Data-driven insurance Article 3: Targeted marketing and sales applications

Data science is well on its way to introducing a sustainable change in the insurance world. As we already discussed in our previous “Data-driven Insurance” series article, processes and customer journeys are already influenced by data science today—and will be even more so in the future. To better understand the huge potential that data science additionally offers for marketing and sales activities, it is important to take a closer look at the B2C relationship.

Imagine the following situation: What if a marketing or sales team were to contact not only those customers whose insurance policies expire within the next few weeks, but all customers who, from a customer perspective (and not from a company perspective!), actually need an updated offer? What if these customers were not simply to receive a standardised offer regarding their currently valid insurance cover, but a product offer that they actually need due to their current life situation—which may have changed several times since they had taken out the original insurance? And what if the sales team knew exactly which customer was not prepared to accept an offer at a higher price?

This exemplary situation highlights problems that can be resolved by means of data science. We can derive two specific use cases, which can be used to increase customer satisfaction and at the same time significantly simplify the sales team’s activities through additional increases in efficiency and higher probabilities of success:

  • Data-supported cross- and up-selling: An automated analysis of the customer history can be used to create a next best
  • Customer churn analysis: Defining the factors that determine whether a customer extends, changes or terminates the con- tract in order to predict the probability of a customer churn.

Data-supported cross- and up-selling
The actual use case implementation is carried out in three steps: First, it is important to understand the customer’s needs such that the most suitable products can be derived and further to address the customer via the ideal sales channel.

Figure 1: Three steps to implementing the 'data-supported cross- and up-selling' use case
  1. Understanding customer needs

A first essential step in creating a next best offer for a customer is to understand the customer’s needs. These specific needs vary from customer to customer, depending on their background, personal interests or financial means. Furthermore, recognising a change in the customer’s life situation also provides the opportunity to offer an updated or new insurance product.

  1. Finding products suitable for the customer

The next step is to find the ideal and most suitable product to meet the customer’s needs and to provide the best possible insurance coverage. The data analyses, which are used to offer the product that best suits this type of customer, will increase both customer satisfaction and conversion rate.

  1. Selecting the ideal sales channel

Customer satisfaction and conversion rates will increase even further when using the most suitable sales channel for this specific customer.

A change in the customer’s life situation, such as buying a home, usually requires some insurance products to be updated. In this case, an automatically generated offer is very advantageous for the customer. Furthermore, the offer includes not only the ideal insurance coverage, but also other essential products that are rel- evant to this specific life situation.

A variety of different methods are available for carrying out these customer classifications and product allocations. Implementing a random forest, for example, is a good option for understanding customer requirements. The apriori algorithm can be used to identify the most suitable product.

Customer churn analysis
This use case can be implemented in two steps: first, the factors that influence customer churn must be identified to then predict the termination.

Figure 2: Two steps to implement the 'customer churn analysis' use case
  1. Survey of customer churn factors

The first step is to determine the factors that are decisive for a customer to extend, change or terminate the current insurance contract. Typically, insurance records contain hundreds of different variables, most of which do not have a significant impact on the customers’ future decisions about their current contracts. To be able to make the right decisions about marketing activities and new contract offers, however, knowing these influencing factors is essential: Ideally, the marketing strategy is selected according to the customers’ preferences, e.g. whether they are highly price-sensitive or place particular emphasis on excellent customer service.

  1. Prediction of future contract continuation

Knowing the essential influencing factors allows dealing with the next step and projecting the future contract continuation. This prediction has at least two advantages: first, knowing which customers will statistically terminate the contract gives marketing and sales staff a valuable edge in contacting customers according to the right priority. Second, detailed sales planning can be carried out at a very early stage.

Various methods are also available for implementing this use case. We have decided to apply the gradient boosting method to the implementation, as this provides a better quality measure compared to other methods.

Instead of just simplifying the sales process and making it more intuitive, faster and customised, data science has a considerable potential to significantly increase sales through more efficient processes. In addition, expected customer satisfaction can be achieved through ongoing customer-centricity, which forms a fundamental part of a successful use case implementation.

Figure 3: Improvements achieved in both use case projects presented above

Marketing and sales, however, are not the only insurance topics where data science can achieve significant improvements; these are probably just the best perceived application areas. Internal risk models are another topic that is predestined for data science application, so as to improve risk protection and to achieve a monetary advantage. Our next article will reveal ways to achieve this.

Alexander Riesner

Manager Office Vienna

Tobias Holler

Analyst Office Munich


Leave a Reply

Your email address will not be published. Required fields are marked *