Trellis Financial
A broker portal, with a third of the requested scope removed
Trellis asked for a broker portal with forty-one features. We built twenty-seven of them. The fourteen we argued out of scope in week three would have cost more than the rest of the build and served under 2% of applications.
- Client
- Trellis Financial
- Sector
- Financial services
- Run
- 5 months · Apr–Aug 2024
- Filed
- 2024-09-24
- Capability
- .NET development · Web development · Cybersecurity · Database solutions
- Stack
- ASP.NET Core · C# · Entity Framework Core · SQL Server 2022 · Entra ID · Azure App Service · React · OpenAPI

Impact
3.2 → 0.9 days
Median application turnaround
14
Requested features argued out of scope
0
High-severity findings at the security assessment
Challenge
Financial services
5 months · Apr–Aug 2024
Brokers submitted lending applications by email attachment, and an operations team re-keyed them into the underwriting system. Turnaround averaged 3.2 days, most of which was queueing rather than assessment. The original specification had been assembled by asking each internal team what it wanted, with no arbitration — it included a bespoke document editor, an in-portal messaging system and a configurable workflow designer, none of which had a named user.
Solution
We priced the specification honestly and presented the fourteen items we thought should be dropped, with the reasoning and the volume data behind each. Trellis dropped thirteen and kept one. What was built is deliberately unremarkable: identity through Entra ID with broker firms as tenants, a structured application form that validates against the same rules the underwriting engine uses, document upload with virus scanning, and a status view that removed most of the chasing calls. Authorisation was designed as a model up front rather than endpoint by endpoint, which is what the security assessment at the end had least to say about.
Plates

Plate 01The form validates against the same rules the underwriting engine applies. 
Plate 02Authorisation as a model, decided before any endpoint was written.
Testimony
They told us in week three that a third of the scope we had asked for should be dropped, and gave us the reasoning in writing. That conversation saved more money than the build cost.