Optimal customer selection for cross-selling of financial services products
In this research a new methodology for optimal customer selection in cross-selling of financial services products, such as mortgage loans and non life insurance contracts, is presented.
This paper addresses the challenge of optimally selecting a subset of customers for cross-selling financial services products to, where the profit of a given cross-sale is unknown and customer specific. Financial services companies tend to possess significant databases and a long relationship with each customer. In this situation the challenge becomes to use the database in general and specific knowledge of the individual target to estimate the probability of a cross-sale, the cost of a cross-sale attempt, the average discounted future profit and the uncertainty of the profit of the entire cross-sale attempt for that individual. Once reliable estimates for the stochastics of the cross-sale process have been established, one can optimise the cross-sale profit according to a variety of criteria including return and risk. In this paper, we first consider the simple question of optimising the average profit, but we also consider one version of adjusting for risk when optimising cross-sale profits. Our extensive case study is taken from non-life insurance, where our sales probability model is provided to us by the company that also provided us with the data.
In this paper, we have introduced a new flexible approach to optimal cross-selling. We solve the optimisation problem of maximising both an optimal mean criteria and a mean-variance criterion. Our profit/risk performance optimisation approach has, to the best of our knowledge, not been previously considered in the context of cross-sales marketing.
For the purpose of solving the proposed optimisation problems, we have developed a stochastic model of the profit, emerging from a successful cross-sale to an individual prospect and a group of prospects. The model is expressed in terms of certain random variables, characterising the occurrence of sale, the price and the cost. When trying our methodology out on real data (we consider a large insurance data set) we get practical and convincing answers suggesting potential cross-sale strategies.
The research paper is available for download below.
