How I Built a Test Infrastructure from Scratch (And Why It Saved Our Project)

Belvinard Pouadjeu — Sun, 19 Apr 2026 10:49:05 +0000

No tests. No confidence. Every push was a risk.

That was the reality of our Internet Banking frontend project. I was working on the Back-Office Client—the mission-critical interface that banking staff use daily to manage customer accounts, transactions, and sensitive operations.

The codebase was growing and new features were flying in, but we had one critical problem: We had zero tests.

The High Cost of "No Tests"

This approach wasn't just a technical debt; it was costing us real time and money:

The "Release Tax": We lost half a day every single release because someone had to manually click through every feature to ensure nothing broke.
The "Refactor Fear": We wanted to improve the code, but every change felt like pulling a brick out of a Jenga tower.
The "Hidden Bug" Loop: A small fix in one module would silently break a transaction tool in another. We wouldn't find out for days—or worse, a user would find it first.

The team was slowing down. Not because we lacked skills, but because we had no safety net.

The Solution: A Modern Test Infrastructure

I decided to build a testing system from scratch designed for speed and reliability. Here is the technical stack I implemented:

Vitest + Testing Library + MSW + GitLab Pipeline

Automated Enforcement (GitLab Pipeline)

Every time a developer pushes code, the pipeline triggers. If a test fails, the merge is blocked. We physically cannot push broken code to production anymore.

Realistic User Simulation (Testing Library)

We don’t just test code; we test behavior. The tests click buttons, fill forms, and navigate exactly like a banking officer would.

API Mocking (MSW)

We use Mock Service Worker to simulate backend responses. This means tests run in isolation, are incredibly fast, and don't depend on an external server being "up."

Developer Velocity

By using Vitest, I ensured the suite stays lightweight. Any developer on the team can now write a new test in minutes.

The Results: From Gamble to Guarantee

Zero Bugs in Production: Every critical path is validated before a single line of code is merged.
60-Second Validation: What used to take 4 hours of manual clicking now runs in 60 seconds automatically.
Confidence to Refactor: We can now clean up old code with the certainty that our safety net will catch us if we slip.
Ready to Scale: The infrastructure is built to handle the next 100 features without falling apart.

The Bottom Line

The best time to add quality checks is before you need them. We didn't add tests because our project was big; we added tests so it could grow.

For developers: Does your project have a safety net, or are you just hoping for the best?
For leads/clients: Would you trust a banking system where changes aren't automatically verified before going live?

Originally published on LinkedIn.

QualityAssurance #SoftwareEngineering #React #FinTech #Testing #CICD #GitLab

DEV Community: Belvinard Pouadjeu