Monday, 7 February 2011

Setting up a cross-selling model

Cross selling campaigns aim at selling additional products to existing customers.
A cross selling model estimates the propensity to uptake an add-on product for each scored customer.

A cross selling model can be built on the results of a test campaign to analyze respondents and identify customers with increased purchase potentials.

An easier approach which does not require the running of a test campaign, is to analyze the profile of customers who acquired the product of interest in the recent past. Here are a couple of examples based on the latter approach:

Banking: Cross-selling an investment product to savings’ customers. 
    • Modeling population: all active customers not owning the investment product at H (end of the historical period). 
    • Target population: those that acquired the product at T (end of the event outcome period). 
    • Scoring population: all active customers not owning the product at present.





Mobile telephony: Cross-selling a telephony service, e.g. Internet usage. 
    • Modeling population: all active customers not using the service before H (no traffic before H). 
    • Target population: those that used the product at T (at least some traffic between H and T). 
    • Scoring population: all active customers not using the service now. 

Get 'Data Mining Techniques in CRM' for more examples and tips

2 comments:

Anonymous said...

Do you use logistic regression for creating cross-selling model? Your book only offer examples for segmentation using PCA & cluster analysis?

Anonymous said...

The second approach actually does not quite resemble the first approach because a test campaign itself would become a factor for affecting the response.