TelTek Project Submission

Name - Preksha Porwal

Student ID - 5320982
Q-1) What is the RFM score of customer 26 and customer 150.

For Customer 26: For Customer 150:

Recency (R) score: 1 Recency (R) score: 1

Frequency (F) score : 2 Frequency (F) score : 2

Monetary value (M) score: 4 Monetary value (M) score: 3

Q-2) Match the targeting strategies with the RFM scores given in the table.
RFM scores Targeting strategy
A. Customers with all the recency, frequency and monetary value i. TelTek should send them notifications about new TV shows
scores of 3 and 4 such as 4-4-4,4-4-3,3-4-3 being screened on its OTT service partner’s platform. In this way,
the firm can make the customers consume more data, eventually
pushing them to recharge more frequently.

B. Churned best customers with low recency scores and high ii. TelTek should ensure that these type of customers have a
frequency and monetary scores such as 1-4-4, 1-3-4, 1-3-3 great experience. Any Conflict or complaint should be resolved
promptly.

C. Customers with low-frequency scores and high recency iii. The firm should use upselling strategy to make these
and monetary value scores such as 4-1-4, 3-1-4, 3-1-3 customers upgrade their existing subscription or subscribe for
additional services.

D. Customers with low monetary value score and high recency iv. The company should develop a strategy to reacquire and re-
and frequency scores such as 4-4-1, 3-4-1,3-3-1 engage these type of customers by offering them incentives and
great customer experience

A-3,B-4,C-1,D-2
Q-3) Which cluster comprises of the most valuable customers? What targeting
strategy should the company use for these type of customers?.

Cluster:

The cluster comprising the most valuable customers is Cluster 4. These customers
have the highest Average Revenue Per User (ARPU) of $2,000. The targeting
strategy for these types of customers should focus on maximizing their lifetime
value and encouraging them to continue using premium services. This can include
personalized offers, exclusive deals, and dedicated customer support to ensure
their satisfaction and retention.
Q-4) Study each metric of cluster 2 and cluster 3. Which cluster is the best
among the two? What can you do to the other cluster to become the best
among the cluster 2 and 3?
Comparing Cluster 2 and Cluster 3:
•Cluster 2:
• Contract expiring in 2 weeks
• Tenure with company: 25 months - 36 months
• Broadband service: Yes Cluster 3 appears to be the better
• TV packages: Yes cluster among the two, with higher
• Postpaid plans: No
ARPU and slightly lower average
• Handset: No
• Avg no. of complaints: 15 - 20 complaints. To improve Cluster 2 to
• ARPU: $700 match Cluster 3's performance, the
•Cluster 3: company could focus on reducing
• Contract expiring in 2 months complaints, potentially by
• Tenure with company: 13 months - 24 months enhancing service quality or
• Broadband service: Yes
providing additional incentives to
• TV packages: Yes
• Postpaid plans: Yes increase loyalty.
• Handset: No
• Avg no. of complaints: 10 - 15
• ARPU: $900
Q-5) How would you approach cluster 3 to make them more profitable?

To make Cluster 3 more profitable, the company could consider the following approaches:

•Enhancing postpaid plans: Since Cluster 3 already has a higher ARPU and includes customers
with postpaid plans, the company could focus on upselling additional premium services or
introducing new features to these plans to increase their value.
•Retention strategies: Implementing targeted retention strategies to reduce churn within
Cluster 3, such as offering exclusive discounts or rewards for continued loyalty, personalized
recommendations based on their usage patterns, and proactive customer support to address
any issues promptly.
•Cross-selling opportunities: Identifying opportunities to cross-sell other services or add-ons
to customers within Cluster 3, such as premium TV packages, additional broadband services,
or value-added services like device protection plans or home security solutions, to increase
their overall spend and profitability.
Q-6) Which metrics will you use in selecting the best classification model?

Metric:

We would use the logistic regression model.

Since we only have to predict or classify the dependent variable which is
categorical in nature based on one or more independent variables
In selecting the best classification model for predicting customer churn, we
will primarily focus on the metrics of sensitivity and specificity. These metrics
provide insights into how well the model correctly identifies churned
customers (sensitivity) and non-churned customers (specificity), which are
crucial for effectively managing churn.
Q-7) In order to predict the customer churn, the firm is required to solve a
binary classification problem. Apart from logistic regression that you have
already seen, there are other models as discussed in the video. Based on the
confusion matrix (given on the platform) for each of these models, choose the
best method for building the propensity to churn model?
Classification model:

Based on the confusion matrices provided, the Random Forest model seems to
be the most suitable for building the propensity to churn model. It demonstrates
a higher true positive rate (130,000) and true negative rate (160,000) compared
to the other models, indicating better performance in identifying both churned
and non-churned customers. We can also go for decision tree model.
Q-8) From the churn analysis, which factors have a significant influence on
customer churn at a 95% confidence level?
Significant factors:

From the churn analysis, the factors with significant influence on customer
churn at a 95% confidence level are:
•Contract bucket
•Customer billing (paperless or not)
•Tenure group

These factors have p-values less than 0.05, indicating a statistically

significant relationship with customer churn.
Q-9) Which channels have a significant influence on the total unit sales at 95%
confidence level?

Significant Channels:

The Channels that have a significant influence on the total unit sales at 95%
confidence level are Print(GRP) and Search(Number of impressions).
Q-10) State the regression coefficients of the channels?

Regression Coefficients:

TV: -8.202800592

Radio: 4.670573449

Print: 39.29067183

Search: 8.459377585

Facebook: -0.882778746
Q-11) State the multiple linear regression equation based on the results
obtained?
Regression equation:

Y= 1688.831472 + 39.29067183(print) + 8.459377585(search)