Solar Engineering B2B SaaS Internal Tool Replacement 2024

Solar Letters

Replacing tools people rely on is a different design problem than building new ones. Replacing Excel (a tool engineers had customized over years and had real emotional ownership of) required understanding their workflow before challenging it, and earning trust before proposing change.

Role: Product Designer (UX/UI), sole designer · ~8 months · In production

20–30

New Orders Per Day

5

User Roles

3

Legacy Systems Replaced

~8mo

Kickoff to Launch

The Design Problem

When a solar installer puts panels on a building, somebody has to certify that the roof can hold them. Castillo Engineering was running that entire pipeline on Excel, Salesforce, and a creaking internal system. My challenge wasn't just building a better tool, and designing something five types of users with very different mental models would trust enough to abandon their existing workflows for.

Image placeholder Overview: engineer calculation workspace
Understanding the Domain

How a Solar Letter gets made

Inputs

  • Building codes (AHJ)
  • ATC/ASCE hazard data
  • Customer project details
  • Site photos + drawings

Calculation

  • Roof geometry + loads
  • Structural analysis
  • Engineer review + QA

Output

  • Certified Solar Letter
  • Building code citations
  • E-seal applied

Delivered to AHJ

Five Roles, One Journey

My IA decision: follow the artifact through the organization, not the org chart. Roles that handle the same artifact share UI; roles that perform genuinely different work get different surfaces. This kept navigation legible without building a separate product for each user type.

Installer

Submits project, uploads docs, tracks status

Jr. Engineer

Runs calculations, drafts letter

Sr. Engineer

Reviews, approves, e-seals

Manager

Assigns work, monitors pipeline

Support

Handles installer questions

Image placeholder Order intake: installer submission view
Image placeholder Review workspace: engineer calculation surface
~8-Month Timeline
Phase 0.0 Blueprint Discovery, flowcharts, wireframes, hi-fi designs, prototyping, design system. Highest design leverage.
Phase 0.1 Foundation Authentication flows, permission UI, branding integration, admin scaffolding
Phase 0.2 Projects Building Codes management, order intake, engineer review workspace, customer project list
Phase 0.3 Audits Document views, reporting surfaces, audit trails
Phase 0.5 Automation Automated calculations, letter generation, e-seal, RFI Flow, Project Changelog, workload views. Launch.

I designed the next phase while engineering implemented the previous one. Overlap was the only way to hit the timeline without losing coherence.

Key Decisions

Replace Excel by removing options, not adding them

The instinct when replacing a spreadsheet is to support every edge case it supported. I made the opposite call: the calculation surface only supports what Solar Letters require, validates inline, auto-fills contextually, and surfaces relevant building codes without requiring lookup. Constraint, not flexibility, was the win. Multiple engineers said it was faster than Excel.

One workspace, two modes, not two separate surfaces

Junior and senior engineers work on the same artifact at different stages. Instead of building two separate surfaces, I designed one shared workspace with a role-aware mode toggle at the top. Review-only controls appear only for senior engineers. Less product to build, less to break, fewer context-switching costs for people who need to move quickly.

RFI as a first-class product object, not an email thread

When an engineer needs clarification mid-review, that conversation was happening over email, invisible to the system, untrackable, easy to lose. I designed the RFI Flow as an inline, threaded object tied to a specific field or section of the project, with visible open/closed state and an audit trail that stays inside the platform.

E-seal as a deliberate, weighted moment

The e-seal is the legal equivalent of an engineer's professional signature, and it has real liability implications. I designed it as an explicit, confirmation-gated action with a document preview before commitment and a clear audit entry on application. The most legally consequential click in the product gets the most considered UI. One-step would have been faster; it also would have been wrong.

Building code expiration: the system doesn't trust itself silently

Building codes change. I designed a verification expiration mechanism that flags codes for re-verification after a defined interval: a soft warning in the workspace, a hard one at letter generation. A design decision in service of the engineer's professional liability, not just system correctness.

Image placeholder E-seal: confirmation-gated document action
Image placeholder RFI flow: inline clarification thread
Challenges

Displacing something people own

Engineers had built their Excel sheets over years. They were customized, familiar, and there was real ownership attached to them. Displacing a tool people have made their own is fundamentally different from replacing something they hate. I dealt with this by making the transition as low-friction as possible: exporting familiar formats, keeping calculation patterns that mirrored Excel's logic where there was no reason to change them, and building the new features around workflows they already understood.

Domain precision without becoming an engineer

Solar engineering has specific calculations, unit systems, and regulatory requirements that I had to understand well enough to design meaningful constraints, without understanding them so superficially that I built something the engineers would dismiss. I handled this by treating every review session as a domain lesson and asking the engineers to explain not just what the system should do, but why those were the right constraints. That "why" usually unlocked better design than the spec alone would have.

Five roles, no single happy path

Five user roles with genuinely different mental models meant I couldn't optimize the IA for any one type without disadvantaging another. Every navigation decision required explicitly asking: "who loses here, and is that acceptable?" The artifact-following IA emerged as the solution to this; the Solar Letter is what every role has in common, so organizing the product around it gave everyone a shared mental model even when their individual tasks diverged.

Outcomes

20–30

New orders processed per day since launch

A throughput that wouldn't have been sustainable on the Excel + Salesforce stack it replaced. The product is also boring in the right way: engineers don't fight the calculation surface. It just works.

Engineering team

Stopped using Excel for Solar Letter calculations. No transition period; they switched and stayed.

Salesforce

Decommissioned. CRM functions now live alongside the engineering workflow they were always supporting.

New engineers

Onboard onto one platform instead of three. The onboarding cost went from learning three tools to learning one workflow.

All Projects Next: ZFDS
tap to unmute