Identifying Hidden Biases in a List Rental Modeling Sample

 As a modeling firm supporting the list rental requirements for several of our clients, we, at Jigyasa Analytics, have literally built several hundred models to help with these efforts. For those unfamiliar with the term ‘list rental’, it is when a firm borrows a set of names from another firm, for marketing purposes. A modeled list is likely to be more responsive and so very desirable in these situations. A valuable lesson we have learnt while supporting the modeling for list rentals, is that while the modeling algorithms, feature selections and stabilization mechanisms are all very important, the best techniques will amount to nothing, unless we address the hidden biases in the data. 

 Much has been written about the recent troubles at NETFLIX and the common prevailing view is that at least some of the loss in its subscriber base could be attributed to the recent increase in their price. In all likelihood, the company, believing its demand to be inelastic, assumed that the price increase would not have an adverse impact on membership. At Jigyasa Analytics, being closely associated with the publishing industry, we've seen several newspapers and magazines, increase prices with great success while others have not fared as well. Having conducted pricing analytics and elasticity estimation studies on numerous occasions, we thought this might be a good occasion to share some of our thoughts on this topic.