Unlocking the secrets to technological transformation in claims
Analytics are changing the way carriers identify and address high-risk claims, and saving them money in the process.
Last year, and as part of a global initiative called Brilliant Basics Claims (BBC), QBE mobilized and launched a functional, multi-year, data-driven claims transformation program to extend and enhance our analytics capability and generate enhanced data and analytics enablement across all global claims divisions and operations. The success of earlier divisional data & analytics team deployments compelled us to elevate claims analytics as a strategic and global area of focus and investment.
As part of the scope for this BBC Analytics program, we also redesigned our Claims Analytics Center of Excellence (COE) with a new business model and scope of services and re-organized/re-skilled our team. For example, one major change was the decision to deploy and share analytics resources on a global basis across different claims divisions, where the analytics talent and tools could be put to work in much closer proximity and collaboration with our claims business stakeholders.
Finding solutions
In our Australia/New Zealand claims division, work had started around a specific use case targeting better recovery of reinsurance on catastrophe claims. The team had experimented on this use case with data analytics using simple business rules for a couple of years and then advanced to incorporate machine learning algorithms. The division’s approach yielded extraordinary insights that culminated in millions of dollars of reinsurance payments that otherwise might have slipped through the net.
The Australia/New Zealand claims division developed an algorithm for use in a catastrophe reinsurance context — the reinsurance we buy to spread the property and business disruption risks we assume from policyholders. This coverage triggers when our underlying aggregate property losses breach a specific catastrophe loss threshold.
The algorithm combed through targeted claims and accident years in combination with our treaty contracts with reinsurers to identify claims that fit the profile of resulting from a catastrophic event but had potentially been overlooked. Due to human error, the oversight came from a simple coding mistake or the claims professional may have missed the catastrophic characteristics of a claim, resulting in its misclassification. Consequently, claims that fit the profile of a potential “catastrophe loss” were missed, causing significant reinsurance leakage.
Added up, these claims exceeded the contracted threshold for reinsurance reimbursements by double-digit millions of dollars. To be fair, in no way could our busy claims staff and reinsurance team manually dig through thousands and thousands of claims to discover and reconcile these missed opportunities.
Algorithms, on the other hand, can rapidly look for claims that align with known catastrophe descriptors. By knowing what to look for, the algorithm ferreted out claims for further analysis and verification by the divisional claims staff. If determined to be valid, the information was then passed on to our reinsurance team for further validation and financial reconciliation purposes.
The impact of analytics
Given the financial value realized from the analytics in the Australia/New Zealand use case, we quickly pivoted and focused the new claims analytics COE on this use case to accelerate and replicate it across our other global claims divisions. This was accomplished successfully and in a fraction of the time it would have taken had this opportunity been left to each of the divisions to handle independently.
Our abiding interest in data and analytics is driven by the remarkable value of these tools in a claims context. Like other insurers, we collect huge volumes of information from our customers and partners such as brokers and third-party claims administrators, in addition to external sources of data. This wide-ranging structured and unstructured data resides in thousands of databases. Some of this data is incorrect, out-of-date or misclassified — a forgivable consequence of human error, given the hundreds of files a single claims professional typically manages.
In many ways, data analytics is a backstop — a technical safety net. Errors and oversights that people fail to catch, the algorithms catch. This is not to say our claims professionals no longer need to perform these reviews; they do, but now they have a better toolset, running an algorithm to ensure nothing was overlooked and all the numbers line up correctly.
A key goal of the BBC analytics program is much more uniformity in the use of data analytics by each claims division. Each division has its own data and analytics team. In the past, if an analytics team in one division built a successful algorithm to address a particular business problem, there was little structure in place to facilitate its adaptation for other claims divisions. No longer is this the case.
We are building a storehouse of remarkable algorithms and playbooks for use and reuse by all divisions, including algorithms that detect the possibility of fraud, speed up workers’ compensation claims settlements, inform more accurate loss reserving, and determine the likelihood of a claim resulting in a lawsuit — to scratch the surface.
We’re also experimenting with the use of analytics to triage complicated claims according to their potential severity. In the past, this assessment was based largely on human experience and intuition, which often resulted in a longer claims adjudication process that increased the likelihood of litigation. Once completed, an algorithm will illuminate complex claims requiring deeper and more immediate evaluation for mitigation purposes.
Following a new roadmap
Going forward, we are exploring other data-driven analytics use cases. For example, we are experimenting with analyzing customer satisfaction scores relative to claims service broadly — at call centers and in the field. We want to determine why it might take two weeks longer for some customers to receive payments for what are essentially the same or very similar claims.
In this quest, we’re developing algorithms to analyze various communications and documents — e.g., case management reports, loss adjuster notes, loss inspection reports, repair estimates and call center recordings — using text mining and Natural Language Processing (NLP) tools.
What is most exciting for us is what lies ahead. The sophistication of the technologies grows by the day. Two years from now, what we’re presently doing may seem rudimentary, given the breathtaking pace of our digital and data transformation. It is a great time to be in insurance.
Jim Kinzie (James.Kinzie@us.qbe.com) is the global head of the claims analytics program at QBE Insurance. Over the past 30 years, he has combined his experience in claims and operations with his expanding knowledge of data analytics to several insurers’ digital and data transformations.
To learn more about how technology is transforming the claims process, join us at the America’s Claims Executive Leadership Forum & Expo in Las Vegas, Nevada on June 24-26.
Related: