Improve Customer Experience and Reduce Risk for Insurers

Improving customer experience and lowering risk is paramount to Azur

The Situation

Azur is a forward-thinking company developing a technology-based platform which enables insurers to work more effectively with brokers and end-user clients. Their mission is to help insurers digitally transform themselves to disrupt the market and ultimately change the way that insurance is sold; making it smarter and more responsive to customer needs. They have taken a new and innovative approach to improving the performance of data by applying Artificial Intelligence and deep learning to create differentiation, reduce risk and improve the customer experience.

The Challenge

Azur developed a new mid-market home insurance product. A key focus of this product was to reduce the time taken to complete the online application process whilst providing accurate information and delivering a great customer experience.

To achieve this, Azur needed to pre-populate as many fields as possible with the most accurate data, automate the process to cut down the number of fields that needed to be manually entered and reduce the completion time.

One of the key issues was the accuracy of the standard Home Rebuild Cost they were currently using. The original starting point for sourcing a Rebuild Cost was based on standard BCIS rates. These proved to be only suitable for ‘standard’ mass market properties and the broker value for properties on file was often time inaccurate – it just followed market value or historic guessed value. It was not suitable for Azur’s mid-market client base.

They needed to find a new approach to calculating a more accurate Home Rebuild cost that reflected our client profile and base it on a more accurate property data set. Clients are likely to have higher specification properties, and this can vary within certain postcode districts, especially in some areas of London where there can be a huge variance in the types of insurable properties. In addition, they have a higher likelihood to live in listed buildings and to live in more affluent postcodes.

The Solution

Outra assisted Azur to create a completely customised Home Rebuild Cost model. Using our more accurate proprietary property data set to base the model on, we aggregated this alongside existing rebuild rates per square metre to create the custom model. By applying deep learning and testing different modelling approaches, we identified non-linear patterns in the data to deliver the most accurate and predictive rebuild cost results. Postcode districts were used as standard, but more granularity was given to more affluent postcodes in Greater London and further granularity to Kensington and Chelsea. The model also factored in listed buildings, with an uplift ranging from 150-200% of the assumed standard rebuild cost depending on the listed grade of the property.

This resulted in being able to accurately calculate the rebuild sum insured for Azur’s client base, producing a unique list of rates per square metre values at postcode district and more detailed values in certain areas which required further granularity. 

Using the Outra API, the property data and Rebuild Cost were automatically pre-populated, using only the address details supplied – increasing accuracy and reducing form completion time.

The Result

The custom Home Rebuild Cost model showed an improvement in accuracy when comparing against the rebuilds held on the existing book of properties under a £1m rebuild, achieving a 48% match (within a 20% tolerance) to the Private Client Group rebuild sums insured. This highlighted the challenges that customers face when asked: “What does your home rebuild cost?”, this is often guessed or incorrectly based on market value and demonstrated the need for a more accurate automated model.

Using predictive modelling through the application of more accurate data and deep learning greatly improved the user experience, enabling us to reduce their risk question set to just five questions. This cut the time taken to complete the online form, allowing the broker to quote and bind a risk within two minutes. This is without any detriment to underwriting performance as we use 67 rating variables (including additional proprietary property data variables and confidence scores supplied by Outra) being captured for a robust actuarial model. This product is fully digital and the first of many being launched on the Azur platform allowing the business to significantly scale whilst containing costs.

We refresh and update the Home Rebuild Cost model on a monthly basis, so that new full postcode rates per sqm are included with the ability to override the sqm value in the model if Azur have appraised a property themselves. We continue to work together to source new data points and these can easily be included in to the API once their value has been assessed.