italiano | español


Churn prevention scoring models


Date: 20-22 June 2012
Where:Nunatac’s premises
Registration: fill out the downloadable form and send it by fax to +39 02 89 01 20 74

The course follows an entire analytical process aimed at constructing a churn prevention model and the successive hypothesis on the use of the results to support a company’s retention strategies.
The churn prevention models are used to estimate churn probabilities (e.g. cancelling or not renewing a subscription, closing a bank account, early financing pay-off) or to predict the reduction of the use of a product or service (e.g. dramatic reduction of telephone traffic on a SIM card or of credit card use).
Like all scoring models, churn prevention ones too estimate the probability that the organization/client carries out or doesn’t carry out a certain action (renewal/non renewal), or assumes or doesn’t assume a certain state (active/ceased). This probability is estimated on the basis of a range of client characteristics and behaviours which precede the “abandonment”.

Who should attend
Statistical analysts, data mining experts, business users.

You should have basic experience in the use of the SAS language and general experience in multivariate analysis statistical techniques. You should also have knowledge of the Enterprise Mining tool.

Course outline
• Company context
• Problem Aims and planning
• Analysis population
• Available data for the definition of target variable and explanatory variables
• Sample extraction
• Preliminary analysis of input variables: management of missing data, transformation of variables
• Outliers identification
• Model construction
• Assessment of discriminant model performance
• Whole population score
• Application hypotheses and tests on other population
• List extraction

The statistical techniques used are: logistic regression, decision trees.

The duration of the course is 3 days.

Public Course