Leveraging image analytics for P&C Insurance – personal & commercial lines
Intelligent Remote Property Assessment
Intelligent Remote Property Assessment
Taking a manual approach to property assessments can lead to rising personnel costs and operating expenses, as well as increasing exposure to inaccuracy assessments and higher loss ratios over time.
Introduction - Property insurers; Are you accurate, agile and cost effective?
The property assessment is an all-important process within the property and casualty (P&C) industry. This task has been largely unaffected by technological disruptions – until now. Using image analytics can improve property assessments end-to-end. Insurers can enhance their capabilities for accurately predicting risks at the location level, and effectively reduce the manual interventions required when adding and maintaining property assessments using this technology.
Traditionally, many insurers have been unable to leverage location-level insights simply because due to a lack of traditional data sources, such as an absence of roof condition data, and processes not designed to take advantage of this intelligence. This lack of relevant data also affects the ability to enforce proactive risk control measures. Additionally, taking a manual approach to property assessments can lead to rising personnel costs and operating expenses, as well as increasing exposure to inaccuracy assessments and higher loss ratios over time.
The underwriter can leverage these scores to access risk at a location level and subsequently use these risk parameters for AI models such as case reserving and loss value prediction.
Many insurers are looking for ways to streamline the property assessment process by using image analytics. For instance, pictures of a house roof provided by satellite imagery can be used to better understand the condition of a house and determine underwriting and loss control risks. Similarly, for extreme risk events such as tornadoes, hurricanes, and wildfires, image analytics can help with real-time and forecasted risk monitoring, management, and adjuster prioritization after the event1. These efforts can provide insurers with continuous risk insights for their book of business, enabling them to proactively leverage risk control measures like premium increases or enforcing structural changes in the property.
Some insurance companies have already begun using image analytics to handle claims for customers. During Hurricane Florence and Hurricane Michael in 2018, image analytics allowed carriers to assess the disaster’s impact immediately by gathering data from the events via aerial images2. Insurers such as AllState have also started collaborating with authorities such as the Geospatial Intelligence Center (GIC) for geospatial data that enables faster inspections, resulting in consumers living in hurricane- and wildfire-prone areas to be able to easily buy an insurance policy within minutes and resolve claims faster after a disaster3.
The right image analytics approach offers increased visibility, lower loss ratios, and reduced operational expenses by extracting property attributes from multiple geo-spatial data sources and combining them with streamlined analytics workflows. The diagram below provides a high-level view of an impactful image analytics practice:
The underwriter’s user journey begins by requesting the risk scores for the relevant property address from the analytics vendor. This enables the analytics vendor to call an API, requesting the image vendor to provide the aerial image associated with this location. The analytics vendor can then go ahead with not only complex computer vision models to create object detection models to identify additional structures like trampolines, swimming pools, and solar panels, but also AI algorithms to extract property attributes like roof condition and roof footprint area. These can then be leveraged to evaluate risk scores for different risks, like fire and flood. The underwriter can leverage these scores to access risk at a location level and subsequently use these risk parameters for AI models such as case reserving and loss value prediction.
Leading insurance companies have begun looking for the right type of image analytics for their property insurance decisions. Some best practices carriers should keep in mind when beginning their image analytics journey include:
- Domain-specific attributes for personal and commercial lines: The advancements in neural network models have enabled the extraction of 100+ attributes from a single aerial image, but not all the attributes are equally important. To add to this, poor domain knowledge might cause some attributes to be overlooked for that are important for commercial but not personal lines, and vice versa.
- Attribute accuracy: Replacing human bias inherent in the traditional assessment process is one of the most important aspects pushing insurers towards adopting image analytics.
- Property coverage: In today’s world of enhanced aerial imagery, insurers should aim for their solution to provide 90%+ coverage of the properties in their book at a minimum.
- Dependable refresh rate: A solution should include a twice-a-year refresh rate of the images in their database to make accurate renewal and claim processing decisions. This enables carriers to make periodic comparisons between image attributes, redefining traditional processes like evidence gathering and premium repricing.
Leveraging property attributes and risk insights for targeted marketing
Targeting and converting the right prospects using forecasted higher loss ratios and low customer lifetime value is one of the leading problems facing P&C insurers looking to grow into new or expand existing markets.
The high-quality property attributes and risk insights generated by image analytics can be coupled with third-party credit and marketing data to create an enhanced targeting model for targeting prospective members. This can lead to a sustained improvement in marketing ROI for insurers as well as:
- Lower cost of acquisition
- Increased conversion rates
- Reduced loss ratios
Granular location intelligence transforms underwriting
Traditionally, P&C carriers have lacked the granular date required to incorporate location intelligence into their operations. However, using image analytics for property assessment provides this data, allowing companies to eliminate manual interventions by leveraging deep learning models to extract property attributes. This enables automated inspections and surveys of properties and their structures by using information from a variety of sources such as:
- Third-party data APIs
- Marketing data
- Weather data
- Property survey data
- Crime data
- Miscellaneous service providers such as 3D interior data
- Client systems
- Geospatial information providers
Providing a complete 360º view for underwriters by integrating third-party data APIs and advanced machine learning techniques is key to plugging the gaps and enabling granular risk and loss insights for insurers. This provides accurate and cost-efficient severity roof detection and identifying building structure damage for underwriters4.
Digital property inspections
Manually conducting property inspections is a tedious task, requiring assessors to climb on roofs, examine building conditions and nearby vegetation. If a manual inspection is required for every single policy purchase and renewal, this can result in increased risks for agent injuries, inaccuracies, and errors. Additionally, inefficient audit scheduling or routing and limited human resource can further complicate this step.
The right approach to property inspections should incorporate image analytics to offer a better alternative to physical surveys and audits. Leveraging deep learning models can extract hard-to-evaluate property attributes like roof conditions and other property parameters. Doing so eliminates time-consuming processes like climbing on roofs and helping carriers effectively utilize their human resources across a higher volume of properties.
Renewals and property surveys
Holistic portfolio assessment
Policy renewals have always been a tricky affair for carriers. Traditionally, insurance companies have resorted to limiting physical verification during renewals to only high risk and high value properties which bypass avoidable claim events. Moreover, any new structures or unfavorable fire or water hazards that have arisen can go completely undetected during policy renewals.
The right policy renewal approach should facilitate a policy administration style that is streamlined and smooth, leveraging automated triggers and alerts via image analysis to establish a continuous risk assessment practice. The relatively lower cost and scalability of image analysis allows insurance companies to leverage it to access 100% of their book value, and assist in the proactive physical verification agent deployment process.
The right approach can leverage image analytics, cutting edge technologies like AI and machine learning, cloud orchestration, and process automation to optimize claim assessments into an optimized process.
Transforming risk assessment and control practices
Insurance carriers have relied on various approaches to assess and rate property risk over the years, going from a simple rule-based approach using actuarial tables to complex statistical models generating risk scores and rating. Now, the assessment process has become more competitive as members demand a more rational approach to pricing.
The right image analytics workflow enables carriers to continuously assess the risk exposure of a property, not just at the time of policy purchase. This helps carriers gain a real-time, multifaceted view of property risk to efficiently manage their risk portfolio. The real differentiator is the ability to incorporate weather feathers and apply it to property data. This can provide insights that can be used proactively by the carriers. Insurers can quickly react to catastrophic events, advising the insured on what need to be done from a risk control perspective and ultimately leading to better customer experiences. The image analytics workflow can also support carriers building ensemble models that incorporate multiple risk evaluation techniques, as well as compliance and subrogation models.
Regularly updated imagery can enable underwriters to better understand when changes on a property occur, be it the addition of a pool or an additional story to a building or encroaching vegetation. Based off this information, underwriters can accordingly give guidance to customers, increase premiums, or cancel a policy5. This results in carriers to better assist policy holders to manage their property risks such as wildfires, floods, or roof damage, and promote safety practices such as preventative maintenance that can minimize the damage in case of a claim event.
Transforming risk assessment and control practices
Physically verifying claims can result in inaccuracies, undue expenses, increased overall costs. The task becomes more complicated due to a lack of efficient resource deployment and customer complications, such as the addition of unaccounted structures and new fire, water, or other hazards. From a member perspective, often the first notice of loss (FNOL) process is a daunting task with lots of paperwork, back-and-forth inquiries, and a lack of transparency that leads to a poor customer experience.
The right approach can leverage image analytics, cutting edge technologies like AI and machine learning, cloud orchestration, and process automation to optimize claim assessments into an optimized process. This can include digitized FNOLs followed by live weather-based automated triaging and auto adjudication. In case the members are not satisfied with the adjudication, they can be provided with digital workflows to settle their property claims.
The right approach can transform the different aspects of the claim function in various ways such as:
- Real-time adjudication for properties damaged by disasters
- Automated triaging of claims
- Automated alerts for fast, precise claim processing
- FNOL claim verification
Optimized loss adjustment and evaluation processes
- Enable digital adjusters to leverage aerial imagery to identify underlying property risks
- Comply with various regulatory requirements during the claim process to avoid any fines
- Assist subrogation models with information gathering such as providing Historical quarter-on-quarter property attribute data
- Provides reconstruction cost analysis based on claims data and offers underwriting models for various risks such as floods, thunderstorms, roof damage, and others
Smart fraud detection
- Damage and natural calamity verification
- Auto-scheduling of special investigation units (SIU)
- Reduction of fraudulent payouts due to periodic comparisons of image-based attributes
- Assistance in verification with historical quarter-on-quarter property attributes
Incorporating image analytics with weather services can be especially beneficial for claims during catastrophic events such as hurricanes and tornadoes. For instance, the satellite images of a tornado path can be leveraged, along with before and after images of the event, to enable insurers to accurately quantify the spread of the disaster and assess damages6.
The insurance industry is adopting digital practices in its value chain, including aerial image analytics. This has opened up avenues for property insurers to gain perform location-level underwriting, better detect fraud, increase efficiency, cut costs, and reduces turnaround times. The pace of this change has been increased by industry need for quick decisions, expense reduction, and optimized customer experience.
Early adopters of image analytics today will reap the benefits, while the rest will be left behind.
Co authored by:
Senior Engagement Manager
Assistant Project Manager