Dataiku +
La Mutuelle Générale

Data Science for More Effective Customer Acquisition in Insurance

One of the reasons data is so integrally important for insurance is because competition in the industry is fierce — organizations are all looking to capture the same type of customer, and the amount of money needed to acquire a new customer has exploded in recent years. 

We sat down with Yann-Erle Le Roux, Data Strategy Manager, and Magalie Didier, Data Scientist and Head of Customer Knowledge, at La Mutuelle Générale to understand how they have made customer acquisition smarter by augmenting the work of humans with data science. 

• More than 70 years of experience in the market.
• Serving over 1.4 million customers and 8,000 enterprise clients.
• More than €1.1 billion in turnover annually

The Project

La Mutelle Générale developed a machine learning-based system that helps sales prioritize their work by assigning an individual probability of conversion to each prospect, whether that prospect is an individual or a group. In order to do this, they first looked at data on existing clients; specifically, their cost of acquisition and lifetime value. They use this analysis to establish “look alikes” for each prospect — that is, an existing customer who has similar characteristics and therefore will likely mirror the future actions of the prospect.

The end result of this system is a tool available for sales that allows them to more effectively prioritize their prospects by providing two pieces of information to consider:

1. Likelihood of conversion

2. Likelihood of recuperation of acquisition costs

The team also created an interactive map containing this data so that any travel to visit prospects could be maximized by visiting other potential prospects nearby.

The People

There were multiple teams and people involved in going from raw data to end product for La Mutuelle Générale’s customer acquisition prediction system. The company has two data-focused teams:

In addition, the customer acquisition project also involved other teams throughout the business to ensure that the final product was aligned with business goals and needs:

• Marketing and sales
• Internal communications
• Customer accounts division

The Processes and Tools

For every data product, including the customer acquisition system, the data teams at La Mutuelle Générale follow the same methodology to ensure a thorough and consistent process.

La Mutuelle Générale uses Dataiku from data collection through operationalization as an end-to-end tool that saves time across all parts of the data-to-insights process, allowing the many different teams and profiles of users involved to contribute directly to the final data product:

1. Data product idea and scoping: This includes making sure that the data team — including technical team members — have a solid understanding of the business problem at hand so that they can find solutions that actually solve the problem.
2. Data collection: From all souces, including both internal (like CSM data, etc.) and external data sources.
3. Data preparation: In addition to cleaning existing data, this is also the stage where the team will identify any missing or problematic data.
4. Model building: La Mutuelle Générale uses state-of-the-art modeling techniques and then leverages Dataiku specifically to quickly compare the performance of different models.
5. Visualization: To go from model to a system that can be used by the business, the team leverages visualization to communicate results in usable formats.
6. Model optimization: After a model is launched (as well as during earlier testing phases), the team observes its behavior to ensure it’s working as intended and tweaks as necessary.
7. Operationalization: The data teams at La Mutuelle Générale always aim to deploy machine learning services live into the places where they are needed. For example, in the case of the customer acquisition predictions, that includes into the account executives’ CRM as well as automatically pushing the information proactively.