Case study · iSeatz
From a legacy SPA to a multi-brand Next.js platform
Moving enterprise travel booking and checkout from a legacy React single-page app to a shared Next.js platform that several brands run on.
Next.js
TypeScript
React
Nx
Vercel
Contentful
Playwright
Problem
Enterprise clients needed the same booking and checkout flows, each with its own brand, content and configuration. The legacy React single-page app made every new client and every content change an engineering task, and it shipped more JavaScript than the first screen needed.
Context
- iSeatz builds travel booking and checkout for enterprise loyalty and travel brands. Checkout spans flights, hotels and cars, with several payment models and feature-flag variants per line of business.
- Work was shared across distributed product, design, backend, QA and engineering teams.
Constraints
- Several brands had to share one codebase while differing in content, locale and configuration.
- Checkout handles payments and personal data, so security and data handling could not regress.
- Checkout performance mattered for users, so the new platform had to load faster than the one it replaced.
Technical approach
- A Next.js, TypeScript and SWR foundation in an Nx monorepo, with shared libraries and per-brand packages.
- Content moved to Contentful and multi-locale i18n, with LaunchDarkly feature flags, so product owners could change content and behavior without a release.
- Storybook with Chromatic checks on every pull request, plus Playwright end-to-end and accessibility tests per brand package.
Architectural decisions
- One monorepo with shared libraries instead of a fork per client, so fixes and features reach every brand at once.
- Nx affected builds and distributed caching, with deploy locks and quality gates in the pipeline, so a large monorepo stays fast to review and safe to ship.
My contribution
- Led the architectural evolution and designed the platform foundation.
- Delivered the white-label capabilities (CMS, i18n, feature flags, payment and security flows) and the CI/CD and quality gates around them.
Outcome
- LCP improved by 50% and new Checkout calls dropped by 30%.
- Up to 90% of unchanged builds skipped and review builds 40–70% faster.
- Product owners update content themselves, and engineer onboarding is 2 days shorter.
Written from my own work. Proprietary code, internal system names and confidential business details are left out.