📄 Migrating a legacy auto claims system for 700+ employees

Research

Conducted ethnographic research (interviews, fly on the wall observation, shadowing, contextual inquiry) for the claims, product, pricing and contract cancellation processes leading to enhanced insights for stakeholders.

First, I was given an overview of the target areas I'd be working on as well as any legacy research that was conducted prior to joining. These included interviews and surveys.

After sitting with all this new information, I made the decision that I still needed more relevant info from the core users who use the system in their day-to-day. My manager agreed and pitched the idea to leadership.

The client had the bandwidth and resources to make that happen, so leadership did the legwork to scheduled a shadowing with several key employees who know the system inside and out. 

6 key users volunteered their time for the testing.

Post shadowing insights included:

• Claims usually came in either a pdf file or CSV so there would need to be a way to upload and parse through those files in the new system.

• Reps would have to memorize multiple acronyms for important key ins for processes. For "TLSA" open a vin search or a name search. There hundreds of these acronyms that the rep had to memorize in order to do their job. User error was common.

• Bulk cancellations have a 45 day turnaround time or more depending on the number of cases.

• Reps kept getting old cancellation requests. This would waste everyone's time.

Design Decisions

1. Claims intake is a step-by-step process. You can't fulfill a step without completing all the mandatory parts of another, but you can go back in the process and change info through proper overrides.

   
   

As a rep, the new process should be able to clearly tell them where they are in the process.

Therefore, I added a path...

so each claim step can be clearly seen and tracked in order for the claim to be considered completed. 

2. Next: Submitting a cancellation quote needed to be done more efficiently, and halting fraud should start on the rep level before escalating

   
   

"Auto insurers lose at least $29 billion a year to auto insurance scams...auto claim fraud includes false vehicle driver reports, faked information about the miles driven annually, and lying to get coverage from the insurance while vehicles are garaged in locations where insurance rates are cheaper." -Verisk Analytics

Testing

• A/B testing revealed that users loved the path idea, with all 6 saying that paths would make their way of working faster and clearer. It can only help and not hinder. Based on time on task, this amount of time improved.

• A/B testing also revealed that 4/6 users liked the stop pay idea but needed more information on who gets permission to do that. In executive stakeholder discussions, they agreed that stopping fraud would be an "all hands on deck" process with ultimate decisions would escalate to senior managers to decide what happens to payments that were stopped.

This was just a taste of the entire process.

If you want to know more about tradeoffs, what didn't make it into production, and challenges, get in touch with me to know more!