Design of a predictive recruiter success model (PRiSM)

Abstract

This thesis describes the formulation and validation of a multiple linear regression model that predicts recruiter success rates. The model's primary purpose is to improve the recruiter selection process by helping to reduce recruiter reliefs. Using recorded information on over 400 members of two active-duty recruiting battalions together with the results of an administered sales ability test, a database was constructed for use in regression analyses. Recruiter success was defined as the response variable in specific, quantifiable terms. Potential predictive variables were identified to reflect the ideal traits of a successful recruiter. The method of Mallow's Coefficient Cp in conjunction with hypothesis tests, was used to develop the final predictive model. Residual analyses, data-splitting, and cross-validation methods assured the appropriateness and adequacy of the final model to describe and predict recruiter success However, this model is limited by the fact that all sales ability data was collected using the present-employee method For the purpose of calculating potential cost savings, an analysis using the Taylor and Russell tables was conducted. Cost savings expected from use of the model amounted to nearly $3.38 million annually.