FXpress · AI Co-Pilot for Cross-Border Remittance · Usability Test Plan
| Property | Value |
|---|---|
| 🏷️ Status | Complete |
| 🔬 Method | Unmoderated prototype task test (Maze) · formative · time-boxed |
| 👥 Participants | 5 real participants — Janice · Jerry · Gabriela · Jin · Ollie |
| 🧪 Artifact Tested | Interactive prototype of the core interactions |
| ➡️ Feeds Into | Usability Debrief · Design Iterations |
Every task in this study tested an interface operating under an explicit guardrail. Fexi does not forecast. This is a governance rule, not a copy preference — it exists for regulatory defensibility, auditability, and honest user expectations.
Prohibited terminology — systematically excluded from UI copy and documentation:
"predictions" · "rate forecasts" · "guaranteed trend" · "target rate predictions"
Approved phrasing — objective, descriptive, historical:
"Fexi describes…" · "rate timing analysis" · "historical trend observation" · "market depth overview"
One finding validates the policy directly: Gabriela responded positively to the non-predictive copy, unprompted. The constraint is not a tax on the experience — it is part of why users trust it.
This was a focused, formative usability pass — five participants, a defined task set, run on an interactive prototype before the functional build. The goal was not a statistically powered benchmark; with five participants, percentages would imply a precision the sample cannot support. The goal was to surface the interaction problems that block comprehension and trust — the kind that show up clearly with a handful of representative users — early enough to fix before building. Results are therefore reported as counts and described behaviorally, not as benchmark scores.
The study was designed to answer four questions that design alone could not: