Thanks to a compelling business case and a proven track record in banking and other industries, a growing number of forward-looking life insurance companies have embraced predictive modeling and are seeing tangible benefits from initial investments in new tools and processes.
To date, predictive modeling in the life insurance industry has primarily been considered for underwriting because of the cost savings and increased speed-to-issue. However, it’s now gaining traction in other areas and functions, such as sales and marketing programs, where the value proposition is similarly clear. If the experience of the property and casualty (P&C) insurance sector is any indication, the growing adoption by life insurers should come as no surprise. Indeed, there is a pervasive sense among life insurers that predictive modeling’s time has come.
Lessons from the P&C Sector
During the last decade, predictive modeling tools and techniques have caused significant shifts in the competitive landscape of the P&C sector, starting with underwriting, where advanced segmentation has resulted in greater growth and profitability. Because of this success, P&C carriers rolled out predictive modeling capabilities in other areas of the business. In product management, for example, it has been used in pricing, feature selection and class plan design. Within claims operations, P&C carriers leverage predictive modeling for resource allocation and anti-fraud efforts.
A number of industry observers and analysts have credited the sophisticated use of predictive modeling as the driving force behind the impressive market share gains and rapid growth trajectory of formerly “middle-of-the-pack” carriers. The alteration of the competitive landscape has confirmed that predictive modeling offers both top-line and bottom-line benefits. It is not surprising, then, that predictive modeling has become an industry-leading practice in underwriting, product management and claims. A few forward-looking companies have successfully expanded predictive modeling into risk management, marketing, and distribution, as well.
The life insurance sector has been laggard in the use of predictive modeling. Yet, leading insurers have begun implementing the same techniques and have started to realize tangible value from relatively modest investments. Effective adoption starts with a sense of urgency about taking action supported by a quantifiable business case and targeted applications based on sound actuarial and business principles.
The Business Case for Life Insurers
Within life insurance underwriting processes, there has been a concentrated push during the last several years to produce more immediate underwriting decisions for as many policies as possible, while maintaining a competitive product offering and profitable customer base. Traditionally, in order to qualify for the lowest possible premium, potential customers undergo a fully underwritten application process, which involves medical tests, such as invasive blood and fluid tests, EKG scans or full medical exams, based on insurance amount and risk factors such as age. Statements from attending physicians may also be required before decisions can be reached. The testing is typically expensive and time-consuming, sometimes taking weeks or months to produce underwriting decisions. Further, the long, traditional process has been a barrier to purchase for many consumers, with higher non-taken rates for those customers who are finally offered policies. Companies not wanting to utilize a fully underwritten application process have so far been limited to simplified issue and guaranteed issue underwriting guidelines. However, because the premiums associated with these guidelines are higher than the best premiums offered in the marketplace, most companies are at a competitive disadvantage relative to their peer group. That is why this option is just not feasible for many insurers.
To differentiate themselves from their competitors, some insurers turned to automated underwriting as a means to reduce costs and streamline these processes. Straight-through processing and electronic interfaces for policy applications eliminate the need for costly and error-prone manual data entry. Furthermore, automated underwriting can utilize electronic applications with “drill-down" capabilities. For many applicants, this means fewer questions and a faster application process. Workflow tools and business rules further streamline the process by automating the ordering of medical tests and by recommending approval of (and in certain instances actually approving) applications without underwriter involvement. Such underwriting systems have the added advantage of being able to automatically assign tasks and cases to the right underwriter at the right time, based on current workloads and resource availability. These capabilities allow underwriters to focus on the risk assessment of borderline cases.
Predictive modeling should be viewed as the next step beyond automated underwriting. Specifically, it streamlines and optimizes underwriting decision-making by applying business rules and enhancing traditional medical underwriting information with external data. This rules-based approach extends the process efficiency gains and produces underwriting decisions in a faster time frame and within a company’s existing underwriting infrastructure.
As early adopters of predictive modeling in the life insurance sector have shown, external data provides an important set of predictors. These additional data sets include prescription history, motor vehicle reports and family history, as well as relevant lifestyle data, like exercise and diet. But early adopters also use complementary data sets, including:
- Demographic information: population density, medical care index;
- Personal attributes: gender, age, occupation, education, marital status; and
- Finances: assets, income, credit history.
Such information is generally less expensive and more readily available than information produced from traditional underwriting tests. The end result is that insurers with the right infrastructure can deliver underwriting decisions to many applicants within minutes or hours – not weeks or months.
In evaluating any data source or element used in the underwriting decision, consideration needs to be given to the importance and acceptance of those data sources. The data source needs to be predictive, but also meet certain public acceptance thresholds and associated legal requirements. However, even when a predictive data source does not meet the acceptance and legal criteria, there are often alternative data sources that can provide nearly the same predictive power, while being publicly and legally acceptable.
Moving Past the Challenges
When considering the deployment of predictive modeling, the most common concerns for insurers are related to pricing, anti-selection by agents and reinsurer attitudes toward these risks. From a pricing perspective, the primary concern involves the risk classification derived from underwriting based on predictive modeling versus that from traditional underwriting. If the goal is anything other than replicating the current traditional underwriting techniques, life insurers must account for all differences that may develop.
Specifically, they will need to estimate the impact in underlying pricing assumptions and the resulting effect on profitability in determining whether premiums should be adjusted to reflect underwriting changes. Balancing expense savings, mortality risk premiums and premium growth potential will be important when pricing new products. Obviously, pricing actuaries have an important role to play in designing an optimal approach to predictive modeling underwriting initiatives.
In terms of agent anti-selection, the critical step is to compare risk-scoring results of individual agents to the risk-scoring results of all agents. The risk-scoring results will include the attribution of predictive modeling scores and traditional medical underwriting scores associated with final underwriting decisions. By analyzing the distribution of risk scores, it is possible to identify outliers and to flag agents who are at the extremes. Remediation steps to address anti-agent selection can include everything from changing of underwriting rules to an evaluation of the agency force.
Finally, some life insurance companies may find that their reinsurers are not willing to offer the same competitive reinsurance rates for predictive modeling underwriting. In such cases, life insurers may wish to apply predictive modeling only in underwriting polices beneath their retention limit. Alternatively, they can seek to negotiate experience refunds for business underwritten using predictive models. The experience refund allows the insurance company to secure reinsurance coverage through their preferred reinsurance provider at a higher premium relative to traditional underwriting techniques. However, if their actual experience develops as expected, the life insurance company then receives a refund on that higher premium. Over time, as experience develops and the reinsurer confirms that underwriting is performing within the company’s expectations, the need for the higher premiums and associated experience refunds will eventually be eliminated.
The value proposition for predictive modeling for life insurers starts with more efficient underwriting.
With automated underwriting and predictive modeling, the goal is to deliver instantaneous underwriting decisions for at least 60-80 percent of new life policies issued. Achieving this goal eliminates the cost of superfluous medical tests and reduces resource needs through the automation of manual tasks leading to greater cost savings.
Premium growth and increased sales are an equally important part of the value proposition. Generally speaking, there is high correlation between the amount of time it takes to underwrite a life policy and the non-taken rate. Policies that are the most time-consuming and expensive to underwrite are also the least likely to be taken. Generating more policies through a more efficient underwriting process will reduce the non-taken rate and increase new business production. In addition, by differentiating your speed to issue relative to your peer group, you have helped your distribution system to drive new business growth and, therefore, made your company more attractive to new producers.
Consider a company writing 100,000 new life policies with an average premium of $800 and underwriting and medical costs of $150 per policy. Cutting those costs to $90 and reducing the non-taken rate from eight percent to five percent would generate $6 million in cost savings and $620,000 in additional profit from increased policy retention. That is an eight percent performance gain on the original premium. In other words, predictive modeling can boost both the top and bottom lines.
Predictive models can be improved and enhanced over time as more data becomes available. Therefore, companies conducting initial feasibility studies would be able to address near-term gains and simultaneously take a longer-term view of the future value to be created as models become more sophisticated.
Looking more closely at the top line, there is also an opportunity for life insurers to bolster their target marketing and lead sourcing programs. Consider the capabilities of banks, credit providers and consumer packaged goods companies to deliver tailored product offerings to individuals just after major events, like home purchases or the birth of a child. Insurers could similarly target fully underwritten insurance options for customers at specific events and milestones, provided they have the capacity to identify the individuals most likely to buy.
Further, predictive modeling will enable lead sourcing programs to effectively balance potential policyholders in terms of risk, retention and profitability profiles. The resulting advantages include more efficient allocation of marketing budgets, improved alignment between marketing and underwriting and optimized acceptance rates of policies that are least likely to lapse. Finally, life insurers can gain clearer insight into the geographic regions with greatest sales and profit potential.
Life insurance companies ready to explore predictive modeling should consider the necessary steps, which have been proven enablers for smooth, efficient deployments at P&C carriers, as well as the leading-edge adopters in the life insurance sector.
First, gain a clear understanding of the value chain components driving the company’s profit and growth agenda in order to craft a feasibility study to build an objective business case. It is often a good idea to test profitability targets by looking back at the last two years of underwriting data to see how predictive modeling might have influenced expense savings, expected mortality and premium growth. Other operational areas (e.g., lead sourcing) may also produce benefits.
Second, brainstorm with senior leaders to chart a course and identify potential initiatives where predictive modeling can deliver the greatest value to the organization. This can also create the necessary cultural consensus to drive large-scale change. Predictive modeling can represent a fundamental shift from the traditional ways of doing business. Therefore, senior executive sponsorship is critical. That is especially true if the rollout of new capabilities affects relationships with agents.
Third, life insurers should assess their current data and technology assets and any specific infrastructure enhancements that may be required. Some process redesign may also be necessary. Insurers that have invested in maintaining and enhancing robust customer data will likely have a shorter path to implementation and, ultimately, a quicker payback on their investment.
But most importantly, life insurers must recognize the imperative to act. Since adoption rates are likely to increase rapidly, predictive modeling will soon reach a tipping point in the life insurance sector as it moves from competitive advantage for early adopters to standard operating practice for all life insurers. The time for action is now.