While the general market impression of the effect of COVID-19 is that new life insurance business demand has dropped in most product areas, protection products appear to be a significant exception.
Such a glimmer of good news is of course welcome, but it does come with some new challenges for underwriting and pricing that could determine how safe it is to write more protection business in the short term and how resilient underwriting remains in a medium- to longer-term, post-COVID world.
For example, how do you get a potentially increased volume of quotes out quickly, accurately and efficiently through your distribution networks to meet demand and competition in the current environment, particularly if your process involves a lot of messy, manual steps? How do you manage a medical underwriting approach where perhaps medical attendance reports are difficult to obtain or being dispensed with out of pragmatic necessity? Or how do you ensure your underwriting is flexible and agile enough to reflect emerging COVID-19 infection and claims trends, not forgetting those of other medical conditions?
Harnessing analytics and automation
Such questions are prompting more insurers to investigate the sophisticated analytics, underwriting and pricing - perhaps incorporating elements of machine learning and artificial intelligence (AI) – that have increasingly become a competitive necessity in many property and casualty (P&C) markets around the world. Indeed, around 75% of the competitive UK protection market, for example, is already served by companies that are using some or all of Willis Towers Watson’s award-winning Radar analytics and pricing suite. Interest extends beyond underwriting and pricing too. Some life and health insurers are also pursuing the use of analytics and automation to better understand trends in experience and customer behaviour that can support, for example, better cross-selling.
In the core areas of pricing and underwriting, key attractions of such investment include enhanced pricing flexibility and agility, more detailed and responsive management information, better market responsiveness, material operational efficiencies and cost savings, and reducing the risk of costly pricing errors. There’s also the not insignificant fact that prices that could previously take up to six months to update, often with extensive back and forth with the IT department, can now be revised within hours, complete with full review and governance. Even changing rating structures or adding new rating factors can be done quickly and easily.
The diagram below shows, by way of illustration, how such benefits are achieved in an agile underwriting process in Radar.
Policyholder information
This is where a customer requests a quote and is asked to provide information relevant to the policy. The system can pull in information from a wide range of sources, so quote information can potentially be enriched with data from a variety of other sources such as Fitbit outputs, zip code/postcode information, distributor information and, where permitted, personal information from other data sources. Automatic data queries will highlight any anomalies in data, such as an unknown address or an unusual sum assured request. The system can also be set up to incorporate automated decision points about whether more information is needed for cases such as where an applicant has a specific named condition or where the sum assured is above a particular threshold.
Risk rates
With the customer data gathered and prepared, attention moves to building the underlying view of the risk of that customer. In other words, to put it bluntly, how likely they are to die or claim.
This could happen in several ways, including calling the existing rate structure, but could also potentially use generalised linear models (GLMs) that incorporate claims experience, models that incorporate medical judgement (such as Willis Towers Watson’s PulseModel), reinsurer rates, machine learning and AI-generated information, or a mix of some or all of them. Significantly, underwriting rules can be changed in a matter of hours if necessary, say to accommodate specific constraints or insights associated with COVID-19, including new questions that might inform an applicant’s susceptibility to the virus.
As an example of the benefits that can come from these approaches, some companies with which we’ve worked that have introduced GLMs have found the increased segmentation capabilities provide risk differentiation of around +/- 50% (identifying that high mortality groups show double the mortality of low mortality groups).
Premium calculation
After the risk assessment comes deciding what you are going to charge. Even with a simple expenses-plus type approach, a loading for risk based on a more in-depth assessment of strong data is likely to have advantages. Similarly, rate structures that are based on supply/demand or attaining a certain competitive position in the market can be applied with less risk of adverse selection using customer demand behaviour data. Such data also opens the way to price optimisation in markets where it’s possible.
Perhaps the key thing though is the speed, flexibility, efficiencies and ease of governance the system provides. It can support updating prices daily, for example, including producing the up to date analysis and management information required to make frequent price changes. It also allows insurers to change rating structures easily, and to ease the transition from a fixed rate table to more advanced formula or algorithm driven pricing.
Market distribution
All of the technical wizardry in the world is worthless if you can’t get your rates to the market in a timely manner. Using industry standard API technology, the system can be called by both internal and external systems, providing channel-agnostic digital distribution. It also maintains complete transparency and flexibility over variable changes to pricing for the likes of advisors, affinity partners such as banks, and price comparison sites. Specific benefits for different distribution options could include making real-time information available to call centre staff and agents about factors such as likely customer value and upsell potential or combining company and agent data to produce channel-specific analysis.
Feedback loops
The system allows information on new business, underwriting decisions, lapses and claims to be fed in to create bespoke management reporting templates quickly and easily. These can be tailored to specific business metrics such as agent performance and market segmentation and used to inform pricing and business decisions. In the case of COVID-19, such reporting could provide early warning signals of claims trends related to, say, socio-economic grouping, occupation or geography, that might influence the underwriting and pricing approach.
The feedback loop also replicates traditional life actual v expected methods, allowing the import of standard tables. In combination with credibility results, the analyses can then feed into the revision of rates used for new business in the risk rates component.
Insurers need protection too
This overall level of sophistication remains the exception rather than the rule in life insurance, but that doesn’t mean that some companies can’t get left behind, especially given the relevance of new operational factors introduced by COVID-19.
The reported surge in interest for protection products relative to other life and health insurance product lines as a result of COVID-19 brings both opportunities and risks. It could indeed be very positive for insurers that have systems that give an in-depth view of the risks and that are able to underwrite and price at the pace and with the agility that will give them an advantage.
For others, the big risk is that ponderous and inflexible systems will result in them being left unwittingly with higher or incorrectly priced protection risks that stand a bigger chance of coming back to bite them some time down the road.
Insurance Consulting and Technology
