Validating the Trust Thesis Against Operator Response
| Property | Value |
|---|---|
| 📅 Project | OmniRisk · Financial Workflow Dashboard |
| 🔖 Phase | Test & Iterate |
| 🛠️ Method | Moderated remote sessions · Think-aloud · Post-task survey |
| 👥 Participants | 3 operators tested — hedge-fund analyst, fintech analyst, FX trader |
| 🗂️ Output | 4 findings held · 3 findings did not hold · 1 iteration shipped · 2 design responses documented for next build |
Each design decision was a hypothesis. Testing was where the hypotheses got tested. The Test & Iterate phase validated the three trust decisions against operator response in moderated sessions with three operators — a hedge-fund middle-office analyst, a fintech middle-office analyst, and an FX trader.
This was a focused validation pass rather than exhaustive usability research. The design had strong research grounding from Discover, sharpened intent from Define, and built decisions from Design. What testing needed to surface was whether the interaction layer worked as the trust thesis predicted — not whether the whole concept was viable. That methodological framing shaped what was tested and what was deferred.
Three operators participated. The head of IT's systems vantage shaped the research but the prototype serves the people doing fund administration, so he was not part of testing. The sessions surfaced four findings that held, three findings that did not hold, and one iteration that shipped within the project window. Two additional design responses to the remaining findings were documented for the next build.
Method: Moderated remote sessions with screen-share. Think-aloud protocol. Post-task survey on trust and confidence with the AI Recommendation modal.
Why validation rather than exhaustive testing: The design hypotheses were grounded in convergent research evidence from four participants representing the workflow's full perspective. Testing was structured to confirm interaction-layer assumptions before they became production risks, not to discover whether the design concept had merit. This is the appropriate methodology when research is strong and the design needs interaction-quality validation before shipping.
Scope acknowledged honestly: This testing pass was focused and timeboxed within the project window. A production-quality testing program would run longer sessions, include compliance officer and operator pairs, and validate behavior under live volatile market conditions. The findings here are real signal from a real validation pass — and they are the appropriate depth for a conceptual case study, with deeper testing flagged for production.
Four design decisions held up cleanly:
Sequential flow. All three participants navigated the Fetch FCM → Run Analysis → View Insights sequence without instruction. The disabled states read correctly on first contact — participants understood why the next button was disabled and what would activate it.
Highlighted rows as AI recommendations. All three participants identified highlighted rows in the Currency Trade Table as AI recommendations without any legend or explanation. In-table highlighting held as a Core feature (see Feature Set) — context proved sufficient signal.