The American Association of Insurance Services (AAIS) serves the US property & casualty (P&C) insurance industry as the only national not-for-profit advisory organisation governed by member insurance carriers. More than 700 carriers across the US use it for policy forums, rating information and data management capabilities.

The AAIS's data engineering and actuarial services team has helped modernise insurance products across the P&C space, including homeowners, auto, personal, and commercial lines.

Its 17-person team worked using alternative data sources and modern technology to change how the AAIS creates its products, which in turn has led to important innovations.

At the core of the innovation is the development of a data lake environment, built on Hadoop open-source software and scaled utilising Amazon Web Services infrastructure.

Historically, insurance data has been copied, reformatted, imported, and analysed in an actuarial analysis software package. Each step added time and the potential for error. To counter this, the AAIS team built an automation tool using technology deployed within the data lake to perform this work.

The team had developed wide varieties of non-traditional data sources, such as government data on fire incidents, a vast array of census data, and targeted data sets assembled by third-party data aggregators to build sophisticated predictive modelling frameworks.

It also built new data sets where no data exists. For example, for a public fire protection methodology development project it calculated drive times from some eight million historic fire incidents for over 50 states to the nearest three fire stations and the traffic density impact at the time of the day the incident occurred.

"Actuarial [insight] is critical right now because the risks are both emerging and evolving and we need the ability to quickly adapt while maintain accuracy with regard to the quantification of the risk," says Phil LeGrone, vice president of data and actuarial at AAIS.