top of page


A major health insurance company asked for a visual update to their internal product. Our research insights surprised even knowledgeable stakeholders, and our user-focused solution solved long-ignored problems.


1 month


UX Researcher, Interviewer, and Designer


The client, a Fortune 100 health insurance company, approached our company seeking an updated skin for their 20+-year-old green screen internal claims processing software. After learning that training on this program took 6 months, and many new processors quit after 13 months, the design team knew there was more to this problem than visuals. We embarked on a revealing research journey before designing the solution.

old computers


We interviewed the people closest to the process and the tool with varying levels of experience: Junior Processors, Senior Processors, and the Training Lead for this program. Client leadership measured success by minimizing the amount of time spent processing, rather than increasing accuracy. Some of the most troubling issues facing our users were:

  • All users identified authorizations, duplicates, and coordination of benefits claims as the most difficult

  • Processors had to leave the program frequently, toggle back and forth between programs, and handle up to 6 standard operating procedures per claim

  • Automated work had to be checked manually, even calculations that the program should have done correctly automatically

  • Much of knowledge transfer was trusted to the extensive, 6-month-long training, but was supplemented with non-standardized tribal knowledge and shortcuts

  • All types of edits and errors manifested differently, and required prior knowledge to understand how to address each one

  • System issues meant that data replicated incorrectly, the program only allowed one instance even when multiple were required, processors had to remember passwords for tens of programs, and incorrect automation logic created more work than it saved.

Current State User Journey whitelabeled.
simplified steps.png


As we analyzed our findings, some key areas of opportunity emerged:

  • Presenting context early and often

    • Surfacing all errors and system flags at once rather than missing one and risking ruining all the work that came before

  • Integrating and optimizing the checklist

    • Shown here: yellow is the first step, purple is the last, and the process was different for all claims, even of the same type. Our proposed flow is much simpler, and is the same for all claims

  • Automatically sorting, checking, and surfacing information

    • Reduce the need for shortcuts and leaving the program, and incorporate their functionality into the new solution


In our solution, we sought to resolve the most trying pains of the claims process, and to capitalize on the ingenuity the processors had already normalized.

  • Aligned tool with process checklist steps to minimize toggling

  • Consolidated interface to minimize leaving the tool

    • Integrated exterior checklist and programs in steps and sidebar to optimize performance and improve accuracy

  • Improved the recognition of errors by surfacing issues as soon as they occur, all in one place

    • Helps reinforce training, narrows knowledge gap

  • Better user experience, with a simple, linear flow that reduces errors overall

    • Friendly reminders to check missed steps

    • Ability to go back

    • Aligned with vertical workstation screens

  • Improved accuracy


"Honestly I'm shocked. I've been working with this program for 10 years and I didn't realize there could be another way. This would save us so much headache."

-Experienced Claims Processor


Recommended metrics to gauge the efficiency of our solution, with a focus on accuracy and correctness, are:

  • Number of claims with errors, Number of errors within all claims

  • Ratio of amount paid vs amount under or over paid

  • Training time

  • Claim processing time

  • Time to independent processing, to gauge processor confidence

Our technical team created some of the automations mentioned above. Incorporating those pieces, along with solving many of the identified pains that were out of scope for this iteration, are worth exploring in future work.

Image by Ashkan Forouzani
bottom of page