Cell2Cell Data Documentation

Download as doc, pdf, or txt
Download as doc, pdf, or txt
You are on page 1of 2

PADM Assignment no.

2: CELL-2-CELL

(Incorporate SAS EM screenshots in the document to explain the results of various models)
Dataset: CELL
SAS Enterprise Miner Settings:
Input Data Source: In step 4 of Create Datasource , click on advanced option and then next. Set
role of CHURN as a Target variable and Set the role of CSA, Customer, Churndep and Calibrat
to rejected. Set the role of variable Id as Id (and level as nominal). Convert the level of other
9 nominal variables to interval.
At step 7, Make sure that there are 79 columns and 71,047 records.
At Step 9, make sure that there is one Id variable, 38 binary input variables, 35 interval input
variables, 4 rejected variables and 1 target variable)
Data Partition: Use the default partitioning method with dataset allocations as 60% for training
and 40% for validation.

1. What are the business objective(s) and data mining objective(s) of the project?
2. Run at least 4 models on SAS - Decision Trees, Logistic Regression, Logistic Regression
with Transform Variables and Logistic Regression with transform variables and interactive
binning and report / compare the cumulative lift value and ROC results .
3. What are the transformation of variables that you tried and Why? What were the impact
of the same on the results/ predictive performance of the regression? From the model
comparison node, enumerate the impact of transform variables on predictive
performance/ cumulative Lift Value (at top 10 percentile) for regression? And compare
the same to the predictive performance/ Cumulative Lift value for Decision Trees? (your
objective should be to improve the results of the regression model as much as possible)
4. Assuming that you would be going ahead with the Final Logistic Regression model,
identify the significant variables/ key factors that predict customer churn (based on t-
scores)? Do these factors make business sense? Why or why not? Which of the
significant variables/ factors are ACTIONABLE (hint: from the point of view of conversion
into incentives) and why?
5. Using a few examples (Decision tree and regression results/ drivers of churn, incentive Cost/ LVC
calculations) explain how would you arrive at the offers/ offer bundles and the incentive cost/ value
of offers that you would make to each customer?
7. You are now asked by your team head to try a different novel data mining modeling
approach that uses both unsupervised learning, i.e clustering and supervised learning,
i.e Decision Tree/ Logistic regression. Explain the Data mining model(s)/ modeling
approach and how the same can achieve better results (w.r.t success criteria of DM and
Business Objectives).
8. Assuming the case is set in 2018 and not 2002, what could be some additional key input
significant variables that you would need to incorporate in the model?
Cell-2-Cell Case (Profit and Incentive Cost Calculations)

Profitability of A Proactive Retention Plan

•  = Baseline churn rate = .02


 = Lift = 1.75 (at top 10 percentile. Assume this value – actual value would be higher)
 = Success rate (fraction of contacted churners who are prevented from churning). E.g for
Lift Value 1.75,  = 1.75-1 =.75 (assuming a 100% success rate)
LVC = Lifetime value of customer
C = Cost of incentive

 Profit = Probability Churner x Success Rate x LVC 1 – C


=  x  x  x LVC – C
= LVC – C

1
LVC = Monthly Revenues x 12* (1+Discount Rate) / (1+Discount Rate – Retention Rate 2), from
simple retention model. Assume Discount rate = 10%.

2
Please note that the retention rate is the annual retention rate, i.e .785

Further Hints:

1) Set Profit = 0 to calculate the max. incentive cost for the customer
2) Use the LVC/ incentive cost calculations and Lift value/ drivers of churn from the DM model to
design an incentive plan

You might also like