TOPIC: Churn Rate Analysis of

Telecommunications Customers

FOR: Prof. Santanu Das

Subject: Data Processing and Visualization

By:
Deepanwita Chowdhury (22PGDMHR12)
Dhairya Vijendra Mehta (22PGDMHR13)
Pranjal Sud (22PGDMHR35)
Siddhartha Bindyeshwari (22PGDMHR
ABSTRACT

INTRODUCTION

Retaining customers is one of the most crucial factors and a primary KPI for any company. With
increasing competition and an almost perfectly competitive market, retention of customers is
becoming more and more difficult. Having a retained set of customers is more cost-effective
than acquiring one. A mere 5% boost in customer retention elevates the company revenue by
as large as 25-30% (Thomson and B, 2005).

As has been shown in the customer retention funnel, the path from being just another
consumer to being a loyalist is a long path with a large number of hurdles for companies.
Providing a trusted platform, offering customized services, and repeated feedback are some of
the steps which help increase the retention rate (Rozenburg et al., 1984).

The rate at which consumers cease doing business with a company is known as the churn rate,
sometimes referred to as the rate of attrition or customer churn. Most frequently, it is
represented as the percentage of service subscribers who cancel their memberships within a
specified time frame. It is also the frequency of employees quitting their positions within a
specified period of time. A company's growth rate (determined by the number of new
customers) must be higher than its attrition rate in order to grow its customer base (Cimpoeru et
al., 2014).

There are certain factors that affect the high churn rate:
1. Natural Causes: certain uncontrollable factors like the untimely death of consumers,
geographic shift, fraudulent behaviors or untimely payment of dues lead to an increased
churn.
2. Customer Dissatisfaction: a gap between customer expectation or promise and the
actual delivery of service can lead to high levels of discontent if the promise is less than
the expectation. Only a very small percentage of customers raise a complaint while the
rest silently churn.
3. Customer disconnect: High customer intimacy leads to higher loyalty and vice versa.
With an increasingly digitized and automated world, the element of the human touch is
diminishing every passing day. Keeping the bond with the consumers alive is a
challenge that organizations need to overcome.
4. Better, Bigger, Faster, and More: Customers are tailored to expect more than they get.
With the world transforming into a global village and everything being within the reach of
individuals, the expectation level is on the rise which if not fulfilled leads to an increased
churn rate.
As per the data on Statista, energy or utilities (89%), IT services(88%), computer services(86%),
and telecommunications(69%) are some of the top industries facing the highest churn rates.
Having multiple options to choose from, the telecom industry is not a popular industry in terms
of customer retention. Customer loyalty is one of the key profitability drivers for the industry.
With complicated billing processes, major connectivity and internet issues as well as issues with
the service provider in terms of customer care services, the stability in terms of retention is
highly volatile leading to a high churn rate.

The study focuses on the churn rate in the telecom industry. The sample involves a combination
of both genders, different age groups, and the availability of different services which might
impact their judgment. With the implementation of CRM, intending to reduce the churn rate, the
industry is still facing an elevated rate. Studies have shown that network quality is one of the
primary necessities for a user followed by booster facilities. For the working class, internet
service and the latest facilities are the driving factors determining the churn rate (Babu et al.,
2014).

REVIEW OF LITERATURE

The telecommunications industry is highly competitive, and retaining customers is crucial for
business growth. Customer churn prediction is an essential aspect of the industry, as it helps in
identifying the factors that lead to customer churn and devising strategies to retain them. In
recent years, there has been a growing interest in developing machine learning-based models
for customer churn prediction.

Several studies have been conducted to explore different datasets, methods, and metrics for
customer churn prediction in the telecom industry. A study by Huang et al., (2019) provides a
comprehensive review of various techniques used for customer churn prediction. They explore
different machine learning algorithms, such as logistic regression, decision trees, and support
vector machines, and discuss their advantages and limitations for churn prediction in the
telecom industry. They also discuss the importance of feature selection, data preprocessing,
and model evaluation in developing accurate churn prediction models.

Another study by Umayaparvathi and Iyakutti,2018 examined the relationship between customer
churn and various attributes in the telecom industry and proposed a method for attribute
selection in the prediction model. They analyze a dataset of telecom customers and identify the
most important attributes that affect customer churns, such as call rates, call duration, and
customer complaints. They then develop a method for attribute selection based on mutual
information and correlation analysis, which improves the accuracy of the prediction model.
(Yabas et al. 2018) propose a machine learning-based approach for customer churn prediction
in telecom services. They employ several machine learning algorithms, such as decision trees,
Naive Bayes, and k-nearest neighbors, and compare their performance in predicting customer
churn. They also discuss the importance of data preprocessing, feature selection, and model
evaluation in developing accurate churn prediction models.
(Ebrah and Elnasir 2019) develop a churn prediction model using machine learning and
recommendation plans. They employ several machine learning algorithms, such as random
forest, logistic regression, and support vector machines, and combine them with
recommendation plans to improve the accuracy of the prediction model. They also discuss the
importance of data preprocessing, feature selection, and model evaluation in developing
accurate churn prediction models.

(Lalwani et al. 2019) propose a customer churn prediction system based on a machine learning
approach. They employ several machine learning algorithms, such as decision trees, support
vector machines, and artificial neural networks, and compare their performance in predicting
customer churn. They also discuss the importance of feature selection, data preprocessing, and
model evaluation in developing accurate churn prediction models.

(Zhang et al. 2018) focus on improving customer churn prediction by segmenting telecom
customers and developing a data-driven approach. They analyze a dataset of telecom
customers and identify different customer segments based on their usage patterns and
behavior. They then develop a data-driven approach that improves the accuracy of the
prediction model by considering the characteristics of each customer segment.
(Oghojafor et al. 2017) employ discriminant analysis to identify the factors that affect telecom
customer churn. They analyze a dataset of telecom customers and identify the most important
factors that lead to customer churns, such as call quality, network coverage, and customer
service. They then develop a model based on discriminant analysis that accurately predicts
customer churn in the telecom industry.

In addition to the studies mentioned above, there have been several other studies that have
explored different aspects of customer churn prediction in the telecom industry. For instance,
(Chen et al. 2018) propose a hybrid model that combines machine learning and fuzzy logic to
predict customer churn. (Wang et al. 2019) develop a churn prediction model based on deep
learning algorithms, such as convolutional neural networks and long short-term memory
networks.

Another aspect of customer churn prediction that has gained attention in recent years is the use
of big data analytics. Big data analytics can help in analyzing large amounts of data and
identifying hidden patterns and trends that can be used in developing accurate churn prediction
models. For instance, (Chen et al. 2019) develop a big data analytics-based churn prediction
model that analyzes data from different sources, such as social media, customer feedback, and
call logs.

Overall, these studies highlight the importance of developing accurate and efficient customer
churn prediction models for the telecom industry. The datasets, methods, and metrics used in
these studies can serve as a basis for further research in this field. The development of
accurate churn prediction models can help telecom companies in retaining their customers and
improve their business growth.
Another area of focus could be on developing models that are more personalized to individual
customers. While many of the studies above focus on developing models for predicting churn at
a high level, there is value in developing models that can predict churn at the individual
customer level.

In conclusion, the telecom industry experiences a higher churn rate of customers compared to
other industries due to several factors. Firstly, the industry is highly competitive, with numerous
service providers offering similar services at competitive prices. This makes it easy for
customers to switch providers if they are dissatisfied with the service. Additionally, the industry
experiences rapid technological advancements, which lead to frequent upgrades and changes
in service offerings, making it challenging for customers to keep up with the changes.

Furthermore, the quality of customer service in the telecom industry is often lacking, with long
wait times and impersonal service. This can leave customers feeling undervalued, and they may
opt to switch to providers that offer better customer service. Lastly, the lack of loyalty programs
and incentives for customers can also contribute to a higher churn rate in the telecom industry.

Overall, to mitigate the high churn rate in the telecom industry, service providers must prioritize
customer satisfaction, offer personalized service, and provide incentives and loyalty programs to
encourage customer retention.

DATASET AND VARIABLES

The data selected for the above research is available on Kaggle. The data contains 19
independent variables portraying the behavior of customers of a dummy telecommunication
company.
The information can be categorized as follows:

A. Demographic Information:
1. Gender: to reveal whether the interviewee is male or female
2. Senior Citizen: whether the client is a senior citizen or not
3. Dependents: whether they have a dependent or not
4. Partner: whether they have a partner or not

B. Services Information
1. Phone Service: Whether they have a usable phone service
2. Online Backup: Whether there is scope for online backup
3. Internet Service: Whether there is a subscription to an internet service
4. Online Security: Whether there is a presence of online security
5. Device Protection: Whether the customer has device protection
6. Multiple Lines: Whether the customer has only a single network or multiple’
7. Tech Support: Whether there is tech support or not
 8. Streaming Movies: Whether there is streaming of movies
9. Streaming TV: Whether there is streaming of TV

C. Customer Account Information

1. Tenure: Duration of stay of the customer with the company
2. Paperless Billing: Whether they have paperless billing or not
3. Contract: Whether month-t-month, one year or two years
4. Payment Method: Whether electronic, bank transfer, or other forms
5. Total Charge: Total billing amount
6. Monthly Charge: Monthly billing amount

HYPOTHESIS

Churn Rate, also sometimes known as Attrition rate, is the average rate at which customers
stop doing business with a company over a specified period. It is one of the prime concerns for
any company as with increasing competition it is always costlier to acquire new customers than
to retain old ones. So, companies are majorly focused on retaining old customers. Thus, a high
churn rate turns out to be a major cause for concern for the companies. The churn rate is
impacted by various factors. Some of them could be psychosocial parameters like the number
of people in a family, age, gender, etc. or they could be impacted by extraneous variables like
Changes in technology, and socio-economic conditions.

Our aim through this study is to determine to what extent psychosocial parameters impact the
churn rate of customers (if they at all) and then through the statistical analysis of data offer
suggestions on how could companies take into consideration these factors to reduce the churn
rate.

Sample: We have based our study and inferences on a sample size of 7043 people and their
answers to a questionnaire based on the Telecomm industry.

Null Hypothesis1: Demographics have a direct relation with the churn rate
Alternate Hypothesis1: Demographics do not have an impact on churn rate
Dependent Variable: Churn Rate
Independent Variable: Demographics (Gender, Age, Partner, Dependents)

Null Hypothesis 2:

METHODOLOGY
The data was downloaded from Kaggle following which data cleaning and analysis was done on
excel.
The graphs presented in the report were created using Python.

RESULTS AND ANALYSIS

After conducting analysis in excel and creating the graphs in python the following results could
be drawn from the data studied.

The following bar graph shows that more than 70% of the general observance do not go for
churn and opt for the same telecommuting network while around 30% churn out for various
reasons.
From the demographic information provided in the dataset, the graphical representation helps
draw the following conclusions:
1. Gender is not a major differentiator for the churn rate. Both males and females have an
almost equal percentage of churn. Thus, gender is not a significant predictor. While
gender may be associated with certain consumer preferences or behaviors, such as
usage patterns or preferences for certain types of services or devices, these factors are
generally not strong enough to predict churn rates on their own.

2. Senior citizens have a churn rate almost double that of the younger generation where 0
is representative of the younger generation and 1 is for the senior citizens. The reason
why senior citizens have over double the churn rate as compared with other consumers
could be because senior citizens are less comfortable with new technology, which can
make it harder for them to navigate and use their telecom services with increased
updating and advancement. This in turn could lead to frustration and a higher likelihood
of switching to a different provider or service. Also, Senior citizens may be on a fixed
income, which can make them more price-sensitive and more likely to switch to a
provider that offers a better deal or discount.

3. Customers without a partner churn more than those with a partner. Customers who have
a partner are more likely to have social ties and connections in their local community,
which can create a sense of belonging and loyalty to their telecom service provider.
They may also have shared financial responsibilities, making them more invested in
maintaining a stable and reliable telecom service.

4. Lastly, those without a dependent almost have double the churn rate than those with a
dependent. Telecom providers often offer value-added services, such as family plans or
shared data plans, which are more appealing to customers with partners. These services
can offer more value and convenience to customers who share a household and need to
manage multiple devices and accounts.
 Independents tend to have higher churn because they possess the free will to change
their provider in case they are nit satisfied with the same, which is not possible in the
case of dependents.

Coming to the customer account information, the following can be drawn:

1. Those with a month-to-month contract tend to have a significantly higher churn rate than
those with a yearly contract which again is more than those with a two-year contract.
The reason could be Service providers often offer promotions and discounts to
customers who sign up for longer-term contracts. Customers with shorter-term contracts
may not have access to these offers, making it less appealing for them to stay with the
service provider. Also, Customers with longer-term contracts may have had a more
positive experience with the service provider, leading to a higher level of customer
satisfaction and loyalty.

2. Customers opting for a paperless electronic mode of billing have almost 20% higher
tendency to churn than those with the traditional billing method. Customers who opt for
electronic billing may experience technical issues such as difficulty in accessing their
bills online, email delivery problems, or issues with the billing system. These issues can
frustrate customers and lead them to switch to another service provider. Customers may
have concerns about the security of their personal and financial information when using
electronic billing. If they do not trust the security measures put in place by their service
provider, they may switch to another provider that offers more secure billing options.

3. The customers with electronic check forms of payment have the highest churn rate while
the rest of the payment methods show an approximately equal pattern. Some customers
may not trust electronic checks as a payment method, either because they have had a
negative experience with them in the past or because they simply prefer more traditional
payment methods like credit cards or cash. Customers who lack trust in electronic
checks may be more likely to switch to a different payment method or provider.
Analyzing the services attributes, the following conclusions can be drawn:
1. Phone services and phone lines do not impact the decision to any significant extent. The
reason, why traditional phone services and phone lines are less important, is the rise of
internet-based communication technologies such as Voice over IP (VoIP) and
messaging apps. These technologies allow people to make voice and video call over the
internet, often at a lower cost than traditional phone services. They also offer a range of
other features such as file sharing, group messaging, and video conferencing.

2. Those with fiber optic internet services have a higher churn rate than the rest.
Customers who have fiber optic internet services often pay a premium price for their
services, which may make them more likely to switch to a cheaper provider if they
become dissatisfied or if they find a better deal elsewhere. Fiber optic internet services
are not available in all areas, which means that customers who move to a location where
these services are not available may have to switch to a different provider.

3. Those without any form of online security or device protection have as high as a 50%
churn rate. If a customer has a bad experience with their telecom provider, they are
more likely to switch to a competitor. Therefore, customers without any form of online
security or device protection are more likely to churn because they are more likely to
experience issues that negatively impact their perception of their telecom provider.

4. Those without any form of tech support, streaming TV or movies have a 50% churn rate
compared to other forms. Without tech support, customers may be more likely to
experience issues with their service, such as poor signal strength or connectivity
problems. If these issues are not resolved in a timely manner, customers may become
frustrated and switch to another provider. With the rise of streaming services and the
increasing importance of technology in daily life, consumers may be more likely to
prioritize providers that offer tech support and access to streaming TV or movies.
CONCLUSION:

Through the statistical analysis of the data set, we came to the following conclusions:

· That gender has a very statistically insignificant impact on the churn rate

· Senior citizens tend to churn significantly more often as opposed to young people.
One of the reasons to justify this inference could be that the age of approximately
60-65, people are more conservative towards budget issues. Retirement and death
could be two major factors leading to this result. This is a common trend across the
telecom industry.

· Third inference that we drew is that people with partners or dependents in their
family tend to be more stable in their choices and hence churn less than people with
no partners or dependents. This could be justified by the fact that perhaps the sense
of responsibility tends to make them more stable.

Suggestions: Companies that offer products or services that remain stable relatively over a
period could focus more on audience segregation based on the above psychosocial
parameters. This would ensure a lower churn rate from a major segment of their customers
(who have partners) and reduce the overall churn rate as well.

Similarly, companies making a relatively cheaper easier-to-use segment of their product for
senior citizens might not be a bad ploy to retain them.
REFERENCES

Thompson, B. (2005). The loyalty connection: Secrets to customer retention and increased
profits. RightNow Technologies & CRMguru, 18.

Rosenberg, L. J., & Czepiel, J. A. (1984). A marketing approach for customer retention. Journal
of consumer marketing.

Cimpoeru, C., & Andreescu, A. (2014). Predicting Customers Churn in a Relational Database.
Informatica Economica, 18(3).

Babu, S., Ananthanarayanan, D. N., & Ramesh, V. (2014). A survey on factors impacting churn
in telecommunication using datamininig techniques. International Journal of Engineering
Research & Technology (IJERT), 3(3).

Sci-Hub | Customer Churn Prediction for Telecom Services. 2012 IEEE 36th Annual Computer
Software and Applications Conference | 10.1109/COMPSAC.2012.54. (n.d.). Sci-Hub |
Customer Churn Prediction for Telecom Services. 2012 IEEE 36th Annual Computer Software
and Applications Conference | 10.1109/COMPSAC.2012.54.

https://sci-hub.se/https://ieeexplore.ieee.org/abstract/document/6340176
Ebrah, K., Elnasir, S., & Publishing, S. R. (2019, November 4). Churn Prediction Using Machine
Learning and Recommendations Plans for Telecoms. Churn Prediction Using Machine Learning
and Recommendations Plans for Telecoms. https://doi.org/10.4236/jcc.2019.711003

Sci-Hub | Customer churn prediction system: a machine learning approach. Computing |

10.1007/s00607-021-00908-y. (n.d.). Sci-Hub | Customer Churn Prediction System: A Machine
Learning Approach. Computing | 10.1007/s00607-021-00908-y.
https://sci-hub.se/https://link.springer.com/article/10.1007/s00607-021-00908-y
Sci-Hub | Attribute selection and Customer Churn Prediction in telecom industry. 2016
International Conference on Data Mining and Advanced Computing (SAPIENCE) |
10.1109/SAPIENCE.2016.7684171. (n.d.). Sci-Hub | Attribute Selection and Customer Churn
Prediction in Telecom Industry. 2016 International Conference on Data Mining and Advanced
Computing (SAPIENCE) | 10.1109/SAPIENCE.2016.7684171.

https://sci-hub.se/https://ieeexplore.ieee.org/abstract/document/7684171
Zhang, T., Moro, S., & Ramos, R. F. (2022, March 16). A Data-Driven Approach to Improve
Customer Churn Prediction Based on Telecom Customer Segmentation. MDPI.
https://doi.org/10.3390/fi14030094

Customer churn prediction in telecommunications. (2011, August 12). Customer Churn

Prediction in Telecommunications - ScienceDirect. https://doi.org/10.1016/j.eswa.2011.08.024

Umayaparvathi, V., & Iyakutti, K. (2018). A survey on customer churn prediction in telecom
industry: datasets, methods and metrics.