Quantitative Study
Rapid Repeat Purchase Scoring Framework (RRPS)
Abstract
As organizations face a tough economic environment, with declines in sales, the business model needs to adapt itself to changing economic realities. In such a scenario the conventional wisdom of mining existing installed base takes precedence over new customer acquisition. We developed a Customer Repurchase Probability model to help predict & target customers with a high likelihood of repurchase in the next n units of time. Our Rapid Repeat Purchase Scoring framework provide a scalable and efficient algorithm to identify likelihood of purchase for each customer in a given time period. The framework aims in improving redemption rate and hence increasing incremental revenue for campaigns, by providing valuable insight to the marketing team in selection the best customers to go after. The framework when applied to the current Back To School marketing campaign by Consumer Exchange (HHO) resulted in $2.3MM in revenues by selecting top 3 groups of customers coming out of the model framework.
Problem statement
Targeting the right customer at the right time is one key aspect of marketing. In this context, knowing or at least, anticipating whether a customer will make the next purchase within next n time period would immensely increase the effectiveness of any marketing campaign. Rapid Repeat Purchase Scoring framework provide a scalable and efficient algorithm to identify likelihood of purchase for each customer in a given time period.
Our solution
We intended to device a new algorithm which can eradicate the drawbacks of the existing model, by utilizing the same information. We divided the problem into two parts: 1. Developed a model that predicts rate of purchase(A) for each customer 2. Calculated the likelihood of churn(B) after last transaction Both the model was based on a regression frame work where generalized additive model (GAM) was used to identify non linear patterns in the data and polynomial
Abstract
As organizations face a tough economic environment, with declines in sales, the business model needs to adapt itself to changing economic realities. In such a scenario the conventional wisdom of mining existing installed base takes precedence over new customer acquisition. We developed a Customer Repurchase Probability model to help predict & target customers with a high likelihood of repurchase in the next n units of time. Our Rapid Repeat Purchase Scoring framework provide a scalable and efficient algorithm to identify likelihood of purchase for each customer in a given time period. The framework aims in improving redemption rate and hence increasing incremental revenue for campaigns, by providing valuable insight to the marketing team in selection the best customers to go after. The framework when applied to the current Back To School marketing campaign by Consumer Exchange (HHO) resulted in $2.3MM in revenues by selecting top 3 groups of customers coming out of the model framework.
Problem statement
Targeting the right customer at the right time is one key aspect of marketing. In this context, knowing or at least, anticipating whether a customer will make the next purchase within next n time period would immensely increase the effectiveness of any marketing campaign. Rapid Repeat Purchase Scoring framework provide a scalable and efficient algorithm to identify likelihood of purchase for each customer in a given time period.
Our solution
We intended to device a new algorithm which can eradicate the drawbacks of the existing model, by utilizing the same information. We divided the problem into two parts: 1. Developed a model that predicts rate of purchase(A) for each customer 2. Calculated the likelihood of churn(B) after last transaction Both the model was based on a regression frame work where generalized additive model (GAM) was used to identify non linear patterns in the data and polynomial
References: [1] Fitting Generalized Additive Models with GAM procedure – Don Xiang