Internal Cross-Sell - II in Anatomy of Decisions

In the last segment, you learned why cross-selling is extremely endeavour for any organisation, especially one selling a push product like insurance. You learned how cross-sell works in the insurance business. You also know that before embarking on any analytics project, we start with a set of assumptions, hypotheses.

In the following video, Ashish talks about the different hypothesis his team had before starting the project. He will also talk about how machine learning models can solve this non-technical problem.

In the above video, you learnt that the team tested the following hypotheses

• Customers with a long association with the organisation have a higher propensity to buy an additional product.

• Customers buying low-risk plans are more likely to invest more money in the insurance sector compared to customers investing in high-risk plans.

• Customers with high income to premium ratio will have more propensity to buy additional insurance products.

• Customers who already have multiple policies may not buy more policies.

• Higher age customers have higher propensity to buy an additional policy.

All of the above hypotheses were successfully validated using data. Furthermore, you learnt how data has been transformed to make it fit for the purpose of modelling. You learnt that a flag variable has been created to identify the cases of cross-selling and the model is trained on this data so that it can successfully predict the customers with a high likelihood of buying another insurance policy.

You saw that once the model was deployed, the response to cross-sell campaigns increased by 15%. In the following segment, you will learn how analytics is used to predict if a policy will result in an early claim.