Better underwriting through better analytics
When Guardian Life Insurance Company of America embarked on a company-wide initiative to become more data-driven, they had no shortage of data. Founded in 1860, there were filing cabinets brimming with historical documents. The real challenge was in getting usable data, without which it would be impossible to perform advanced analytics or even basic exploratory data analysis. “We couldn’t broadly visualize our data or discover actionable insights with our data being constrained by a paper source format before Vidado.” says Andrew Holshouser, Director, Underwriting R&D Analytics for Guardian
Challenges in Accessing Data
Guardian set out with the specific goal of creating top-notch predictive analytics for their underwriting department. The first step was digitizing hundreds of thousands of paramedical forms going back 5 years. When taken in aggregate, the data can offer highly accurate insights into applicant mortality, disclosure, or behavioral aspects of assessing life insurance risk. Guardian had built out a data analytics department that was poised to begin doing the robust modeling necessary for this initiative, but they found themselves frequently limited by data availability. These documents contained handwriting, checkboxes, and were sometimes marred by low resolutions or artifacts from being faxed or scanned - making digitization through traditional solutions costly and ineffective.
Guardian investigated using human data entry teams and OCR-based solutions but neither could offer the accuracy and ROI the project needed. That’s where Vidado’s AI came in. “Vidado has a pretty long track record of success.” says Frank Chechel, 2VP, Underwriting R&D at Guardian. “We knew they had a track record of doing this and doing it well.”
After selecting Vidado, the Guardian team and the Vidado team began meeting weekly to lay out a project plan and execute it. Over the engagement Vidado digitized the forms and Guardian was thrilled with the accuracy of the extracted data, especially around the lower-quality and handwritten documents. Andrew Holshouser appreciated the fact that when the Guardian team encountered a challenge, Vidado never told them it couldn’t be done, but always found a way to fix or work around the issue without impacting the project timelines. “There’s a lot of vendors out there that are making a lot of promises. You guys said you were going to do something, you had a project plan for executing on it, and you delivered on it.” said Frank Chechel.
The Situation Today
With the help of Vidado’s AI, Guardian was able to automatically digitize over 500,000 handwritten pages corresponding to 75 different form types at an accuracy level of 96.17%. Guardian has already started utilizing this data as part of their analytics efforts. But that’s only the beginning. Going forward, Guardian plans to combine this data with other datasets to perform more advanced analytics. With the best data at their fingertips, Guardian is ready to take its underwriting process to new heights and redefine what it means to be data-driven.