Sales predictive model design

Client Needs & Objectives

Data teams provided an extraction containing clients data and commands. An extraction of the default variable was then transmitted.

The objectives were as follows:

  • Reconstruct & test a complete database from the two extractions provided
  • Assess the possibility of establishing a predictive model given the database

Our Approach

1st study:

  • Reconstruct & test a complete database from the two extractions provided. The quality of the database has then been tested and proved to be satisfactory
  • From the reconstructed database, the following was done:
    • Study of the distribution of actual and constructed variables
    • Univariate, multivariate and correlation analysis
    • Study on the ability to define a predictive model of default risk

2nd study:

  • Transform the explanatory variables retained in binary mode (0 or 1) then classify them by taking the sum of the binary variables
  • The results of such an exercise allow to conduct more thorough studies

Client Benefits & Main Results

Establish the good general quality of the database provided

Establish a typical client profile and command profile

Conduct univariate, multivariate and correlation analysis to highlight the explanatory variables to be used preferably in a predictive analysis

Realise the feasibility of the construction of a score given this database

The feasibility of predictive analysis (data mining & scoring) is established and they should deliver good results in terms of risk management and/or commercial pilot

The final use of such a classification hasn’t been discussed yet but will be