This study attempted to formulate a regression model that identifies the characteristics that influence whether a customer is probable to switch telecommunications providers (Churn). We started with a full model, performed variable selection using AIC and BIC through the stepwise method and moved on to other models, using LASSO, or a few simple (aggregate) transformation of certain predictor variables. We concluded that the best logistic regression model we could find was produced with an aggregate transformation of the variables that concern domestic charges for various times of the day (Day Charges, Evening Charges, Night Charges).
2. The Problem
Over the previous period
~15%
Churn Rate
*based on a sample of 3333 customers
Which customers are probable to churn?
Our preliminary analysis shows, that:
§ Gender and Area Code don’t seem to
play a role
§ There is some variability between States,
but didn’t know if it’s important
§ Having an International Plan and a Voicemail
Plan seem to be important factors
3. Deeper Analysis
using Logistic Regression
We attempted to formulate a model with a good fit, that
pinpoints the characteristics of customers probable to churn.
Model_ID Description McFaddens R^2
Cox and Snell
R^2
Nagelkerke R^2
Hosmer
Lemeshow
p-value
Model15
Unifying the charges under one variable “Domestic.Charge” > Stepwise with
AIC (multicollinearity fixed) 0.257 0.185 0.338 0.247
Model2
Starting with all the variables, and performing Stepwise Selection with AIC
(multicollinearity fixed)
0.258 0.186 0.338 0.117
Model3 Starting with all the variables, and performing Stepwise Selection with BIC 0.258 0.186 0.338 0.117
Model9
Unifying the number of calls under one variable “Domestic.Calls” > Stepwise
with AIC (multicollinearity fixed)
0.258 0.186 0.338 0.117
Model6
Unifying the minutes under one variable “Domestic.Mins” > Stepwise Selection
with AIC (multicollinearity fixed)
0.258 0.186 0.338 0.108
Model12
Domestic.Calls + Domestic.Mins (aggregates) > Stepwise with AIC
(multicollinearity fixed)
0.258 0.186 0.338 0.108
4. We managed to produce a model with a good fit and only 6 variables,
that ranks well in all key metrics and has an accuracy of 88%.
Variable Estimate Std. Error Significance 95% Conf. Interval
(Intercept) -9,18 0,51 *** [-10.20 , -8.20]
CustServ.Calls 0,54 0,04 *** [0.46 , 0.62]
Has_Int.l.Plan 2,08 0,15 *** [1.78 , 2.38]
Has_VMail.Plan -1,29 0,16 *** [-1.62 , -0.97]
Intl.Calls 0,15 0,03 *** [-0.21 , -0.10]
Intl.Charge 0,39 0,08 *** [0.24 , 0.55]
Domestic.Charge 0,10 0,01 *** [0.08 , 0.11]
The model we produced
Confirming our preliminary analysis
§ Having an International Plan and a Voicemail
Plan are indeed of major importance
§ The number of Customer Service Calls is a
worrying indicator
§ Regular (domestic) as well as international
charges play an important role
5. Customer A has 1 more Customer
Service Call than Customer B
Customer Service Calls Having an International Plan Having a Voicemail Plan
# of International Calls International Charges Domestic Charges
Customer A has an International Plan.
Customer B doesn’t.
Customer A has a Voicemail Plan.
Customer B doesn’t.
Customer A has 1 more
International Call than Customer B
Customer A has 10 more USD in
International Charges than Customer B
Customer A has 10 more USD in
Domestic Charges than Customer B
A
B
Odds of Churn for A are
1,7 times higher
than customer B
A
B
Odds of Churn for A are
8 times higher
than customer B
A
B
Odds of Churn for A are
70% lower
than customer B
A B
Odds of Churn for A are
14% lower
than customer B
A
B
Odds of Churn for A are
14,8 times higher
than customer B
A
B
Odds of Churn for A are
11,1 times higher
than customer B
6. Review the characteristics of our international plans
and make them better
Compare international and domestic rates
with our competitors and adapt
Promote VoiceMail Plans as they build switching barriers
Implement a “pampering” period for customers with 2 or more
recent customer service calls to get us back in their good graces
How can we solve the problem?
7. Thank You
for your attention
Sotiris Baratsas
sotbaratsas@gmail.com
MSc in Business Analytics