This is the kind of post I wish existed when I was making technology decisions for eCommerce builds three years ago. The advice out there is either too abstract or too opinionated without context.
So, here's my actual thinking, with tradeoffs included.
Frontend
Next.js is the dominant choice right now for a reason. Server-side rendering, static generation, image optimization built-in, strong ecosystem. For most eCommerce frontends, it's the right default.
If you need something lighter and are comfortable with a slightly smaller ecosystem, Astro is genuinely excellent for content-heavy store pages.
Avoid building a heavy single-page application (React or Vue without SSR) for a public eCommerce store. Crawlability and first-load performance both suffer.
Backend / Commerce Engine
Shopify + Storefront API if your client wants a managed solution with minimal ops overhead. The headless capability is solid and the ecosystem is unmatched.
Medusa.js if you want a truly open-source, self-hosted commerce backend with real flexibility. It's younger but it's mature enough for production now. I've used it on two projects and it held up.
Magento/Adobe Commerce for enterprise complexity, large catalogs, complex B2B pricing, multi-store setups. But be honest with clients about the total cost of ownership. It's not cheap to run or maintain.
Database
PostgreSQL for almost everything transactional. It's boring advice but it's correct. Don't reach for NoSQL in your commerce core unless you have a very specific reason.
Redis for session management, caching, cart state. Non-negotiable at any real scale.
Search
Algolia or Typesense (self-hosted, lower cost) for product search. Native database queries will not give you the faceted filtering, typo tolerance, and search-as-you-type experience that modern shoppers expect. This is not a place to cut corners.
Lessons Learned Working on Large eCommerce Builds
Over-engineering early is a real failure mode. I've seen teams spend three months on infrastructure for a store that had 500 products and 200 daily visitors. Build for where you are, with architecture that can grow, not architecture built for 10x traffic you don't have yet.
Database indexing on product and order tables gets forgotten until things slow down. Don't wait for the performance complaint to add them.
Image optimization is never done. It always comes back. Automate it or it will always be a problem.
API rate limits from third-party services (payment gateways, shipping, tax) bite you in load testing if you haven't planned for them. Always plan for them.
And finally, work with partners who understand the full picture. A great ecommerce development company isn't just writing code. They're making architectural decisions that affect your business two years from now.
Top comments (0)