What we walked into.
The execution platform was technically excellent and operationally fragile. The desk relied on tribal knowledge: which script to run, which sheet to update, what to check before market open. New traders took six weeks to onboard. A single mistyped command could halt strategies.
The team had tried twice to build an internal console and both attempts stalled. The second one shipped but went unused. The product question was harder than the engineering question.
How we worked.
We treated the desk as the user research and started from their day: market open, mid-session monitoring, strategy adjustments, end-of-day reconciliation. The console mirrored that rhythm rather than the underlying system architecture.
On the engineering side, we built it as a real internal product: versioned, observable, auditable. Not a one-off tool. Every action through the console produced a structured event the risk team could replay.
The work, broken down.
Desk workflow study
Three weeks observing the trading desk. Catalogued 80 distinct operator actions and the pre/post-conditions for each.
Console UX
A dense, keyboard-first interface organised around the four phases of the trading day. Every action two keystrokes from anywhere.
Build
React + WebSocket front-end, Go control plane, Postgres event log. Sub-50ms p99 action latency.
Audit and observability
Every console action emitted a structured event for the risk and compliance teams. Replay tooling for incident review.
What shipped.
- Workflow study
- Action catalogue
- Risk & audit model
- Console UX
- Keyboard-first IA
- Component library
- Onboarding flow
- React console
- Go control plane
- Event log
- Replay tooling
- Desk rollout
- Runbook
- Onboarding curriculum
- Handover to platform team