 24 months and
25+ months, which are coded as 5, 4, 3, 2 and 1, respectively. This method is
known as hard coding (McCarty and Hastak, 2007). For frequency, although this
scoring method appears to solve the sensitivity problems, it still encounters similar
problems as customer quintile method in the frequency measure.

6
Chapter 3

Methodology Planning

Methodology and planning is a systematic approach to designing, executing and

delivering a project. It is a clearly defined combination of related practices, meth-
ods, and processes which determine how to best execute the project.
A summarized research and planning spread light upon the required tech stack
and various algorithms, modules to be taken under consideration.

3.1 Hardware Software requirement

Hardware:

• Processor: Intel core i3 6th gen onwards

• RAM: 2 GB or higher

• Hard Disk Space: 20 GB or higher :

Software

• Python

7
3.2 Algorithm
Step 1
We will start with a dataset containing information about individual orders
performed on a platform. The data is represented in a transaction level i.e. each
row of the dataset contains features associated with a single transaction, such as
date, time, payment method, client id of the user that made the purchase, etc.

orderd ate orderi d customer grandt otal

2022-01-01 IN-52653 Khushi 90
2022-01-02 IN-52654 Naziya 150
2022-01-03 IN-52655 Pranjal 70

Table 3.1: CSV file with all the orders

Step 2
The next step in building an RFM model is to assign Recency, Frequency and
Monetary values to each customer

• Recency is simply the amount of time since the customer’s most recent trans-
action (most businesses use days, though for others it might make sense to
use months, weeks or even hours instead).

• Frequency is the total number of transactions made by the customer (during

a defined period).

• Monetary is the total amount that the customer has spent across all trans-
actions (during a defined period).

Step 3
The third step is to divide the customer list into tiered groups for each of the
three dimensions (R, F and M).
This results in 64 distinct customer a (4x4x4), into which customers will be
segmented. Three tiers can also be used (resulting in 27 segments); using more

8
 Recency Frequency Monetary
R-Tier-1 (most recent) F-Tier-1 (most frequent) M-Tier-1 (highest spend)
R-Tier-2 F-Tier-2 M-Tier-2
R-Tier-3 F-Tier-3 M-Tier-3
R-Tier-4 (least recent) F-Tier-4 (only one transaction) M-Tier-4 (lowest spend)

Table 3.2: RFM Segmentation

than four, however, is not recommended (because the difficulty in use outweighs
the small benefit gain from the extra granularity).

Step 4
The fourth step is to select groups of customers to whom specific types of
communications will be sent, based on the RFM segments in which they appear.
It is helpful to assign names to segments of interest. Here are just a few
examples to illustrate:

• Best Customers – This group consists of those customers who are found in
R-Tier-1, F-Tier-1 and M-Tier-1, meaning that they transacted recently, do
so often and spend more than other customers. A shortened notation for
this segment is 1-1-1; we’ll use this notation going forward.

• High-spending New Customers – This group consists of those customers in

1-4-1 and 1-4-2. These are customers who transacted only once, but very
recently and they spent a lot.

• Lowest-Spending Active Loyal Customers – This group consists of those cus-

tomers in segments 1-1-3 and 1-1-4 (they transacted recently and do so often,
but spend the least).

• Churned Best Customers – This segment consists of those customers in

groups 4-1-1, 4-1-2, 4-2-1 and 4-2-2 (they transacted frequently and spent a
lot, but it’s been a long time since they’ve transacted).

Marketers should assemble groups of customers most relevant for their partic-
ular business objectives and retention goals.

9
 Step 5
The fifth step actually goes beyond the RFM segmentation itself: crafting
specific messaging that is tailored for each customer group. By focusing on the
behavioral patterns of particular groups, RFM marketing allows marketers to com-
municate with customers in a much more effective manner.
Again, here are just some examples for illustration, using the groups we named
above:

• 1. Core - Your Best Customers

RFM Score: 111

Who They Are: Highly engaged customers who have bought the most recent,
the most often, and generated the most revenue.
Marketing Strategies: Focus on loyalty programs and new product introduc-
tions. These customers have proven to have a higher willingness to pay, so don’t
use discount pricing to generate incremental sales. Instead, focus on value added
offers through product recommendations based on previous purchases.

• 2. Loyal - Your Most Loyal Customers

RFM Score: X1X

Who They Are: Customers who buy the most often from your store.
Marketing Strategies: Loyalty programs are effective for these repeat visitors.
Advocacy programs and reviews are also common X1X strategies. Lastly, consider
rewarding these customers with Free Shipping or other like benefits.

• 3. Whales - Your Highest Paying Customers

RFM Score: XX1

Who They Are: Customers who have generated the most revenue for your
store.

10
 Marketing Strategies: These customers have demonstrated a high willingness
to pay. Consider premium offers, subscription tiers, luxury products, or value add
cross/up-sells to increase AOV. Don’t waste margin on discounts.

• 4. Promising - Faithful customers

RFM Score: X13, X14

Who They Are: Customers who return often, but do not spend a lot.
Marketing Strategies: You’ve already succeeded in creating loyalty. Focus on
increasing monetization through product recommendations based on past pur-
chases and incentives tied to spending thresholds (pegged to your store AOV).

11
3.3 Code Implementation

12
13
Mail System

14
Email Received

Coupon on website

15
3.4 Work Flow

16
Figure 3.1: Project Roadmap

17
Chapter 4

Conclusion

The primary goal of this project is to develop a system that categorises the cus-
tomers on the basis of the recency - that is how recent did they purchase an item,
frequency - that is how frequent do they purchase and monetary - that is how much
money are they willing to spend. This categorisation allows the business owners
to develop unique and personalized marketing scheme for each of the cluster of
customer segments. This project makes use of RFM analyses as well as k means
clustering.
Furthermore this categorisation is used for generating coupons. These coupons
will be generated specifically with respect to customers purchase history and RFM
results leading to custom coupons for every group of customers.

18
Chapter 5

Future scope

A relevant implementation to the current project would be to add a marketing

technique that profits the company’s revenue. When a store issues promotional
codes, they’re providing customers with an incentive to buy, which benefits both
the customer and the business. The coupons The discount can either be a per-
centage or a specific dollar amount. Promo codes also provide customers with free
shipping or gift-wrapping. This marketing strategy essentially gives customers yet
another reason to buy your products. These incentives even have a direct impact
on the overall shopping experience, according to research conducted by the Center
for Neuroeconomics Studies at Claremont Graduate University.
Furthermore promotional codes are an effective way to track your marketing
efforts to see which platforms are generating the most traffic or leading to conver-
sions, according to Inc. magazine. Store owners can track their efforts by attaching
a certain promo code to a specific marketing campaign or advertisement. Hence
creating a database by analysing the user experience with the promotional codes
can add more

19
References

20
 Acknowledgement

I have a great pleasure to express my gratitude to all those who have I would
like to express my deepest appreciation to all those who provided me the pos-
sibility to complete this report. A special gratitude I give to our project guid,
Ms Ankita Nagmote, whose contribution in stimulating suggestions and encour-
agement, helped to coordinate project especially in writing this report and has
invested her full effort in guiding the team in achieving the goal..
A special thanks goes to the team, Pranjal Ambulkar, Naziya Inamdar, Khush-
numa Shaikh , who helped to assemble the project and gave suggestions to improve
the project. Last but not least, I have to appreciate the guidance given by other
supervisors as well as the panels especially in our project presentation that has
improved our presentation skills thanks to their comment and advice.
Finally, we are very thankful to Usha Mittal institute of technology, for giving
us a chance to complete our thesis project within the scheduled time.

Date:

Pranjal Ambulkar - 01

Naziya Inamdar - 24

Khushnuma Shaikh - 56

21