What we walked into.
Decisions were slow not because data was missing, but because it lived in five places. Each analyst maintained their own informal pipeline of CSV exports and tab-bookmarks. The cost was not just time; it was inconsistency: two analysts looking at the same opportunity would price it differently, depending on which view they checked first.
Compounding this, the existing tools had been built incrementally over five years by three different vendors. The IA was an archaeology of business priorities, not a product.
How we worked.
We started with a four-week discovery: shadowing analysts, mapping the decision workflow from idea to execution, and identifying the seven data primitives every screen actually used. From there, we designed a single workspace organised around the analyst's task, not the underlying systems.
The build was deliberately incremental. We shipped a minimal decision shell in week six, then layered in real-time data, charting, and OMS integration over the following two months, replacing the old tools one workflow at a time, so adoption never blocked production.
The work, broken down.
Workflow audit and analyst shadowing
Two weeks embedded with three desks. Mapped 41 distinct decision pathways down to seven recurring patterns.
IA and product strategy
Defined the workspace model, primary navigation, and a phased migration plan from the five legacy tools.
Design system and component library
A dense, data-first system built on a 4px grid with token-based theming and a 60-component React library.
Production build and rollout
Next.js + tRPC stack, integrated with the existing OMS and market data feeds. Rolled out desk-by-desk with no production downtime.
What shipped.
- Workflow audit
- IA & product roadmap
- Migration plan
- Design system
- Component library
- Workspace UX
- 60+ screen flows
- Next.js workspace
- tRPC API layer
- OMS integration
- Auth & RBAC
- Phased rollout
- Internal training
- Documentation site
- Handoff to in-house team