Magento to Next.js — faster pages, lower hosting, better SEO

The headline

A Magento storefront for an industrial equipment distributor was carrying years of accumulated weight — slow pages, expensive hosting, brittle deployments, and a technical SEO setup that was getting harder to maintain. I rebuilt it as a Next.js site with the product data sourced from the existing ERP, dropped hosting cost to effectively zero, and rebuilt the SEO and AIO foundation along the way.

Context

• An aging Magento 2 install on a self-managed VPS — running, but expensive to keep current and slow to deploy
• A four-figure monthly hosting bill that was eating margin without delivering anything the catalogue actually needed
• Core Web Vitals failing on mobile: LCP and CLS both red, INP marginal
• Deployment friction: every meaningful change involved a 20-minute staging deploy, a manual smoke test, and a window when the team didn't dare touch the site
• Plugin sprawl: dozens of extensions, several abandoned by their maintainers, contributing more attack surface than functionality
• The catalogue was the constraint: large, niche, with long-tail SEO value worth preserving

Why rebuild rather than optimise

The honest calculus: the Magento estate was costing more to keep on life support than a focused rebuild would cost to deliver. The plugin sprawl was the structural blocker — optimising a site you can't deploy confidently is sunk cost. Pointing Next.js at the existing ERP as the source of truth turned out to be a structurally simpler architecture than fixing Magento in place, and removed an entire category of data sync bugs along the way.

Approach

1. Audit and URL inventory

• Full crawl of the existing site, every URL captured
• 301 map built before any code was written — preserving SEO equity was non-negotiable
• Top-traffic pages identified, page-by-page brief written for the rebuild

2. Architecture

• Next.js App Router, SSG for catalogue and category pages, ISR for product pages
• Product data sourced from the ERP, not duplicated in a CMS — single source of truth
• Image pipeline: next/image with the Vercel image optimisation pipeline
• Search: Algolia for catalogue search — cheap, fast, and one less moving part for the team to maintain

3. SEO and AIO foundations

• Single H1 per page, semantic heading hierarchy
• Structured data on every product (Product, Offer, AggregateRating where applicable)
• FAQ schema where the page warrants it
• Plain-English definition paragraph after H1 on every concept and category page — extractable by LLMs
• llms.txt at root, robots.txt permitting GPTBot, ClaudeBot, PerplexityBot, Google-Extended, OAI-SearchBot
• Sitemap, canonical tags, and OG cards generated server-side

4. Migration and cutover

• Staging deployed under a subdomain for parallel review
• 301 map applied at the edge before DNS flip, so the redirect surface was live before traffic moved
• DNS cutover during a low-traffic window
• Rollback plan: DNS could revert within minutes, and the previous host was kept warm for the first two weeks in case it was needed

Outcome

• LCP roughly halved on mobile, comfortably inside the Core Web Vitals threshold
• Page weight reduced by close to two-thirds — fewer JS bundles, smaller images, no plugin baggage
• Monthly hosting cost dropped from a four-figure VPS contract to effectively zero on Vercel's Pro plan
• Organic sessions roughly doubled in the six months after migration — a combination of restored crawl efficiency, faster pages, and the AIO foundation paying off as LLM surfaces started routing traffic
• Deployment frequency went from cautious-monthly to confident-daily; the team can now ship copy changes without an engineer on call

What we learned

• The 301 map is the work. The rebuild is almost easy by comparison.
• Sourcing product data directly from the ERP eliminated an entire category of sync bugs. The "no separate CMS for product data" call is one I'd make again on any catalogue rebuild.
• AIO foundations are the bit clients underestimate. Doing them in the rebuild costs almost nothing; bolting them on afterwards costs a lot. The traffic curve from AI surfaces six months later was the evidence that decided it.

Stack

• Framework: Next.js 14 App Router
• Hosting: Vercel
• Data source: Microsoft Dynamics 365 Business Central via its REST API
• Search: Algolia
• Analytics: Vercel Analytics (cookieless, no consent banner required)
• Images: next/image with the Vercel image optimisation pipeline
• CDN/edge: Vercel edge network

Related capability

This is the heart of the SEO + AIO + GEO service — performance, architecture, and answer-engine readiness as one programme rather than three. The ERP-as-source-of-truth pattern is the same one used in the multi-brand ERP consolidation.