Case Study Deskcasestudydesk.com
InsuranceSourced

Case Study: IFFCO-Tokio saves over $1M annually on fraud with H2O.ai

IFFCO-Tokio General Insurance Case StudySourced & dated by Case Study Desk
Key facts · TL;DR
Company
IFFCO-Tokio General Insurance
Industry
Insurance
Challenge
A three-person team needed on-premise fraud prediction for claims
Headline result
Motor-claim fraud models save about US$1M a year and lift retention 3%

Key results

$1M
Expected annual fraud savings
~70M Indian Rupees per year
3%
Increase in customer retention
since July 2021 trial

The challenge

IFFCO-Tokio, a joint venture between Indian Farmers Fertilizer Co-operative and Japan's Tokio Marine Group, wanted to detect and predict fraudulent insurance claims. It needed to do so with a small in-house team of three developers with limited data-science experience, while keeping data on-premise for privacy reasons.

The solution

Guided remotely by an H2O.ai Kaggle Grandmaster, IFFCO-Tokio used H2O AI Cloud's automated machine learning with pre-built algorithms to build and deploy an on-premise fraud-prediction model for motor insurance claims in July 2021, later extending it to health insurance claims.

H2O AI Cloud is a platform that my team can learn by themselves. I find the platform to be very friendly for people who are not data scientists or have little knowledge of data science.

SG
Seema Gaur
Chief Information Officer and Head of IT, IFFCO-Tokio

The results, in context

Using models built on H2O AI Cloud, IFFCO-Tokio sieved out fraudulent claims and expects to save almost 70 million Indian Rupees (about US$1 million) annually. Since the online-customer trial began in July 2021, customer retention increased by three percent.

Products used

H2O.ai H2O AI Cloud