Headless vs composable vs staying themed

Three words get used interchangeably in this market, and they describe projects that differ by an order of magnitude in cost and vendor count. Sort them first.

• 01
  Headless means decoupling the storefront from the commerce engine, keeping the engine. Shopify Plus stays as the backend; the frontend is rebuilt in Hydrogen or Next.js against the Storefront API. Adobe Commerce stays; the frontend moves to PWA Studio or Next.js. This is what most 2026 'headless' projects actually are, and it is the only version most mid-market brands should consider.
• 02
  Composable (MACH) means swapping the engine for best-of-breed services too. commercetools for commerce, Algolia for search, Sanity for content, Stripe for payments: four vendor contracts where there was one. Forrester puts full MACH composable TCO at $2.4M–$4.8M over three years for mid-market retailers, roughly 3–5x a decoupled frontend on an existing platform. Of the 67% of enterprises on decoupled architecture, only about 19% run full composable builds.
• 03
  The mid-market rule from agency practice: headless against a single platform delivers about 80% of composable's benefits at 30% of the complexity. For $5M–$50M GMV brands, composable is almost never the answer; the question is headless-on-Plus versus a well-built theme.
• 04
  Staying themed is the correct answer more often than the market admits. The 75%-talked-out-of-it figure exists because Shopify's theme layer closed most of the historical gap: Liquid stores already post strong Core Web Vitals baselines, Checkout Extensibility handles most customization on Plus, and the delta headless adds is smaller than it was on a Luma-era Magento store. Read the ECM Shopify Plus vs Shopify guide before deciding a theme has hit its ceiling.
• 05
  Spring 2026 changed the framework math. Hydrogen was rebuilt to run on any stack including Next.js, and Next.js has 5–10x more available developers than Hydrogen or PWA Studio. New Adobe Commerce headless builds now split roughly 60% Next.js with custom GraphQL adapters, 40% PWA Studio. Talent availability is now a first-order input into the framework decision, not a footnote.

What headless actually costs, including what proposals skip

Most headless pricing quotes the build and omits the rest. The full cost structure has four lines, and the last three are where budgets die.

The build

• →
  Mid-market headless with one commerce engine and a custom frontend: $50K–$150K. Enterprise with multiple regions, channels, and high SKU volume: $150K–$500K+. A Hydrogen rebuild on an existing Shopify Plus backend runs $150K–$250K over 16–24 weeks; a comparable themed Plus build is around $100K.
• →
  The multiple is the honest headline: headless costs 1.5–2.5x themed on the same platform and adds 8–16 weeks. Adobe Commerce themed at $150K becomes PWA Studio at $250K–$400K. If a proposal's headless number is close to the themed number, something was left out.
• →
  Timeline claims need translation. Launch-focused Hydrogen builds can close in 6–8 weeks and full platform-plus-headless migrations inside three months in the best agency data, but the industry's own warning is blunt: '4–6 weeks' usually means 4–6 months for anything beyond a demo. Apply the standard 25–40% contingency and a skeptical read of any timeline under a quarter.

The lines proposals leave out

• →
  The Plus subscription. Most production headless builds land on Shopify Plus (~$2,300/mo on a multi-year term) for API rate limits, Checkout Extensibility, and Functions access. Hydrogen is free and open-source; the platform tier it needs isn't.
• →
  The headless CMS license. Decoupling the frontend means content needs a home: Contentful, Sanity, or equivalent, with its own recurring license and its own integration work. It is the most commonly omitted line in headless proposals.
• →
  Ongoing maintenance and the developer market. A theme is maintained by the platform; a custom frontend is maintained by you. Budget continuous developer capacity, and price it against a Hydrogen and PWA Studio talent pool 5–10x smaller than Next.js. Run the three-year view, not the build quote, in the ECM platform calculator before committing.
• →
  Search, personalization, and the app ecosystem. Platform apps that 'just work' on a theme often need re-integration or replacement on a custom frontend. Inventory every app-dependent feature in a tech stack audit before scoping; each one is either an API integration or a rebuild.

What headless delivers when the use case is real

Strip the vendor averages and the production data is still good — for the right brand. The mechanism is almost always speed, and speed's effect on revenue is measurable.

• 01
  Page load drops 20–50% versus monolithic baselines, with mobile seeing the largest gains. Google's own data ties a one-second LCP improvement to a 2–7% conversion increase, which is the mechanism behind nearly every credible headless ROI claim.
• 02
  The honest conversion range is 6–18% from reviewed production migrations. Worth pursuing, and worth modeling against your own traffic: a 10% lift on a $10M store is $1M, which clears a $200K build in a quarter. The same lift on a $2M store takes longer to pay back than the build takes to ship.
• 03
  Organic traffic follows Core Web Vitals. Digital Commerce 360's 2026 SEO research shows sites moving from CWV 'poor' to 'good' see roughly 22% organic growth over six months; frontend migrations on slow legacy stacks (Luma-era Magento to Hyvä) post 15–40%. The Shopify-to-Hydrogen delta is smaller because Liquid baselines are already decent — which is the whole point about who benefits.
• 04
  Named production results: Denham's move from Salesforce Commerce Cloud to headless Shopify made the site 3x faster. Veloretti's headless storefront holds an 80+ Lighthouse score while running a configurator with 100+ bicycle combinations across four languages — the kind of interactive complexity a theme genuinely struggles with. Both are Ask Phill builds; both had a use case a theme couldn't serve.
• 05
  The agentic-commerce argument is emerging, and vendor-flavored. API-first storefronts with structured data are positioned as the foundation for AI shopping agents, with projections of 15–25% agent-mediated revenue by 2027. Treat the projection as a scenario, not a forecast, but the underlying point holds: clean, structured, fast product data is increasingly what gets cited. ECM tracks this in weekly Insights.

Who should go headless, and who should stay on a theme

Assembled from agencies that build both and turn away most headless requests. Score yourself before the discovery call.

Headless earns its cost when

• →
  Your storefront has an interactive requirement a theme can't serve — a configurator, a builder, a marketplace layer, a content-heavy experience with commerce embedded — and the requirement is worth more than the 1.5–2.5x cost multiple.
• →
  You're on a slow legacy frontend (Luma-era Magento, an aging SFCC build) where Core Web Vitals are actively costing organic traffic. If the backend is also aging, the ECM Magento to Shopify guide covers replatforming the whole stack instead. The 15–40% organic recovery from a frontend migration is a fix for a problem Google is penalizing you for, not an aesthetic upgrade.
• →
  Multi-region, multi-brand, or multi-channel at real scale, where one frontend layer serving several commerce contexts beats maintaining several themes, and where an engineering team exists to own it.

Stay themed when

• →
  The motivation is design freedom or 'modern architecture' rather than a specific requirement. This is the 75% that a good agency redirects, and the six-figure lesson otherwise.
• →
  Your team can't fund continuous frontend engineering. A theme's maintenance is the platform's job; a custom frontend's is yours, forever, in a talent market that's tight for Hydrogen and PWA Studio.
• →
  You're under roughly $5M–$10M in revenue and your current Lighthouse scores are already in the green. The 6–18% lift, applied to your revenue, pays back slower than the build ships. Fix the checkout funnel with a CRO specialist first; it's cheaper and often larger.

The strangler-fig sequence that de-risks the build

Big-bang headless cutovers are how six-figure projects turn into twelve-month recoveries. The pattern successful implementations share is incremental extraction from the monolith.

• 01
  Phase 1 (months 1–3): migrate the high-traffic pages first. Product detail and product listing pages move to the headless frontend while the platform theme keeps serving everything else. Lowest risk, highest visible impact, and the phase that produces the first measurable speed and conversion data.
• 02
  Phase 2 (months 3–6): personalization and optimization. The features that justified headless — the configurator, the content-commerce blend, the segmented experiences — get built on the pages already migrated, against the baseline Phase 1 established.
• 03
  Phase 3 (months 6–9): checkout and account pages. On Shopify, checkout stays platform-native through Checkout Extensibility regardless; the account and post-purchase surfaces move last, because they're where authentication and payment edge cases live.
• 04
  Phase 4 (months 9–12): decommission the legacy frontend — and never before the new one is validated at 100%. Feature flags that switch the new frontend on by user segment turn every phase into a controlled experiment with a rollback path.
• 05
  Non-negotiables at every phase: server-side rendering (both Hydrogen and Next.js support it; client-only rendering is an SEO failure waiting to index), a complete redirect map wherever URL structures change, schema rendered server-side from day one, and the pre-migration baseline pulled from your core KPIs. The replatforming SEO discipline applies in full — the ECM SEO impact guide covers it — because a fast frontend that Google can't render is slower than the theme it replaced.