A cost-effective approach to casualty claims analytics

  • Print
  • Connect
  • Email
  • Facebook
  • Twitter
  • LinkedIn
  • Google+
By Ravi Kumar | 27 April 2012

An insurance company’s financial condition and its customer satisfaction levels are both decided in large part by the organization of its claims operations. Therefore, streamlining those operations can provide a significant business advantage.

This is especially true for small insurers. One of the strengths of a small insurer is its proximity to its customers. Oftentimes, the smaller the company, the greater the potential for designing its claims operations to meet both customer needs and profitability goals.

Claims analytics does not need to be front-loaded with high costs. The whole process can be designed in such a way that the effort pays for itself in a very short time frame.

Claims analytics can be an effective tool for streamlining claims operations, but is it also cost-effective? It may seem to be a high-cost, technology-driven activity affordable only to large insurers with big budgets—but it doesn’t have to be so. Claims analytics needn’t require a large up-front investment; an optimal approach is designed in such a way that it pays for itself very quickly.

This article discusses a cost-effective, results-oriented approach to claims analytics that can provide immediate benefits to any organization, and particularly to small insurers.

From business objectives to implementation

Claims analytics is an objective tool used by insurance companies to identify, correct, control, and monitor performance issues in their claims operations. Many large insurers have implemented sophisticated claims analytics solutions that seamlessly integrate their day-to-day business processes with their corporate goals.

Claims analytics uses empirical information available in data collected by the insurance company and helps answer high-value business questions about claims operations, resource allocation, vendor management, cost containment, litigation management, fraud detection, and other areas. For example, can we identify a potentially unexpected property, demographic, or customer characteristic consistent throughout a recent loss trend? Or, are there unusual attributes of properties that show up more often in a certain segment of claims, or in our handling of certain segments of claims?

Claims analytics as a sequential process

A claims analytics project has many phases, each requiring its own specific set of skills.

First, the claims department needs to define high-level objectives, which are derived from the company’s strategic goals, known issues facing the department, and a detailed understanding of day-to-day operations.

The second phase of the project consists of identifying the most relevant data sources and preparing the data for analysis. This data preparation phase is usually both people- and computer-intensive.

Then, statistical techniques are used to develop models that provide objective insights on the selected business problems. This phase is generally done by actuaries or other people with statistical skills.

Finally, the insights from the models need to be translated into business actions and implemented in real business processes. Implementation is both business- and IT-intensive.

Standard approaches to claims analytics (the old, expensive way)

Early adopters defined claims analytics as a project with a predefined, sequential set of tasks similar to a typical project undertaken within an insurance organization. As a result, standard approaches often apply best practices from project-based initiatives to claims analytics.


Because advancements in technology have made claims analytics possible, thought leadership has mainly come from technology vendors and consultants, and all phases of claims analytics initiatives are often technology-driven.

Large-scale data organization is emphasized in the data preparation phase, while, in the modeling phase, technology issues play an important role in statistical tool selection. Sophisticated models based on machine learning and other techniques rely on advancements in computing power.

In the IT-implementation phase, large resources are spent in areas such as enterprise-wide data warehousing and business rules engine and business intelligence implementations.


Typically, most projects use a sequential approach. Traditional software development life cycle methodologies have been used for technology selection, project planning, staffing, and response to business requirement changes. Business value is often delivered all at once at the end of a project.

In this approach, IT is given well-defined data requirements up front. IT prepares the data and hands it off to actuaries, who typically spend a few months analyzing the data and designing models. Then, the claims people are introduced to the models. Once the models are acceptable to the claims leadership, the IT implementation phase takes a few more months to integrate the steps into the claims process.

All these sequential steps would be appropriate if the claims analytics process was a well-defined process as predictable as widget-making. Unfortunately, claims analytics is more like a discovery process. As with any discovery process, we do not know beforehand which paths to take and which paths lead to dead ends. Many times we do not even know what we will be discovering.

In a sequential model, a trial-and-error-based discovery approach can be accommodated only to a small extent and at considerable cost. As a result, projects that are expected to take only a few months often exceed those timelines and run for years. Many projects realize very little business value compared to the investments made.

Large teams, large IT infrastructure, high costs, and budget overruns are common—but are they inevitable?

An alternative, results-oriented approach

A more effective approach views claim analytics as a business-driven, iterative, continuous activity, rather than as a technology-driven, sequential, one-time project.


It used to be difficult to find someone who understood claims and also the immense value of analytics, but most business leaders in the claims area now realize that valuable objective insights can be found in data.

Claims analytics is a bespoke craft used to gain a competitive advantage. Industry standard solutions or best practices borrowed from others will not provide that core advantage. Claims department leaders should take it upon themselves to develop a solution tailor-made for the specific culture and specific needs of their claims organization.

The claims department can leverage people from other parts of the organization and/or hire consultants who complement their internal skills, but it should make sure not to depend on those resources in the long-term. For example, a strong IT team can set up an initial infrastructure, but the claims department should not be dependent on IT for the day-to-day data needs of an analytical project.

Iterative and incremental

Claims analytics does not need to be front-loaded with high costs. The whole process can be designed in such a way that the effort pays for itself in a very short time frame.

The diagram in Figure 2 shows the different phases of a claims analytics life cycle. The goal is to experience all phases of the claims analytics process within a reasonably short time frame, all the while preparing to do it again very soon. Each iteration enables the team to learn and to improve. There is gradual cross-training across teams. Over time, the people who fit best tend to stay on the team, and the appropriate technologies are put in place. Business leaders are directly involved; they will learn in this process why certain decisions work and others do not.

These low-risk, small-budget iterations allow for the design of a long-term solution that is tailor-made to the organization’s unique needs. And, approaching a project in such an agile, iterative way can allow the team to potentially see benefits in a matter of weeks rather than the old-style approach which could take a year or more to begin to deliver value.

Iterative approach to claims analytics


To be effective, claims analytics should not be a one-time or once-every-few-years project. Models require constant calibration as claims adjusters, customers, and vendors react to changes already in place. In addition, claims analytics must incorporate changing business needs on an ongoing basis.

A core team should stay close to the business to provide the needed services and continue to play a role in every phase of the claims analytics process.


Practical and effective claims analytics strategies are vitally important for insurance companies. A business-driven, iterative, and continuous approach results in a solution that best serves their business needs.

Small insurance companies are uniquely placed to take advantage of this approach. They are already used to deploying small but effective teams to solve other issues in their organization and do not have a “bigger is better” mindset or the organizational silos that often affect larger insurance companies.

Small insurance companies can therefore more easily design and implement cost-effective solutions that are tailor-made to their unique needs—an approach that is long overdue.