What is headless commerce? The meaning, the examples, and the honest trade What is headless commerce? The meaning, the examples, and the honest trade

Headless commerce means the storefront — what the shopper sees — is decoupled from the commerce engine that runs the catalog, cart, orders and checkout, and the two talk over an API. The “head” (the front end) is removed from the platform, hence the name: you build your own head, and the engine serves any head you point at it.

What you gain is total front-end freedom. What you take on is a front end that does not exist until an engineer builds it — and the most useful cost figure in this article comes from a vendor that sells headless. Both halves below, honestly.

Headless, explained in one architecture

A traditional platform — a themed Shopify store, a Webflow site — is monolithic: presentation and commerce live in one system, and you customise within its theme model. Headless splits them:

• The engine (catalog, cart, checkout, orders) runs as an API — Shopify's Storefront API, or a dedicated engine like Medusa, Saleor or commercetools.
• The head is a web app your team builds — usually React or a framework like Next.js or Shopify's Hydrogen — deployed to edge hosting, calling the engine for products, carts and orders.

Because the engine doesn't care what calls it, one API can serve a website, a mobile app, a kiosk and a marketplace feed at once. That's the architectural promise.

And “composable commerce”? The same idea, taken further: instead of one platform, the whole stack is assembled from best-of-breed components — commerce engine, CMS, search, payments — each API-first, each swappable. Headless describes the storefront split; composable describes building your entire stack that way. Every composable stack is headless; not every headless build is fully composable.

The platforms, with real prices — examples of what “headless” actually names

All read on the vendors' own surfaces, 26 July 2026:

Three details from that table that most explainer content gets wrong:

• Saleor's “Forever Free” Sandboxes tier sits under “Non-Commercial Use Only” — it is not a free plan for a store.
• Hydrogen doesn't require Shopify Plus. Oxygen hosting is free on any paid plan; the real gate is the storefront count — one below Plus, 25 on Plus (shopify.com/pricing, 26 Jul 2026).
• Going headless on Shopify does not mean losing Shopify's checkout. Hydrogen storefronts hand the shopper to Shopify's hosted checkout; keeping it is the default. Decoupling checkout itself is a separate, unpriced enterprise conversation.

What a brand genuinely gains

Headless is not a scam, and the engineers advocating it aren't wrong about the ceiling:

1. Total front-end control. No theme system, no Liquid, no template model. If the design team can specify it, the front end can be it.
2. Real performance headroom. Edge rendering, your own caching, none of the third-party script weight a theme accretes. The ceiling is genuinely higher than a themed store's.
3. Many surfaces, one engine. Web, app, kiosk, marketplace — one API underneath.
4. Ownership without lock-in. Medusa (MIT) and Saleor (BSD-3-Clause) can be self-hosted and run indefinitely; Medusa states plainly there's no lock-in and data exports any time.
5. Sane economics at the top end. commercetools charges by order, not GMV — their pitch: “no hidden fees or penalties for scaling.” For a high-AOV brand, order-based beats GMV-based comfortably.
6. On Shopify, it's cheap to try. Hydrogen is MIT, Oxygen is free on any paid plan.

What it costs — in Shopify's own words

The number that should anchor every headless evaluation comes from Shopify's own enterprise blog, selling headless (read 26 Jul 2026):

“The pricing for an enterprise headless project could cost anywhere from hundreds of thousands to millions upfront, plus annual maintenance costs.”

That's the vendor's figure, not a critic's. The same piece concedes the decision itself: “if your business is prospering with a traditional architecture, it may not be worth the financial and time resources to invest in headless.” Where that money goes:

1. The storefront doesn't exist yet. Cart drawer, variant picker, search, filtering, account pages — every component is a build.
2. Marketing loses the edit button. No theme customiser, no visual editor unless someone builds or buys one. Every banner, landing page and copy change enters a sprint. This is the single biggest operational cost, and it never appears in the licence comparison.
3. The app ecosystem mostly stops working. Theme-injected apps have no theme to inject into; reviews, upsells and loyalty each become an integration you own.
4. Checkout becomes a decision. On Shopify you keep theirs (good — say so). On Medusa or Saleor you build it, and inherit PCI scope, payments, tax and fraud.
5. You operate infrastructure. Medusa self-hosted wants Postgres, Redis, a server instance and a worker instance; someone is on call for that now.
6. Launches slow by a category. The realistic unit for a headless replatform is quarters.
7. You lose measurement by default. None of Medusa, Saleor or commercetools ships analytics, session replay, heatmaps, experimentation or error tracking. Their pricing pages meter orders, GMV, compute and requests — not insight. That whole stack gets assembled afterwards, by the same engineers, out of the same budget.

Who should actually go headless

An honest filter, not a strawman: headless earns its cost when you have a real engineering team that will own the storefront permanently, multiple surfaces (app, kiosk, international storefronts), and a design or performance ceiling you've genuinely hit — and when marketing can live with engineering owning the edit button, or you'll budget for a visual-editing layer on top. That's a real population. It's mostly not a 1-to-20-person DTC team, which is Shopify's own “may not be worth it” sentence in different words.

The opposite bet — and the version where you're already headless

Mtrix is the opposite bet on the same two promises. Headless buys speed and control by adding engineering; Mtrix removes it: build the store in a visual builder with no developer, edit live and undo in one click, publish to global edge hosting instantly. The speed the shopper feels is the point — the site's own published line is “6.3 seconds to 1.8,” and “Under 2 seconds on low-4G. Every page. Guaranteed.” One platform, one line item, and the measurement layer — analytics, replay, heatmaps, experiments, errors — is already inside. What headless promised as composability, and delivered as integration work, arrives assembled.

And if you're already committed to headless? Then don't replatform — instrument. Mtrix runs on the store you have via npm or CDN, and a headless stack is actually where its server-side testing is most at home: variant assignment is decided server-side on every test, and the server-rendered integration is documented for Hydrogen, Next.js and SvelteKit — about 35 lines in your load function. No experiment caps, a first test live in about five minutes once installed, and no Mtrix transaction fee metering your revenue (payment-processor fees only — a pointed contrast with GMV-overage pricing). To a team whose every experiment is currently a deploy, that's the missing half of the stack. When the storefront itself becomes the bottleneck, migration is a shipped feature.

FAQ