Full-StackCRMProduct Build
Connectr: CRM Platform, Built from Zero
How we led the full-stack development of Connectr's CRM platform — from architecture decisions to production deployment — on a startup timeline and already scaling beyond initial projections.
On time
Delivered to schedule
6 weeks
MVP to production
2×
Projected volume, already
## The Brief
Connectr came to us with a product spec, a tight runway, and a problem shared by many early-stage startups: they needed a senior engineering team to build their core product, but didn't have the time or capital to hire one in-house.
Their CRM needed to handle complex relationship tracking, multi-step pipeline management, and a real-time activity feed — all within a clean, fast interface their sales team would actually want to use.
They gave us six weeks to MVP.
## Architecture Decisions
Our first deliverable wasn't code — it was an architecture decision record (ADR) covering every significant technical choice and the reasoning behind it.
**Frontend**: Next.js App Router with server components for data-heavy views and React client components for interactive elements. This gave us fast initial loads with full interactivity where needed.
**Backend**: Node.js API layer with tRPC for type-safe client-server communication — no REST endpoints to maintain, no type mismatches between frontend and backend.
**Database**: PostgreSQL with Prisma ORM. The CRM's relational data model (contacts, companies, deals, activities, pipeline stages) maps naturally to a relational schema. We modelled this carefully upfront to avoid expensive migrations later.
**Auth**: Clerk for authentication — handles SSO, magic links, session management, and MFA without custom implementation work. The saved time went into product features instead.
**Infrastructure**: Deployed on Vercel (frontend) and Railway (API + database). Both auto-scale, both have preview environments on every pull request, and neither required a DevOps specialist to configure.
Every decision in the ADR had three sections: what we chose, what we considered and rejected, and what the trade-offs were. Connectr's CTO reviewed and approved before a line of code was written.
## The Build
We ran two-week sprints with a fixed structure. Monday planning, Wednesday check-in, Friday demo. The Connectr team saw working software every two weeks, not a big reveal at the end.
**Sprint 1**: Data model, authentication, core contact and company views.
**Sprint 2**: Pipeline board (kanban-style), deal creation and tracking, activity logging.
**Sprint 3**: Search, filtering, bulk actions, email integration (sending via Resend, tracking via webhooks).
At the end of week six, we had a production-ready CRM with real data, real users, and real email flowing through it.
### What We Were Careful About
**Database performance**: CRM queries can get complex fast — filtering 10,000 contacts by multiple criteria while joining activity counts. We added indexes thoughtfully, profiled queries with `EXPLAIN ANALYSE`, and wrote the Prisma queries to avoid N+1 patterns from day one.
**Type safety end-to-end**: With tRPC, the TypeScript types for every API procedure are shared between the Next.js frontend and the Node backend. A breaking change in the API is a compile error on the frontend before it ever hits production.
**The activity feed**: Real-time felt like a nice-to-have initially, but Connectr's team quickly realised it was core to how their salespeople would actually use the product. We implemented it with Supabase Realtime subscriptions — a focused addition that didn't require re-architecting the backend.
## Handoff
At project close, Connectr received:
- Full codebase on GitHub with a documented README
- Architecture overview explaining every decision and how to extend the system
- Database migration guide for future schema changes
- A 90-minute handoff call walking through the codebase with their incoming backend hire
The codebase is documented well enough that a developer unfamiliar with the project can be productive within a day.
## What Happened Next
Within three months of launch, Connectr had grown their user base to twice the initial projected volume. The architecture held. The database performance held. Nothing required emergency re-engineering.
They moved onto a monthly retainer with us for ongoing feature development. We know the codebase better than anyone — it made sense to keep that continuity.
> "Ten Peaks led the development of our CRM platform from the ground up. Clean architecture, documented codebase, and it's already scaling beyond volume we didn't expect to hit this soon. We've already signed them for the next phase."
> — James Calloway, CEO, Connectr
Building a product from scratch?
We specialise in taking a spec and shipping production-quality software on startup timelines. Let's talk about what you're building.
Tell Us About Your Project