A pharmacy claims review process spread across three disconnected applications. Auditors lost their place constantly, training took months, and throughput sat flat at thirty reviews a day. We collapsed it into two screens and took it through two rounds of moderated testing before a line of production code was written. Post-launch it runs at forty-five quality reviews a day.
Pharmacy claims get reviewed in a short window before payment. Catch a discrepancy in that window and it is corrected. Miss it and the pharmacy is paid, the audit happens later, and somebody spends months chasing the money back.
So the review window is the whole value of the process, and it was being spent on navigation. Completing one review meant working across three separate applications, moving out of the workflow and back into it repeatedly to gather data that should have been on the page.
What the users said. Four people, three auditors and a supervisor, walked through a realistic scenario the supervisor wrote. Even with every piece of data available and nothing blocking them, it took too long. Two comments did most of the work in the research summary: the interface made it easy to lose track of what had been audited and where you left off, and the page carried a great deal of information nobody needed while missing things they did.
Both of these are the same tool at two stages. Axure RP, which is worth saying out loud, because the second video is only possible because of what the first one made cheap.
Why the tool matters here. Both artifacts came out of one environment, which meant the wireframe was not thrown away to start the prototype — it was raised. And because the prototype actually behaves, usability testing could run real scenarios against it rather than asking people to imagine what a button would do.
Not every piece of UX was budgeted for the timeline, so the work was scoped to what would change the outcome and sequenced accordingly.
Watching the work, not asking about it.
The constraint was the design.
Against the prototype, before development.
Built against the client’s design system as it stood in 2022 — an older system than the one they run now, and the work was done within it rather than around it.
The application was built and launched. These are measured post-launch numbers, not projections.
Throughput rose by half. From roughly thirty reviews a day to forty-five, and the word the client used was quality reviews — the increase did not come at the cost of the work being done properly.
The per-claim time tells the clearer story. Twelve minutes down to seven. Across a working day that returns about two and a half hours to every auditor, which is the number that actually gets noticed in an operations meeting.
And two effects nobody asked for. Input errors fell, because there are fewer places to make them when data entry sits beside data display. And training time halved, because a new auditor learns one screen instead of the seams between three applications.
This redesign has exceeded our team’s expectations. There were some metrics around reduced clicks and improved workflow, but the ease at which the users can now complete tasks is very impressive. In addition, onboarding new trainees will be much easier with this new process.
This is what the engagement produces. The tooling has moved on since; the method has not.
We sit with the people doing the work and watch them attempt real tasks. The original flow here met every requirement in its specification and was still painful to use — those are different tests, and only one of them is in the document. You get the observed flow, the workarounds nobody wrote down, and a defect list ranked by what each one costs per day.
Two rounds of moderated testing against a prototype that responds like the product, not a clickable picture. Defects found at that stage cost an afternoon to fix. The same defects found after the build cost a sprint, and some of them never get fixed at all because by then the code is load-bearing.
We work inside the one you already have, as it exists today, not a nicer one invented for the pitch. A redesign that ignores the system in place produces screens your engineers cannot ship. If the system itself is the problem, that is a separate conversation and we will say so.