Leveraging Customer Value Management

How insurance firms respond to these challenges in this economic landscape can have repercussions affecting not only short-term profitability, but also long-term financial prosperity and survival. Many facing this new economic reality will be forced out of specific market segments or whole markets because of poorly defined customer management, risk, pricing, and customer service strategies.

From the insurance industry's perspective the integration of different product lines into a single business unit sounds like a prime opportunity to reduce operational costs, grow revenues, and improve profitability. In practice, however, achieving this convergence is a much more difficult task. There are significant data challenges, training issues, product silos, incentive pay policies, internal processes, and cultural barriers to overcome, including:

Data: Available at the product level, "owned" by different departments. In many cases, lacking a common customer identifier (index) to enable an integrated, enterprisewide view of the customer.
Sales force: Fierce internal competition for customer "ownership." Marked territoriality and resistance toward sharing information across departments.
Service functions: Highly specialized with a strong product orientation.
Incentive pay: Cultural practices that encourage a product-focused culture to the detriment of a more holistic customer view.

The complexity in integrating these different and often adversarial functions explains in part the limited success the industry has had in their cross-selling efforts. Therefore, one of the most pressing and daunting challenges facing insurance industry executives today is providing actionable, cost-effective solutions that answer the following key strategic questions:

The need has never been greater for solutions based on an improved understanding of customer behavior and customer lifetime value. An integrated view of claims and behavioral data contributes to an improved understanding of customer potential value by emphasizing granular risk segmentation and by developing predictive risk models that better reflect the underlying risks of policies being underwritten.

The benefits of a proactive view of customer value
Few industries offer the rich variety of customer, product, sales, marketing, geographic and multichannel usage data as the insurance industry does. From the moment the phone rings at the customer service desk, a claim is filed, a claims adjustor is dispatched, to the time a claim is paid or settled, opportunities are abundant to improve the bottom line through process automation and improved risk-based decisions.

Whereas insurance companies have made significant IT investments and are increasing their adoption rate of CRM technologies to capture enterprisewide data, the industry lags others in leveraging this information to improve customer profitability and gain competitive advantage in the market.

Customer Value Management (CVM) views data as a strategic resource that enables the organization to make better and more informed decisions. Its focus is prediction and explanation: the identification, quantification, and prioritization of factors having an impact on the value chain and customers' current and potential value.

Being able to accurately identify and rank order customer characteristics and behaviors having the highest impact on risk, premium leakages, and servicing costs can yield immediate, bottom-line benefits to insurance companies. This information, in turn, can be leveraged in many forms across the enterprise: adjusted product premium tiers, revised customer contact rules, incentive pay alignments, sales and training programs.

By and large, many organizations attempt to address these data integration and data analysis challenges internally. However, more often than not, these efforts fail to meet ROI expectations. The reasons for this lack of success are several:

Data quality: Legacy systems still dominate much of the functionality of claims, risk management, and billing at many large insurance carriers. In many instances, these systems have been patched together and lack a robust, operational architecture. As a result, it is difficult to maintain data integrity across all of them and coding errors are common.
Rating errors: For example, industry statistics show that rating miscalculations cost auto-insurers over $13 billion annually (nearly 9.6 percent of total personal auto insurance reserves nationwide).
Fraud: Of the $369 billion in P&C premiums collected by the industry last year, more than $29 billion was lost to fraud. It is estimated that 10 to 20 percent of claims are fraudulent or contain exaggerated claim information.

Industry-wide, technology adopters have responded to these challenges and have deployed specialized underwriting databases and automated rules-based systems. Results, however, have been mixed. While these systems have in some cases improved claim-processing efficiency, the bulk of revenue leakage and premium loss due to these problems remain unresolved. The reasons for this lack of success are several, however, the "curse of data dimensionality" seems to be a common underlying factor.

This concept explains in part why it is difficult to find intuitive patterns in data and incorporate them into rules-based systems, except perhaps in cases involving extremely low dimensionality, for example, two or three variables. Beyond this number, it becomes extremely difficult, even for the most experienced insurance personnel, to find obvious, statistically valid patterns and relationships among variables. Typical rate-making decisions in the insurance industry usually require sifting and analyzing through hundred of variables to identify the few that matter and herein lies the problem.

To illustrate this point, consider an over-simplified automobile insurance rate-making example involving 10 decision variables, each at two levels:
Policy coverage
Claim frequency
Claim severity
Mileage
Policy age
Vehicle weight
Driving record
Risk
Age
Gender

On one hand making independent decisions about the impact of these variables one at the time can be misleading because it fails to incorporate linear and nonlinear interdependencies among them. On the other hand a full factorial enumeration yields 1,024 risk cells, clearly, an intractable number.

CVM methods, however, make full use of the information space contained in these 1,024 risk cells, but do so via highly optimized search methods that require only the examination of a few, statistically selected nonredundant conditions. In this example, 10 "orthogonal" combinations selected from the full outcome space can provide information that is equivalent to a full examination of each one of the 1,024 rate-making cells. Clearly, with this highly optimized approach insurance companies can more quickly identify which factors are significant to drive optimal segment risk decisions, improve pricing and refocus sales and marketing programs.

Even though expert judgment encapsulated as automated underwriting rules can be leveraged to simplify the outcome space, such an "experience based" approach is unlikely to yield consistent results that account for interactions among claims risk factors across time. Consequently, results so derived yield sub-optimal returns because of an inherent static bias, representative of "average" risk factors that automated underwriting decision rules represent.

Plugging the sources of revenue leakage
There are two things that matter most: reasons and results. Reasons do not count.

The transition from an operational (retrospective) approach to a proactive view of customer value management need not be a radical "big bang" solution to shake-up the whole organization. On the contrary, efforts that yield the highest benefits are those based on a measured and graduated transition plan that starts small and gains momentum gradually, driven by market results. Successful CVM projects are driven by the following basic principles:

Data: Less is more. Often, most of the requisite data is already available. Data extraction via stratified sampling plans that meet specified precision and reliability parameters.
Insights: Tailored analytics based on segmentation and predictive modeling.
Action: Quantification and prioritization of recommended market test pilots and response optimization.

In contrast to solutions that emphasize the creation of large and specialized data warehouses, our experience in the insurance industry indicates that small datamarts based on representative statistical samples from data already available in the organization can provide valid and reliable information to drive analysis and facilitate identification of the key drivers of revenue growth.

The result of this proactive CVM approach to insurance are:
Risk segments: Groups of customers and policy claims having similar risk profiles and identification of key risk factors and value differentiators across segments.
Propensity-to-claim scores: Predictive models that quantify the likelihood of submitting claims, size of claim, fraud propensities.
Loss elasticities: Quantification of the potential revenue loss due to unit changes among the risk factors.
Prioritized Market Test pilots: Recommended list of potential pilots to test in the market.

The strategic benefits to those organizations that adopt a CVM approach to improve the bottom-line are many and are centered on improvements in three major areas:
More effective acquisition campaigns: Improved response rates via highly optimized targeting lists aimed at profitable, high potential value prospects.
Increased customer retention: Reduced customer attrition through proactive modeling of customer defection risk conditioned on expected customer lifetime value.
Improved Cross-selling: Greater customers' "share of wallet" through targeted offers that meet customers' evolving security and financial needs.

In addition, the organization benefits from institutionalizing the use of relationship based variables, as part of the decision making process that will contribute to:
Reliable metrics of Customer Lifetime Value (CLV): Improved algorithms that incorporate the propensity to respond to product offers, attrition probability, propensity to lapse, propensity to file claims, and fraud risk scores.
Greater risk granularity: Detailed segment level risk granularity that facilitates an improved understanding of loss exposure across smaller behavioral groups.
Improved underwriting: Faster, accurate, dynamic underwriting models that complement traditional, experience, rule-based systems.

The integration of transaction activity, product usage patterns and customer preferences invariably leads to other improvements in customer lifetime value and profitability, as well.

A winning strategy for improved claim risk prediction and improved customer profitability
Without segmentation you are spending too much money on some customers and too little on others, do you know which ones?

The CVM approach expands the scope of actuarial information by supplementing it with external relationship-based customer data and by deploying a disciplined market test process. This relationship view of customer lifetime value is what we call the next evolution in revenue management. It helps insurance companies realize the many benefits from positioning their products and services to meet customers' evolving security and financial needs.

The analytical methods of CVM have as their main objective the identification of the key drivers of customer lifetime value: loss ratios, reserve requirements, defection probabilities, claims lapse, and fraud propensities. The goal is explanation and the formulation of winning strategies to help insurance companies strengthen their profitability by identifying and eliminating low-hanging revenue leakages in the claim risk underwriting process.

The components of this winning strategy are straightforward and will allow leading organizations to:

This material is property of DiamondCluster International. Further reproduction or distribution is prohibited without permission.

Jay Norman is a partner with DiamondCluster International and leads the Financial Services consulting practice. He has worked in the financial services industry with leading insurers on a variety of business, operations and technology engagements throughout his career. Specifically, he has worked extensively with clients on the topics of customer analytics and customer value management. Contact him at Jay.Norma@diamondcluster.com.