illimity Tech: Benchmark Monorepo Tools

Why monorepos?

At illimity Bank, our frontend stack grew to span multiple apps — customer-facing portals, internal dashboards, and shared design-system packages. A monorepo was the natural next step to share code without the overhead of separate npm releases.

Tooling landscape

We evaluated four main approaches:

• Nx — Opinionated, full-featured, built-in affected computation
• Turborepo — Fast pipeline caching, minimal config
• Lerna (v7+) — Package publishing focus, improved with Nx backend
• PNPM workspaces — Bare-minimum orchestration, maximum flexibility

Benchmark criteria

1. Build time with cold and warm cache
2. Affected project detection accuracy
3. Configuration overhead
4. CI/CD integration complexity
5. DX for onboarding new engineers

Results summary

Nx won on raw developer experience and affected detection. Turborepo won on simplicity and cold-start adoption cost. For a team already invested in React + TypeScript, Nx's generators and IDE plugin tipped the scale.

Takeaways

• Caching strategy matters more than the tool itself
• Affected detection quality directly impacts CI speed at scale
• Choose the tool your team will actually maintain, not the theoretically optimal one