Skip to main content
Frameworks

React

Choose the application before the component library.

Your developer has answered “React” before the business has agreed what kind of application it is building.

React can be an excellent interface layer for a dashboard, a complex workflow or a product with many reusable states. It does not decide how pages are routed, where data is loaded, what runs on the server or how the application is deployed. Those choices sit around React and often cost more to change later, so the application architecture is what needs approving.

“React gives us a mature ecosystem” is only the first decision

The pitch is fair. React has a large community, a broad package ecosystem and component patterns that many frontend developers understand. It can be added to an existing project or used as the interface layer inside a larger framework.

That flexibility is useful for a product with many interactive states. A well-designed component can own its behaviour and be reused across routes without duplicating the implementation. The same knowledge also carries into React Native when the business later needs a mobile application.

Flexibility creates a long shortlist of choices. Router, data fetching, forms, state, styling and testing can each come from a different project. Ask which set the team intends to support together and why each dependency is needed.

Create React App is deprecated; inherited builds need a route out

Create React App was once the standard answer for starting a client-side React project. The React team deprecated it in 2025 and now recommends a framework for many new applications, or a current build tool such as Vite for projects that need a client-only setup.

An existing Create React App product does not stop working because of that notice. It does acquire a maintenance decision. The team should identify custom overrides, environment variables, service-worker behaviour, test configuration and deployment assumptions before proposing a migration.

Do not let “deprecated” become permission for an open-ended rewrite. A representative route and production build can prove the target setup before the whole codebase moves.

React does not choose routing, data loading or rendering

React’s current application guidance recommends starting new apps with a framework because routing, data fetching and code splitting work better when designed together. The listed frameworks can still produce client-side applications and static output; choosing one does not force every route onto a live server.

A focused authenticated dashboard may work well with React, a router and Vite. A public product that mixes content, accounts and server-rendered data may benefit from an integrated framework. Adding these decisions package by package can leave the company maintaining a private framework nobody intended to build.

Ask for a route map showing which pages are public, which require authentication and where each page gets its data. That reveals more about the proposed architecture than a component demo.

React state libraries can outlive the problem they solved

React includes local state, context and hooks. Larger applications may add a server-state cache, a global client store or a form library. Each can be useful, but installing all of them at the start creates overlapping places for data to live.

Separate server data from temporary interface state. A customer record fetched from an API has caching and invalidation rules. Whether a drawer is open belongs close to the component. Putting both in one global store makes ordinary changes travel through more code than they need.

Ask the team to explain where one representative piece of data lives, who changes it and how stale data is handled. If the answer crosses several libraries, the state model needs simplifying before the product grows around it.

React performance has to be measured in the product

React is not a performance guarantee. The customer still downloads JavaScript, and component structure determines how much work happens when state changes. Large tables, charts and editor screens can expose poor boundaries that a small demonstration never reaches.

React provides developer tools and a Profiler to show which components render and how long they take. Use them on the slow interaction before adding memoisation throughout the codebase. React’s own useMemo guidance says caching should remain a performance optimisation, not something the application requires to work.

Measure route payload, time to useful content and interaction latency on a representative device. A faster development build is a separate result from a faster customer experience.

React Server Components create a server patching duty

Server Components and Server Functions are supplied through supporting frameworks and bundlers. They can reduce client-side JavaScript and coordinate server data with the component tree, but they also move React concerns onto the server.

That changes the update risk. The React team published urgent Server Components security fixes across late 2025 and January 2026. Client-only React applications without the affected framework or bundler support were outside that issue.

The lesson is narrow: know whether the application uses these packages and who watches their advisories. “We use React” no longer describes one runtime model.

Questions to ask before approving React

  • “What kind of application are we building, and which routes need a server?” Settle the shape first.
  • “Are we using a React framework or assembling the application around Vite?” Name the owner of routing and data.
  • “If this is Create React App, which custom behaviour must survive the migration?” Inventory it.
  • “Where will server data and temporary UI state live?” Keep their lifecycles separate.
  • “Which dependencies are part of our supported foundation?” Avoid a package for every preference.
  • “What customer performance budget will we test?” Include a representative device and route.
  • “Do we use Server Components, and who owns security updates?” Make the server surface visible.

When React is a sound product foundation

React suits an application with substantial interaction, reusable interface states and a team that can own the surrounding architecture. It is a strong choice for operational dashboards, collaborative workflows and data-rich products where component boundaries reduce repeated work.

The case is also good when the organisation already maintains React components, TypeScript conventions and tests. Reusing that capability is worth more than chasing a smaller framework on paper.

A current React framework can be the right route for a new public application when the team needs integrated routing and rendering. For a focused client-side tool, React with Vite can remain the smaller commitment.

When React adds more application than the requirement needs

Keep server-rendered HTML when the page has limited interactivity and already loads quickly. A targeted component can enhance one part without turning every route into a React application.

Do not introduce React into an organisation with a healthy, maintained frontend stack unless the product need pays for a second set of skills and dependencies. Framework familiarity across the team usually matters more than a generic popularity ranking.

And pause any migration whose success is defined as “the app builds.” The route architecture, production behaviour, accessibility and performance need acceptance criteria. Otherwise the business has funded a tooling change without proving the product is better supported.

Alexander De Sousa, Founder of Digital Royalty
Developer’s verdict

Choose the application architecture before the React component tree.

I use React for interfaces with enough state and reuse to justify an application layer. I will not approve it as a substitute for deciding routing, data loading and rendering. If a few enhanced server pages solve the problem, that is the smaller product and usually the better spend.

Alexander De Sousa · Founder, Digital Royalty · LinkedIn

Decision evidence

What turns React from a popular library into an owned application

React covers the component tree. The decision needs to account for the framework, shipped JavaScript, state model and security surface around it.

The React commitment in six decisions

Role in the stack
A user-interface library. Routing, data loading, server rendering, deployment and much of the state architecture come from surrounding choices.
Customer effect
Rich, responsive workflows when component boundaries are sound. Customers can also inherit excess JavaScript and slow interactions when they are not.
Adoption cost
Moderate for a new application foundation. Migrations vary with build overrides, state libraries, routes, tests and rendering changes.
Ongoing owner
A frontend team that owns React, the chosen framework or router, package updates, performance budgets and accessibility as one system.
Exit cost
Moderate to high. API contracts and domain logic can travel, while the component tree, hooks, state and framework conventions need replacement.
Proof required
A representative route measured for payload and interaction speed, plus a clean production build and an explicit dependency ownership list.

Application routes to consider before defaulting to React

  • Server-rendered HTML with targeted components

    Right when: Most routes already work well on the server and only a few workflows need a richer client-side interface.

    Watch for: Give embedded components clear data and lifecycle boundaries so they do not grow into an undocumented application shell.

  • React with Vite

    Right when: A focused client-side application can own routing and data without the broader conventions of a server-capable React framework.

    Watch for: The team must supply and maintain those missing application decisions itself.

  • A recommended React framework

    Right when: Routing, data fetching, code splitting and route-level rendering need an integrated production model.

    Watch for: Framework updates, hosting behaviour and server features become part of the product's maintenance surface.

  • Keep the organisation's current frontend stack

    Right when: The existing team, tests and component system already solve the product need without adding a second frontend capability.

    Watch for: Check that the decision reflects maintained expertise rather than an old stack nobody has budgeted to improve.

Research behind this React position

Get started

Tell us what you need

A few quick questions, then a straight answer from a real person — usually within a few hours.

What happens next

  1. 1 We reply A real person reads your answers and gets back to you, usually within a few hours.
  2. 2 Questions or a short call If anything isn't clear, we ask by email or book a short call. Before any call we audit your website and make the quick fixes it finds, free.
  3. 3 Your fixed quote A clear, fixed price in writing, within 48 hours of any call. There is no obligation to go ahead.

Tell us what you're working on

Whether it's a new site, a platform, or a process that shouldn't be manual any more — we'll tell you honestly if we can help.