Process mining: Uncover how well your insurance processes are really working

It's fair to say claims organizations have had a lot on their plate in 2020, as a year dominated by a worldwide pandemic continues to deliver pressure and unknowns for insurers. With a gloomy economic outlook ahead, it's important for carriers to operate at peak efficiency and balance the holy trinity of expense, indemnity accuracy, and customer service to thrive in the future. And in a year that has seen the shift to online become mainstream, this includes identifying the right digital assets to help deliver real improvement and change.

To make effective digital investments, companies traditionally first gather resources to conduct human-led process deep dives to identify areas ripe for transformation.

But I think there's an alternative to this approach. Insurers should give serious consideration to applying process mining instead to identify opportunities for transforming operations. Process mining aims to discover, monitor, and improve real processes, not assumed processes, by extracting knowledge from event logs readily available in today's information systems. Doing this can ensure that organizations are successfully applying the right digital interventions, automations, and improvements.

Process mining software enables companies to see the different steps across a process and captures the view of how any employee is accomplishing the process. Capturing all this system detail paves the way for analytics, supported by machine learning and artificial intelligence, to build an overall view of the process patterns. It produces views of the most common paths taken and identifies bottlenecks and other common process variations. It really does visualize what's going on behind the scenes, such as frequency and time in process perspectives. It helps to identify problems such as leakage, delays, and waste by looking at all the data, not just a sample from a handful of claims resources. It presents the data in an objective view, without the bias of subjectivity that can sometimes affect root cause analysis.

For me, process mining is most interesting in how it shows all the different ways employees work on a process. It will show where the common standards in processing are and reveal how often it breaks down into inconsistent workflows across employees. For example, when it comes to subrogation, claims handlers need to recognize the issue, identify recovery potential, chase a claimant for their information, and pursue recovery efforts. What process mining revealed for one of our clients was ways to automate various investigative efforts in common data resources outside of the carrier, where to apply RPA to manual processes internally, and a set of best practices for working on a process in order to train and manage employees in those areas.

The granularity and breadth of data derived from process mining allows insurers to identify best practices and benchmarks for processes across different lines of business as well as compare one office to another, one person to another, or one unit to another. This sheds more light on best practices, where they are or aren't being followed, and where performance improvements need to be made.

These outcomes don't just present a snapshot of historical data. They can also continue to review claims in process and help flag and drive performance to the updated standards. They really do enable companies to operationalize findings and ensure they are adhered to moving forward.

We can all appreciate the effort and the challenge of digging into a process to look for wasted efforts, choke points, non-value adds, or root cause analysis. We all know that the challenges of limited time and resources can affect these efforts and result in a less than complete knowledge of the process assessment. Process mining can overcome these challenges and represents a significant change in how to prepare to transform claims operations.

Watch our series of intelligent automation in insurance videos for more insights into process mining.