Scribd Logo
Upload

Welcome to Scribd!

  • 632 views

    Question 1 Is Wall's Belief About The Dependence of Churn Rates On Customer Age Supported by The Data?

    Uploaded by

    jhadarshna
    The document discusses Wall's belief that customer churn rates depend on age and whether this is supported by the data. Several models are analyzed to predict churn probabilities for specific customers. The best model suggests churn is highest for customers aged 6-14 and identifies the top drivers of churn for different age clusters. The top 100 highest risk customers and their key drivers are provided in an Excel sheet.

    Copyright:

    © All Rights Reserved
    Download as PPTX, PDF, TXT or read online from Scribd

    Question 1 Is Wall's Belief About The Dependence of Churn Rates On Customer Age Supported by The Data?

    Uploaded by

    jhadarshna
    632 views9 pages
    The document discusses Wall's belief that customer churn rates depend on age and whether this is supported by the data. Several models are analyzed to predict churn probabilities for specific customers. The best model suggests churn is highest for customers aged 6-14 and identifies the top drivers of churn for different age clusters. The top 100 highest risk customers and their key drivers are provided in an Excel sheet.

    Original Title

    BA

    Copyright

    © © All Rights Reserved

    Available Formats

    PPTX, PDF, TXT or read online from Scribd

    Did you find this document useful?

    Is this content inappropriate?

    The document discusses Wall's belief that customer churn rates depend on age and whether this is supported by the data. Several models are analyzed to predict churn probabilities for specific customers. The best model suggests churn is highest for customers aged 6-14 and identifies the top drivers of churn for different age clusters. The top 100 highest risk customers and their key drivers are provided in an Excel sheet.

    Copyright:

    © All Rights Reserved
    Download as PPTX, PDF, TXT or read online from Scribd
    Download as pptx, pdf, or txt
    632 views9 pages

    Question 1 Is Wall's Belief About The Dependence of Churn Rates On Customer Age Supported by The Data?

    Uploaded by

    jhadarshna
    The document discusses Wall's belief that customer churn rates depend on age and whether this is supported by the data. Several models are analyzed to predict churn probabilities for specific customers. The best model suggests churn is highest for customers aged 6-14 and identifies the top drivers of churn for different age clusters. The top 100 highest risk customers and their key drivers are provided in an Excel sheet.

    Copyright:

    © All Rights Reserved
    Download as PPTX, PDF, TXT or read online from Scribd
    Download as pptx, pdf, or txt
    of 9

    QUESTION 1

    Is Wall’s belief about the dependence of churn rates on customer age supported by the data?

    The histogram clearly shows that churn rate of people in


    the age group 10-14 are higher than others
    QUESTION 1

    Is Wall’s belief about the dependence of churn rates on customer age supported by the data?
    Contd..

    The jitter plot shows that :-


    1. The number of customers of the age group 0 –
    10 is very high when compare to those of
    higher age groups.
    2. It is also seen that the majority of churn is seen
    among the people of age group 6 - 14
    QUESTION 1

    Is Wall’s belief about the dependence of churn rates on customer age supported by the data?
    Contd..

    Age.Group
    <6 6-14 >14 This bar chart shows that customers between
    10 0

    the age group 6 and 14 discontinue more than


    any other age group
    80

    From the initial graphs we can conclude that


    60

    Wall’s belief is on the right track


    P erce n t

    40
    20
    0

    No Yes

    Churn..1...Yes..0...No.
    QUESTION 2.A

    Run a single regression model that best predicts the probability that a customer leaves.
    a. What is the predicted probability that Customer 672 will leave between December 2011 and February
    2012? Is that high or low? Did that customer actually leave?

    The probability that the customer will leave between Dec2011 and
    Feb2012 is 3.83%. It is Low.
    From the dataset, it can be seen that the customer did not actually
    leave.
    QUESTION 2.B

    Run a single regression model that best predicts the probability that a customer leaves.
    B. What about Customers 354 and 5,203?

    The probability that the customer 354 will leave between Dec2011
    and Feb2012 is 4.85%. It is Low.
    The customer did not actually leave.

    The probability that the customer 5203 will leave between


    Dec2011 and Feb2012 is 4.20%. It is Low.
    The customer did not actually leave.
    QUESTION 3.
    How sensible is the approach with a single model? Can you suggest a better approach?
    a) Provide updated estimates of probabilities that Customers 672, 354, and 5,203 will leave.
    b) What factors contribute the most to the predicted probabilities that these customers will leave?

    The probability that the customer 672 will leave


    between Dec 2011 and Feb 2012 is 3.45% according to
    the model V3. It is Low.
    And the dataset shows that customer did not actually
    leave.
    The older model gave a probability of 3.83%. Hence
    this new model is not better.
    QUESTION 3.
    How sensible is the approach with a single model? Can you suggest a better approach?
    a) Provide updated estimates of probabilities that Customers 672, 354, and 5,203 will leave.
    b) What factors contribute the most to the predicted probabilities that these customers will leave?

    The probability that the customer 354 will


    leave between Dec2011 and Feb2012 is
    8.06% according to the model V2. It is Low.
    The customer did not actually leave.
    The older model gave a probability of 4.85%.
    Hence this new model is not better.
    QUESTION 3.
    How sensible is the approach with a single model? Can you suggest a better approach?
    a) Provide updated estimates of probabilities that Customers 672, 354, and 5,203 will leave.
    b) What factors contribute the most to the predicted probabilities that these customers will leave?

    The probability that the customer 5302 will


    leave between Dec2011 and Feb2012 is 0.72%
    according to the model V1. It is very Low.
    The customer did not actually leave.
    The older model gave a probability of 4.20%.
    Hence this new model is much better.
    QUESTION 4.
    Answer Wall’s “ultimate question”: provide the list of 100 customers with highest churn probabilities
    and the top three drivers of churn for each customers.

    The Excel sheet contains the top 100 customers who are most
    likely to leave.
    Microsoft Excel
    The Key Drivers for :- Worksheet
    1. For Cluster 1(customer age <6)
    a) CHI.Score.0.1
    b) Days.Since.Last.Login.0.1
    c) CHI.Score.Month.0
    2. For Cluster 2(customer age >=6 and <14)
    a) CHI.Score.Month.0
    b) Customer.Age..in.months
    c) Days.Since.Last.Login.0.1
    3. For Cluster 3(customer age >=14)
    a) CHI.Score.Month.0
    b) Customer.Age..in.months
    c) Views.0.1

    You might also like