Definition of the micro frontend
In the article “Micro Frontends” (martinfowler.com, 19 June 2019, read on 12 November 2026), Cam Jackson defines the micro frontend as “an architectural style where independently deliverable frontend applications are composed into a greater whole”. The frontend is the part of a web application that the user sees and interacts with.
Michael Geers, on the micro-frontends.org website (read the same day), suggests thinking of a site as a composition of features, each in the hands of an independent team, and sees in it the idea of microservices extended to the interface.
The expected benefits of splitting
Cam Jackson attributes four benefits to splitting: incremental upgrades, smaller and decoupled codebases, independent deployment, autonomous teams. Each micro frontend has its own delivery pipeline, and teams form around slices of business functionality rather than technical skills.
Incremental upgrading ties in with the step-by-step replacement of a legacy application: the entry on the strangler fig pattern describes the pattern, our page on application modernisation how it is run.
The integration approaches reviewed in the article
In the article, a container application displays the common elements, handles authentication and navigation, and places each micro frontend in the page.
| Approach | Principle | Author’s remark |
|---|---|---|
| Server-side composition | The server assembles the page from HTML fragments | A technique that is nothing new |
| Build-time integration | Each micro frontend is a package, on which the container depends | Not recommended: any change requires the whole to be redelivered |
| Iframes | Each micro frontend is displayed in an isolated frame | Routing, history and deep links are more complicated |
| Run-time JavaScript | A script exposes an entry function, which the container calls | Probably the most flexible, in his view |
| Web Components | Each micro frontend defines a custom HTML element | A variant of the previous one |
Tools support run-time integration. The webpack documentation describes Module Federation: separate builds form a single application and load modules exposed by others. The single-spa documentation presents a framework that brings together several JavaScript micro frontends, each registered with a name, a loading function and an activation condition.
What splitting costs
The article identifies three drawbacks. Separately built JavaScript bundles may embed the same dependencies, which increases download size. A micro frontend developed outside its container may behave differently in production. Repositories, tools, delivery pipelines and servers multiply.
The article deals separately with two consistency issues. CSS styles, global by nature, may conflict from one team to another. A shared component library should carry only display logic.
Micro frontend, microservices, backends for frontends, headless: not to be confused
Microservices architecture splits the server side of an application. Michael Geers calls “frontend monolith” the single-piece interface placed on top of microservices.
The backends for frontends pattern, presented in the entry on the API gateway, belongs to the server side. Cam Jackson notes that it can extend to one backend per micro frontend.
The “headless” approach, covered in the entry on composable architecture, separates the interface from the services it calls, without saying how that interface is split.
An estate agency network’s website run by three teams
Imagine the website of a network of estate agencies, divided between three teams: property search, property details page, owners’ area. Each delivers its micro frontend. A container application carries the header, sign-in and navigation. The search team releases a new filter without waiting for the other two.
Two trade-offs appear. Without agreement on the version of the display library, the visitor may download it three times. A naming convention prevents one team’s style sheet from altering the screens of the others.
The architecture document described on our page on IT architecture is where such a decision is set out, with its options and consequences.