Business Objective
Our client is a Fortune 100 P&C insurer, leveraging telematics data to assess driver risk profiles and offer discounts on insurance premiums. Telematics data can be sourced from multiple vendors that collect data from a variety of devices. Our client engaged with us to
- Evaluate the effectiveness of 4 telematics data vendors over the quality of high-frequency sensor data
- Identify ways to improve model performance (model predicts the premium discounts for varying levels of driver risk)
Challenges
- Understanding of the complex telematics data landscape
- Keeping pace with the rapidly evolving Telematics space
Solution MethodologyÂ
- Assessment of various data capture devices, types of telematics data, and the various data providers
- Leveraged raw telematics data to detect events, and external map data to determine driving scenarios
- Event detection, driving scenarios, and smartphone usage data were leveraged to determine contextualized events
- Vendors were evaluated based on data comprehensiveness, data quality, data type and breadth, costs, provision channel, etc.
- The developed model predicts the premium discounts for varying levels of driver risk
Business Impact
- Identified vendors that meet long-term business requirements
- Identified data feed issues resolved by some of the vendors leading to improvement in data feed quality
- Harmonization of data received from across devices and vendors, leading to improvement in overall data quality
- Lift in downstream models’ accuracy due to improvements in data quality
