Web performance consultant for Medusa.js

Medusa.js, modular headless commerce with performance to frame

Medusa.js has established itself as the open-source alternative to Shopify for teams that want to own their e-commerce stack end-to-end. Version 2.0 rewrites the core into a modular architecture (Modules, Workflows, Links), with a strict backend / storefront separation — typically a Next.js storefront calling the Medusa API through the Store API or Server Actions.

That modularity is a design strength, but opens performance debt angles that compound as soon as the catalog passes a few thousand products or traffic climbs. The web performance optimization angle on Medusa is multi-layer: Postgres profiled on catalog and pricing queries, Redis configured for cache and events, workflows split between sync and async, Next.js storefront on ISR or streaming with a strict JS budget.

Pricing engine and catalog, first work-stream

Medusa's pricing engine resolves complex rules (price lists, currencies, customer groups, region, sales channel) on every request. On a product listing rendering 50 cards with prices, the calculation can trigger cascading queries if relations aren't preloaded correctly.

Diagnosis via the Medusa logger and Postgres profiler (pg_stat_statements, EXPLAIN ANALYZE) exposes slow queries immediately. Moving to targeted select + relations instead of full fetch, plus a Redis cache on stable price calculations, drops API TTFB 40 to 70% on catalog endpoints. On high-SKU stores, it's often the lever that gets the storefront green on LCP.

Workflows and events, async by default

Medusa 2.0 introduces the Workflow Engine that orchestrates multi-step operations (order creation, refund, payment capture) with idempotency and compensation. It's powerful, but a poorly-split workflow runs steps synchronously that should be async: confirmation emails, ERP sync, Meilisearch or Algolia index updates.

Refactoring — critical steps synchronous (payment capture, stock), secondary steps async via Subscribers + Redis events — drops perceived order creation latency by hundreds of ms. Same logic as Laravel queues, but carried by Medusa's workflow primitives.

Next.js storefront, the Core Web Vitals battleground

The storefront is typically separated from the backend (Next.js, Remix, Astro). That's where LCP, CLS and INP — what Google measures — actually play out. Classic traps: revalidate: 0 everywhere instead of ISR on catalog pages, full hydration instead of RSC + streaming, Stripe or PayPal SDK loaded on the home when it's only useful at checkout.

The storefront audit covers JS bundle (250 KB gzip target on the home), product images (AVIF format, correct sizes, lazy below the fold), CDN cache on product pages, and precise RSC/Client splitting. On a properly framed Next.js Medusa storefront, you hold sub-1.5s LCP and sub-200ms INP even on 3G mobile.