Project Title: Customer Churn Analysis
Sl. No Description Source Details & Status
Module Solution
1 Project Project Role: Data Complete
Definition & Planning Analyst
Scope Objective:
Analyze
historical
customer data
(2020-2025)
to identify
churn drivers,
build a
predictive
model, and
provide
prescriptive
recommendati
ons to reduce
customer
attrition.
2 Data Loading Descriptive Action: Loaded Complete
Analysis the
customer_Chu
rnData(jio).xlsx
dataset into a
pandas
DataFrame.
Solution:
Ensured the
file path was
correct to
avoid
FileNotFoundE
rror.
3 Initial Data Descriptive Action: Complete
Exploration Analysis Reviewed the
first few rows
(.head()) and
data types
(.dtypes) to
understand
the structure
and features
of the dataset.
Result:
Identified a
mix of
numerical and
categorical
data, including
the target
variable
'Churn'.
4 Data Descriptive Action: Complete
Cleaning: Analysis Checked for
Duplicates and removed
duplicate rows
using
df.drop_duplic
ates().
Solution:
Ensured each
record was
unique,
preventing
bias in the
analysis.
5 Data Predictive Error: The Close
Cleaning: Analysis TotalCharges
Missing column
Values contained
missing values
and was
incorrectly
typed as an
object.
Solution:
Converted
TotalCharges
to a numeric
type, forcing
errors to NaN.
Then, imputed
the missing
values using
the median, a
robust method
against
outliers.
6 Feature Predictive Action: The Complete
Dropping Analysis CustomerID
column was
dropped as it
is a unique
identifier with
no predictive
value for the
model.
7.0 Descriptive Descriptive Objective: To Complete
Analysis & Analysis identify initial
EDA patterns and
form
hypotheses
about churn
drivers.
7.1 Churn Rate Descriptive Insight: Complete
Calculation Analysis Calculated the
overall churn
rate, which
was found to
be 24.94%.
Conclusion:
The significant
churn rate
confirmed the
business
problem and
the need for a
predictive
solution.
7.2 Demographic Descriptive Insight: Complete
Analysis Analysis Analyzed
churn rates
across
different Age
Groups and
Genders.
Conclusion:
Found no
significant
difference in
churn rates,
indicating
these are not
strong
predictors.
7.3 Contract & Descriptive Insight: Complete
Payment Analysis Customers
Analysis with Monthly
contracts have
a significantly
higher churn
rate (26.7%)
than those
with Annual
contracts
(19.3%). Cash
Recharge
payment
methods also
showed higher
churn.
7.4 Support Descriptive Insight: A Complete
Interaction Analysis clear positive
Analysis correlation
was found
between the
number of
customer
support calls
and the
likelihood of
churn.
8.0 Predictive Predictive Objective: To Complete
Modeling Analysis build and
evaluate
models to
predict which
customers are
likely to churn.
8.1 Class Predictive Error Close
Imbalance Analysis Description:
The target
variable
'Churn' was
imbalanced
(75% No vs.
25% Yes),
which could
lead to a
biased model.
Solution:
Applied the
SMOTE
(Synthetic
Minority Over-
sampling
Technique) to
the training
data to create
a balanced
dataset for
model training.
This was done
after the train-
test split to
prevent data
leakage.
8.2 Categorical Predictive Challenge: Close
Feature Analysis Categorical
Encoding features
needed to be
converted to a
numeric
format.
Solution: Used
One-Hot
Encoding
(pd.get_dummi
es) instead of
LabelEncoder
to avoid
creating an
artificial
ordinal
relationship
between
categories.
8.3 Model Predictive Action: Complete
Training & Analysis Trained three
Tuning models:
Logistic
Regression,
Random
Forest, and
XGBoost. Used
RandomizedS
earchCV to
efficiently tune
hyperparamet
ers.
8.4 Model Predictive Result: Complete
Evaluation & Analysis Random
Selection Forest and
XGBoost
significantly
outperformed
Logistic
Regression.
The Random
Forest model
was selected
as the best-fit
model due to
its high ROC-
AUC score
(87.0%) and a
strong balance
of Precision
(92.5%) and
Recall (68.3%).
8.5 Feature Predictive Action: Complete
Importance Analysis Extracted and
Analysis visualized the
feature
importances
from the best-
fit (Random
Forest) model.
Insight:
Identified
PaymentMeth
od,
CustomerSupp
ortCalls, and
ContractDurati
on as the top
predictors of
churn.
9.0 Prescriptive Prescriptive Objective: To Complete
Analysis Analysis convert
predictive
insights into
actionable
business
recommendati
ons.
9.1 Retention Prescriptive Challenge: Close
Strategy Analysis Model
Formulation predictions
needed to be
translated into
direct
business
actions.
Solution:
Developed a
series of
prescriptive
recommendati
ons based
directly on the
top feature
importances.
For example,
for "Contract
Duration," the
recommendati
on is to
"Incentivize
Long-Term
Contracts."
9.2 Dashboard & Prescriptive Challenge: The Close
Workflow Analysis client needed
Proposal a way to
operationalize
the insights.
Solution:
Proposed a
workflow
where the
model's
predictions
could be used
to create a
churn risk
score. High-
risk customers
would be
flagged for a
manual review
or targeted
with specific
retention
campaigns,
which could be
managed via a
BI dashboard.
9.3 Financial Prescriptive Action: Complete
Impact Analysis Estimated the
Analysis potential ROI
of
implementing
the proposed
retention
strategies.
Result: A 10%
reduction in
churn could
save
approximately
₹12.5 Million
annually,
providing a
strong
business case
for the project.