[Customer knowledge]

Improving the performance of target marketing in insurance

The leveraging of data and the addition of artificial intelligence models offers ways to better define risk profiles and high value profiles. This paves the way for more and more individualized insurance, such as putting forward an offer to owners of a second car with little or no information about them.

To identify and target customers with a second car in order to provide them a personalized offer.

For who?

A French online insurer

Difficulties encountered

▪ No information on the 2nd car
▪ Low client input (Last name / First name / Address / Email / Age / Profession code)
▪ Low representative sample of the French population (5000 persons)

How?

Our approach is essentially based on the use of external data and of Open Data:

▪ Yellow pages
▪ White pages
▪ Facebook
▪ Linkedin
▪ Previous acquaintances/partners
▪ Best agents
▪ Best Street Maps
▪ INSEE

The insurer’s data was enriched and allowed the modeling of an algorithm.
It considers, among other things, population density, housing characteristics, income level, Socio-Professional Category, proximity to public transport…
A score and a classification are then attributed to the initial sample. This service makes it possible to move from an unsupervised machine learning model to a supervised one.

Samples are expanded and customer data is further enriched.

The reliability of the algorithm is ultimately improved.

And the results?

The algorithm has been adapted to the CRM of the insurance company. It now makes it possible to establish the probability of who owns a second car (old and new customers). This model has made it easier to identify prospects. Sales and marketing teams can customize offers to increase productivity and improve ROI.

octopeek icon

Retrouvez l'essentiel de l'actualité "Big Data" en vous inscrivant à la newletter Octopeek

 

 

Félicitation, vous allez recevoir le livre par email