Composable commerce, three years in
Composable wins on velocity and team scale. It loses to monoliths on weeks one through four, and most teams quit then. Here's how to survive it.
Composable commerce is a real architecture, not a marketing slide. After three years on commercetools and adjacent platforms, the pattern is consistent: weeks 1–4 feel slower than a Shopify monolith. Weeks 12+ feel five times faster.
The trap most teams fall into is trying to recreate Shopify's full-stack DX on top of a headless platform. Don't. The whole point is that you own the storefront layer, embrace it. Own your routing, your cart, your checkout flow. The platform owns the catalog, the orders, the entitlements.
What I'd do differently: invest harder in the contract layer. Codegen everything from the GraphQL schema. Treat the API client as a first-class internal package, not a folder. The teams that did this shipped 3x faster a year in; the teams that didn't paid the tax in every PR.
The other thing: don't underestimate the content layer. Amplience, Contentful, Builder, pick one early, model it properly, and stop trying to ship marketing pages out of code. Every storefront I've seen regress to a monolith did it because nobody owned content.