Jim Kaiser is CEO of Casentric.

The use of data to gain deeper insight into claims performance is growing rapidly. Technology and tighter process management is fueling this growth. Legacy system replacement provides offers the opportunity to collect more data as do other solutions such as medical bill review, auto and property damage estimating or injury and liability evaluations. Business intelligence systems designed to improve our ability to consolidate and analyze this data are growing in capability as well.

Before plunging headlong into the sea of data, it is useful to take stock in the historical challenges of data analysis in claims to avoid having history repeat itself. In particular, it is vital to understand what kind of sense can be made of the data we create—can it tell us “why.” To explore this question, this article lays out data goals, the challenges that have impeded achieving these goals, and comments on how these insights impact the next generation of data analysis.

The Goals of Data Usage in Claims

In order to effectively consider and use data, it is critical to appreciate the goals of using it at all. As the Cheshire Cat famously pointed out, “If you don't know where you're going, any road will take you there.” This is important to mention because discussions about data these days can quickly gravitate to functionality—the path you can take—instead of outcomes—where you are going. By making this leap, it is easy to overlook the challenges data has presented to reaching the destination.

The goal behind collecting data is to be able to understand why and how a claim reached its ultimate resolution. This tells us whether a claim varied from a desired outcome, why and by how much. By understanding variance, we can take action to reduce it and, thereby, improve the accuracy and timeliness of settlement.

Historical Challenges with Claims Data

The temptation to design information capture that is overly broad is great. This is driven by efforts to “force” consideration of important claims factors and, at the same time, account for all possible fact sets. It is also necessitated, in some cases, by rules engines that require the data to produce outputs.

On the most basic level, forcing the collection of more data generally has limited effect. Adjusters and managers start to suffer from data-gathering fatigue. The consequence is data collection that is rushed and often inaccurate. Because of this, requiring more data collection is rarely effective at exposing variance or its causes.

The more significant issue for analysts is the data's relevance to understanding variance. Most often the data collected represents “attributes” or facts of a case. For example, liability attributes can include such items as the location where the loss occurred, the number of traffic lanes or speed limit. This information tells us the “what” in a claim, but does not tell us “why” a case was resolved for a particular amount.

It is, for example, far more important to understand whether a party failed to yield right of way and what evidence supported that decision than it is to detail the nature of the loss site. Unfortunately, this “why” information is largely lost in unstructured claim notes, if it is ever captured at all.

On the other end of the data collection spectrum lies interfaces designed to run rules engines. While these solutions can be useful, the data types collected are rigid and tend to mask the relationship between the data entered and the output produced. Aside from thwarting learning (removing cause and effect relationships has this effect) they can also require organizations to focus excessively on managing accurate data inputs rather than successful claim outcomes.

The issue presented by analysis of data that lacks a connection between attributes of a claim and its outcome is that it leads to broad reactions rather than surgical intervention. These reactions, in turn, generate unintended consequences—a whole new variety of issues. For instance, if I examine average personal injury settlement amounts in cases where the payment for property damage was low, I might rush to a judgment that low impact cases are improperly valued. This can cause a new set of procedures and, possibly, more data fields.

If, however, we know how low impact information is being used to adjust the value of the case we achieve both a well-reasoned case and tremendous strategic insight. For instance, capturing the fact that low impact information was used, how much it affected value, and the reasons it impacted value by that much, our comprehension of variance is immediately raised.

Obligatory data entry of marginally relevant data also creates challenges for understanding how a claim developed. When users are “forced” to fill out screens, they tend to put it off as long as they can. This means loss of insight into when information was received and it affected the course of the case at that point. For instance, it is very useful to know how and why a claim's value changed between the evaluation prepared to discussion resolution and the amount for which a case was actually resolved.

Data Planning and Design

To serve the goals we outlined above, and others, claim data must move beyond “what” to provide insight into the following questions:

Why did we arrive at a value?
What path did the claim take to arrive at resolution?
Who made provided the justification?

By re-thinking the data we capture, we can simplify it. While it is not an easy task to catalog the conclusions an adjuster reaches, there are far fewer conclusions reached and reasons given for those conclusions than there are facts and attributes to a case.

There are only so many reasons that an individual can be negligent in a loss, but there are myriad circumstances in which that negligence can arise. In a world where we have less time to analyze and act on data, having less data that provides deeper insight allows more rapid and specific identification of issues.

Special Reports /

Deep Impact: Using Data to Analyze Claims Performance

Related Stories

Recommended For You