A three-pronged solution: it takes management commitment, error evaluation and mobilization of new technologies to control rating error.

Rating error in personal auto lines costs the property/casualty industry nearly $14 billion every year. In my April 2004 column, I presented an overview of the problem and the associated financial, risk-management and moral hazard costs. Now, I want to review the available solutions.

Research conducted by Quality Planning Corp. has shown that effective control of rating error has three necessary components: management commitment, error evaluation and mobilization of new technologies.

* Management Commitment to Rating Integrity. The first necessary step to rating accuracy is to make it a priority. Lax underwriting is easy--underwriting discipline is difficult. In the absence of strong direction and commitment from senior management, day-to-day underwriting and sales operations tend to slip into a pattern of tolerating and then ignoring rating errors.

* Evaluation of Rating Error. The second step toward rating integrity is to evaluate current rating error. More than a decade of conducting premium audits has convinced us that every carrier is indeed unique. The amount and kind of rating errors vary by factors unique to each carrier: rating factors used in their plan; existing verification systems; characteristics of the book of business; geographic location; distribution channels; systems history; regulatory environment; relationships with sales agents; and underwriting standards. Carriers need to conduct a thorough analysis to determine their particular rating-error profile.

* Mobilization of New Technologies. The lessons from failures of mail and telephone questionnaires point to the solution. A four-part approach can address these problems:

1) Use automated error detection. It is necessary to move beyond shotgun approaches to correcting rating error and mobilize verification databases and pattern detection technology to target likely errors. Hundreds of databases exist that can aid in detecting rating errors. For example, it is possible to determine whether the address is a mail drop, such as Mail Boxes Etc. Similarly, we have used geographic databases to identify policies with farm use discounts in downtown Chicago. In addition, underwriting data can be subject to pattern analyses to identify potential errors. For example, decodes of work phones to addresses can be used to validate reported commuting distances. Each pattern is associated with the probability of a rating error. Automated error detection combines technologies of statistical pattern analysis, expert systems, and database verifications to provide an in-depth review of every policy at a low unit cost.

2) Custom underwrite a plan to correct likely errors. The optimal correction method varies by type of error. Some errors, such as giving a factory-installed airbag discount on a 1965 Chevy, can be corrected directly. Most likely errors require direct customer contact to update and verify questionable information. For such errors, it is necessary to develop custom interview scripts. For example, if the automated policy analysis discovered that a policy garaging address was associated with a contracting business and the insured vehicle was a three-quarter ton truck rated for pleasure use, a reunderwriting script should be developed that asks specific questions about that vehicle and its potential business use. The script should display knowledge of the issue so that misreporting becomes much more difficult.

3) Develop a cooperative, not adversarial, customer relationship. It is commonly assumed that most rating error results from intentional misreporting to avoid premium. This is not necessarily the case. Average insureds are unlikely to know what constitutes business use of their vehicle. They may not understand which members of their households should be rated on a policy and which should not. The definition of "principal user" of a vehicle is not always clear. In these circumstances, we have found that the right customer approach produces accurate rating data.

4) Employ modern phone center technologies. The fourth element of successful reunderwriting in personal lines is the mobilization of modern phone center technologies. Efficient out-going call centers can overcome the inefficiencies inherent in reunderwriting by staff underwriters. Fully scripted interviews preclude the need for highly trained underwriters to complete the calls. Scripts must be carefully written to promote a cooperative and non-adversarial customer relationship. Supporting systems can make mapping and other tools available to interviewers to support full error detection and correction. Calls can be rotated by time-of-day and day-of-week to provide maximum opportunities for successful customer contact. Interviewers fluent in multiple languages can respond to the range of customer needs.

Daniel Finnegan is president of Quality Planning Corp., a San Francisco-based provider of analytic methods to the auto insurance industry. He can be reached at [e-mail].