×
A major insurance brokerage predicts its loan applications
Discover how Micropole's teams enabled our client to model the forecasted volume of activity using machine learning.
Context
Our client is a broker in loan insurance and individual providence for 30 years now, the company develops its business model on the European territory. It is the intermediary between the insurer and the distributor (banks, credit brokers, etc.) for the management of insurance, loan and personal protection contracts. The role of our client is to accompany the insured throughout the life of his contract.
Some numbers:
- 197 million € in sales
- 730 employees
- 15 million contracts under management
- 1 out of every 4 home loan insurance policies handled by our client in France
Challenges
When Micropole's teams were called in, the volume of home loan applications fluctuated greatly over time, thus impacting the availability of resources dedicated to managing applications. As the training of a loan insurance teleconsultant required several months, the management and planning of resources represented a real challenge for the company.
Methods
Micropole modeled the forecasted volume of activity several months in advance in order to adjust recruitment needs and/or contract durations as best as possible, to anticipate drops and peaks in activity, and to guarantee a good quality of training for teleconsultants.
The Design Sprint method was used here to control the quality of the final product
- Sprint #1: an optimization of the ML algorithms (carried out after a report of the inability of the available data to meet the objectives of the project).
- Sprint #2 : un modèle produisant des prévisions d’excellente qualité à 3 mois (MAPE < 10%) suite à l’intégration de nouvelles données OpenData axées sur les recherches de mots-clés spécifiques sur Google
Benefits
Better management of human resources and anticipation of recruitment and training needs for new teleconsultants
Better staffing of the call center team and improved customer experience due to optimized waiting time