Drop out prediction in sport centres. Definition of models and reproducibility

Abstract

Purpose: One of the main problems detected in the sport centre field is abandonment of physical activity by members. The aim of this estudy is to evaluate the reproducibility of predictive models of abandonment in sports centres, based on information collected through access control and the Customer Relationship Management system. Method: Monthly data corresponding to the actual behaviour of of two sport centres in Spain were recorded over the course of one year (12,545 members). Based on these variables, two predictive models were designed using the absence or permanence in the centre as an independent variable. Subsequently, the regression formula obtained was applied on all members of the same sports centres a year later (12,986 members), checking the effectiveness and reproducibility of the model. Results:. Logistic regression models based on customer historical behaviour were useful to predict abandonment. The effectiveness of the models was around 70% a year after. Conclusions: The segmentation of users based on the likelihood of drop out and the identification of behaviour variables that predict such drop out and the confirmation of the effectiveness of the predictive model, can help to optimize the design of loyalty strategies to prevent the abandonment in sport centres.