CMS Rebuild in React & Next.js

In 2023, I proposed and led this migration, deciding to introduce micro-frontends (iframe) and the Next.js App Router architecture, guiding a team of four members in the refactoring of the LINE Shopping CMS from PHP to a React & NextJS app router architecture. In addition to actual development tasks, I was responsible for tech-stack decision-making, internal training, schedule planning, task allocation, and risk assessment. This project was also one of the early initiatives within LINE Taiwan to adopt the NextJS app router.

How the Change Was Driven (persuading the engineering team to adopt a new solution)

Situation: The legacy system used a PHP backend with a server-side template language. The company had only one PHP backend engineer, so every update to the system hit major obstacles, and the frontend was hard to get right.

Goal: Split the whole system into a decoupled frontend and backend, and adopt the Next.js App Router, still very new at the time, as the frontend framework; future upgrades would be owned by a dedicated frontend team, and the PHP engineer would only maintain a clean backend.

Resistance: The migration could not ship in a single release; my manager required zero downtime and disaster recovery capability. Most frontend teammates specialized in Vue and were somewhat reluctant to adopt Next.js.

Action: Hosted several Next.js sharing sessions, demonstrating the advantages of the App Router step by step; designed an iframe-based micro-frontend architecture with an online feature toggle so that a problematic new page could be switched back to the legacy page immediately, without redeploying or shipping a fix.

Result: The plan resolved my manager's concerns and earned their support, and once teammates got to know Next.js they were willing to help deliver it. Using extra time that did not affect main product development, we migrated the legacy system to the new architecture in batches over about 1 year. The PHP frontend code was then retired, new features could be built with the usual frontend/backend split, average development time dropped by 30%, and PHP maintenance went from understaffed to adequately staffed.