Micro-Frontends: A Reality Check After 3 Failed Implementations

We implemented micro-frontends on three projects between 2022 and 2025. All three were consolidated back into single frontend applications. Here is the post-mortem.

Project one: a B2B SaaS dashboard split into four Module Federation micro-frontends for a four-developer team. It worked for three months. Then the shared component library became a bottleneck -- version mismatches caused visual inconsistencies across sections. State management across boundaries required a custom event bus that grew increasingly complex. Build times went from two minutes to twelve. After eight months, we consolidated into a single Next.js app. Migration took two weeks. Team velocity increased 40 percent.

Project two: an e-commerce platform using iframes so marketing could deploy landing pages independently. Performance overhead, fragile cross-frame communication, and responsive design nightmares killed it in four months. A single app with static marketing pages and client-rendered catalog achieved the same independence without the pain.

Project three: the canonical use case -- twelve developers across three teams, distinct product areas. It still failed. Inter-team coordination overhead was enormous. API contracts changed frequently. The UX suffered because each team made independent design decisions, creating an app that felt like three products stitched together.

When do micro-frontends actually work? Only when all of these are true: more than twenty frontend developers, genuinely independent product areas with minimal shared state, and a dedicated platform team maintaining the infrastructure full-time.

For everyone else: use a monorepo with strong module boundaries. Turborepo or nx, clearly defined packages, team-owned modules, single build and deploy. You get organizational ownership boundaries without the operational overhead of independent deployments.

The frontend community has a pattern adoption problem. Patterns designed for organizations with hundreds of developers get evangelized at conferences, adopted by teams of five, and cause unnecessary suffering. For 95 percent of teams, micro-frontends are unnecessary complexity masquerading as good architecture.