Unleashing the power: a streamlined approach to data warehousing enables insurers to use claims data as a competitive advantage.

Many senior leaders at insurance companies have tried with varying success to unleash the power of data with major information-technology investment and long, complicated implementation time lines. Particularly with claims, the losses and expenses of which typically consume 60% to 80% or more of the premium dollar, they suspect that there are clear and considerable opportunities to make claims data pay big dividends.

Those suspicions are correct-drilling down into loss adjustment expense and severity can uncover important trends, enable new capabilities and identify opportunities for generating significant returns on claim technology investments. Yet, many carriers justifiably have grown leery of the operational costs and risks inherent in such data projects, having experienced the difficulties of building large-scale claim data warehouses. With huge amounts of data and transaction information spread across many different source systems, files and documents, the potential pitfalls are numerous.

"Typical Insurance Data Warehousing Environment," at right, is a visual depiction of a typical data warehouse installation that would be considered problematic and suffer from many of the pain points noted above and in "Risks of Large-Scale Warehousing" on page 74. It represents the typical "dumping" of legacy data from many different sources into a massive data-warehouse environment. It should be noted that, where data warehouses are effectively designed and data is managed properly, the data warehouse acts as a good agent to supply effective decision-making data.

Data warehousing without tight quality and governance systems, however, can drive significant escalation in operational costs and create serious system performance issues. Also, failure to redesign jobs and business processes to take full advantage of new information and capabilities often contributes to decreased productivity, potential gaps in customer service and lost opportunities.

A streamlined alternative approach to developing large-scale data warehouses can avoid some of these problems. The following discussion will use auto claims processes and technology to illustrate the significant value that can be generated through the strategic use of data mining. The opportunity is universal, however, and does not stop at auto claims, but rather extends across multiple business functions and industry lines of business including life, health, property/casualty, specialty and reinsurance.

Better Data in Real Time

The key to capitalizing on the strategic power of data is the organization's ability to quickly analyze key business metrics to support operational analysis, decision making, benchmarking and new programs development. The ability to predict outcomes and adjust strategies, plans and activity in real time rather than react to key performance metrics will provide a significant competitive advantage to those companies that use data efficiently and effectively.

A number of development alternatives can mine information through the use of business intelligence tools. Solution-specific data marts can be built to provide repositories for various domains of data and associated extracts. A data mart brings together disparate platforms and database management systems such that a trained user can access structured or ad hoc queries to meet unique business analytics requirements.

"Streamlined Alternative Approach" above, is a graphic representation of the model. Creating numerous smaller data marts allows users to address various reporting needs while allowing access to more manageable databases. Disparate data is "piped" to multiple data marts with appropriate business data dictionaries providing common definitions of the data (semantics). An overall governance framework

shown at the bottom of the illustration ensures all stakeholders understand the metrics, rules and models.

The Streamlined Alternative Model provides a user-friendly environment that the business can use with minimal training and involvement from the IT department. By implementing this model, increased confidence in the data from the business community occurs as a result of better usability, performance and quality of information. The typical cost for data-mining-related projects using the streamlined approach can be about 15% to 20% or less of a large-scale data-warehouse project using the typical data warehousing approach. This could decrease the overall expenditure by hundreds of thousands or even millions of dollars. Additionally, multiple departments or business units are no longer on the same critical path under the Streamlined Alternative Model, resulting in faster implementation time for business customers focused on a solution to a specific need.

Strategic Claims Management In Auto Repair

In the United States, more than $63 billion is spent per year on automobile physical-damage claims. There is tremendous opportunity to use data mining to effectively help direct scarce resources responsible for managing the total spending in this area. Insurance companies typically work with thousands of auto-repair facilities. Given this scope, it is critical to become increasingly adept at analyzing all of the variables involved in managing and paying claims throughout the claims management process and auto-repair supply chain.

For carriers, on the loss cost management side, other opportunities exist in the areas of investigating, evaluating and paying claims to minimize costs and become more efficient. Access to customized data regarding automobile physical damage claims is a big part of the answer.

Zeroing In

Repair costs, and hence loss severity, can vary considerably between auto repair facilities and types of vehicles for many reasons. The costs continue to escalate in the United States, with the average personal lines claim for auto physical damage being around $4,000.

Effective data mining enables oversight of higher risk claims by analyzing data in real time. This oversight process determines where field inspectors, damage estimators and other analysts can have the most beneficial impact on particular loss events. It can analyze every repair-estimate line item and flag material deviations or variances from expected standards or professional agreements. Through wireless access and mobile computing linked to global positioning systems, real-time assignments can be generated and modified to direct scarce field resources to those claims that offer the greatest economic opportunity for effective loss cost management.

Over time, underperforming repair facilities that are skillfully managed using analytics generated through data mining should potentially close the gap on repair estimate variances in order to qualify as premier repair vendors for a particular carrier. The net result is all parties involved win, from the insurance company to the repair shop to the customer. We estimate that the use of effective data mining combined with redesigned systems, roles and processes can yield total auto physical damage savings of at least 3% or $125 per auto repair. For large and small carriers both, tens and even hundreds of millions of dollars in loss costs can be avoided. Also, the longer a business analytics solution is in use, the smarter the underlying applications and users become through analysis of trends and the iteration of business rules. This yields additional opportunities for even more savings in the long run.

Analytics also can allow the company to identify certain vehicles or damage scenarios that have the greatest amount of claim variability and repair costs and flag them to have field representatives inspect those damaged vehicles more often, thus going a long way toward managing severity and claim variability. As opposed to random inspections or re-inspections, targeted oversight allows more efficient management of skilled resources. This enables improved decision making and ultimately lower claim costs.

Processing Claims More Efficiently

Claims managers and their organizations are typically measured in terms of severity management and loss adjustment expense optimization. It's important to pay the right amount for each claim, meaning no more and no less than called for pursuant to the underlying policy. Rules-based intelligence combined with iterative data mining is an important tool for managing claims. Using automated rules to determine what work gets done when and by whom drives operational efficiencies, thereby positively impacting both severity and LAE.

One example of leveraging rules and iterative data mining is the segmentation of claims based on historical experience trends. In other words, claim experience data can determine the optimal method of auto damage inspection based on customer value, claim history and loss information. The overall claims operating model for a carrier can involve many options for the inspection of a damaged vehicle--from carrier owned drive-in centers to independent and internal field adjuster visits to external repair facilities. Each inspection option carries higher or lower levels of historical severity and LAE.A nimble data mart supported with other enabling technology can produce significant benefits across all dimensions of cost and service.

Accurate auto repair estimates also are essential to reducing LAE. Rules can be derived by mining data related to repair shop trends or performance, negotiated rates, quality and customer satisfaction. The historical estimate data can be used in an oversight capacity allowing claim adjusters and management to hold shops accountable for certain levels of performance, including quality, customer satisfaction and cost containment. Even historical estimating trends from internal staff adjusters can prove valuable in terms of targeted training opportunities based on pattern analysis.

Detailed analytics give claims managers a better basis for repair approval in a given facility on a given claim. This can be a delicate situation since some repair facilities send in lower estimates to get initial approvals and then provide revised estimates for additional repairs or supplements. Some are adept at identifying supplemental repairs and understand how to stay under the supplemental limits which trigger queries and challenges. Historically, some repair facilities have been able to take advantage of insurance companies by doing unnecessary work, or over charging, knowing that field claim adjusters can't possibly look at every claim estimate.

Claims managers can look at historical data, line by line on an estimate if necessary, in a proactive mode to see which repair facilities tend to game the system, and intervene as needed to drive the appropriate behavior and service levels. In essence, data mining helps drive improved work-force management for field adjusters and supply chain optimization for carriers by helping determine optimal assignments and prioritized tasks.

Critical Success Factors

Regardless of the approach taken by an organization to mine its data, the keys to success involve effective governance, internal strategy alignment, use of qualified outside experts, attention to organizational change management, and measurable value that can clearly be attributed to the new infrastructure investments.

Governance is sometimes looked at as painful, but it is essential that an enterprise IT governance strategy, and an appropriate organization for implementing the governance strategy, be established with the backing of top business and technical leadership. Clarifying who owns the enterprise's IT resources, who the data stewards are, and in particular, who has ultimate responsibility for their integration is an essential part of a successful enterprise IT governance strategy.

Another key is job redesign. In many cases, some form of process and job redesign will be needed to fully harness the benefits and capabilities offered by data mining. New metrics and insights may also require changes to incentive and compensation programs. As with any major change, rigorous attention to organizational change management, including communication, training and transition planning, is critical to the success of the initiative.

Claims is clearly in every insurance organization an area that has broad opportunities to significantly lower loss costs and improve service. It is one of the best and highest yield areas to prove the business case for developing and rolling out improved data management capabilities across the entire organization. Accelerating time to market via the Streamlined Alternative Approach to large scale data warehousing can get you a big jump start on your competitors.

Key Points

* Creating numerous smaller data marts allows users to address various reporting and analysis needs by allowing access to more manageable databases.

* The typical cost for data-mining-related projects using the streamlined approach can be about 15% to 20% or less of a large-scale data warehouse project.

* Claims is one of the best areas to prove the business case for rolling out improved data management capabilities across the entire organization.

Keys to Data-Mining Success

* Effective governance

* Internal strategy alignment

* Use of qualified outside experts

* Attention to organizational change management

* Measurable value that can clearly be attributed to the new infrastructure investments

* Job redesign

Risks of Large-Scale Warehousing

Inaccessibility. Legacy systems present significant challenges to insurance companies when combined with large data warehouses, numerous business reporting subsystems and the disparate platforms they run on. Such challenges are further compounded by the problem of timely information access and usability. With dozens, if not hundreds, of data stores and silos, internal clients who wish to access data out of "mega" warehouses too often experience frustrating delays. They lack direct access to the data, and requests cannot be filled in a timely manner. Even worse, when the data arrives it is unusable without extensive reformatting and manipulation. That data, from many sources, is typically dumped in the warehouse with a variety of definitions, standards and reference models.

Significant Costs. Data warehousing projects can suffer from many of the same pitfalls as major application development efforts. It is not uncommon for initiatives to take years and total tens of millions of dollars in development and maintenance costs. Unclear business and technical requirements, inadequate project management, failure to identify measurable benefits, and lack of comprehensive process redesign and organization change management are typical obstacles encountered. These all add to the time, risk and expense of implementing a large-scale data warehouse. Ultimately, competitors will pull away from companies that have not architected efficient and effective data strategies and related business operating models.

Contributors Dan Spaeth and Robert Robertson are in the Insurance & Claims practice of Navigant Consulting. Spaeth, a director, can be reached at [e-mail] Robertson, an associate director, can be reached at [e-mail].