Cortana Analytics helps insurer If P&C tackle Big Data analytics workload
This post is co-authored by Katherine Lin, Data Scientist, Trond Brande, Principal Solutions Specialist, and Tao Wu, Principal Data Scientist Manager, at Microsoft
Data has been an important part of the insurance industry for centuries, but the emergence of big data has brought unprecedented challenges to insurers around the world. Today, insurance companies large and small need to ingest, process, analyze and act on massive amounts of data from heterogeneous sources quickly and cost effectively. Even as legacy analytics tools fail to provide the necessary capability and agility for new big data workloads, insurers are discovering Cortana Analytics Suite (CAS), a fully managed big data and advanced analytics suite, as an enabler for their new analytics needs.
If P&C Insurance, a leading property and casualty insurance company serving three million customers in the Nordic region, worked with Microsoft and completed a pilot project on CAS, with a focus on Azure Machine Learning. The main goal of the collaboration was to evaluate how well CAS handles different aspects of predictive modeling as a replacement for If P&C’s on-premises legacy data analytics platform. The success of the pilot led If P&C to replace their legacy SAS platform with the Cortana Analytics -based solution. “With the same or better results as the existing solution, and cost savings of 90%, we have a serious contender in Cortana Analytics Suite. This will have significant business impact for us,” explains Ole Christian Kjelsaas, Head of Data Warehouse at If P&C.
Major advantages that If P&C is seeing include:
- Performance. In all use cases evaluated in the pilot, the performance of CAS-based solution met or exceeded If P&C expectations.
- Cost effectiveness. If P&C estimates that it will realize significant cost savings by using CAS.
- Integration. CAS components are designed to work with one other. Using Azure ML web services, If P&C can quickly integrate the output of predictive analytics into an end-to-end data pipeline.
- User productivity. If P&C finds that the ramp time for data scientists and engineers is much shorter with CAS.
Three use cases were evaluated in the pilot:
- Churn prediction
- Upsell prediction
- Email text analytics
The purpose of the Churn Model was to predict whether or not a customer will cancel the policy in a 40-day window surrounding their renewal date. The Upsell Model was used to predict the probability of success of a potential upsell communication to a given customer. The data used for these two models includes the following:
- Age
- Duration of the policy
- Product composition
- Payment solution
- Household data
- Contact points (phone/web)
Using Azure ML Studio, a series of data preprocessing steps were first performed on the raw input data, followed by various steps of feature engineering. For both the churn and upsell predictions, binary classification models were then developed. Azure ML has the capability to quickly deploy and compare the performance of multiple algorithms in parallel. The best performing algorithm was deployed as a web service which could be consumed programmatically as part of a fully automated end-to-end pipeline.
Possible applications of the churn and upsell models include:
- Churn prevention communications (telemarketing/event-driven marketing/inbound marketing).
- Upsell communications (telemarketing/event-driven marketing).
- Upsell lead generation.
- Next-best-action recommendations.
The Email Text Analytics project helps If P&C classify inbound email. This part of the work is critical for If P&C to transition from a push to pull marketing strategy. The following list highlights the potential business value:
- Email dispatching. Inbound email traffic will be channeled to specialized teams in customer service.
- Churn prediction. Messages with negative sentiment may indicate increased risk of customer churn.
- Customer feedback. Automated notification of positive and negative customer feedback will be sent to business units.
Besides Azure ML, other CAS components such as Azure Data Factory and Azure Data Lake are also being adopted at If P&C. In addition, integration with If P&C ‘s instance of Dynamics CRM has also begun.
By utilizing Cortana Analytics Suite’s advanced analytics capability, If P&C is now able to uncover the value of data quickly and cost-effectively.
