Skip to content
Dashboard

Headless commerce turns page speed into revenue

In Portent's 2022 study, business-to-consumer sites that loaded in 1 second had a 2.5 times higher conversion rate than sites that loaded in 5 seconds.

Storefront speed depends on how the page renders, what the browser downloads, and what it has to wait for. If an uncached page needs the result of a slow catalog query before it can render, shoppers wait too.

Headless commerce separates the storefront from the commerce backend. Frontend teams can improve page rendering and see whether those changes improve conversion without replacing the backend.

Copy link to headingWhat is headless commerce?

In headless commerce, APIs connect a separate storefront to the backend that manages commerce data and transactions.

The backend can be a platform like Shopify, BigCommerce, or commercetools, or an engine like Medusa or Saleor. It manages the catalog, pricing, inventory, and checkout.

The frontend runs as a separate application and controls what shoppers see and how they interact with the store. It uses GraphQL or REST APIs to exchange data with the backend.

Backend engineers keep data accurate and transactions running. Frontend engineers manage the storefront's layout, interactions, and rendering, and deploy on their own schedule. If the storefront fetches or revalidates pricing data, the backend team can change prices without a frontend deploy. Frontend teams can redesign routes without changing the commerce database.

Copy link to headingWhy coupled storefronts lose revenue

A coupled commerce platform packages the storefront and commerce engine together. The platform determines how pages render and how much control teams have over rendering and caching. A slow catalog query, a script that blocks rendering, or a delay in the backend can slow the store.

When the storefront and backend share a release pipeline, the frontend team may have to wait for a full-stack release and regression testing to ship a layout change or rendering improvement.

Proprietary templates can tie a storefront to a vendor's rendering system. Replacing the frontend doesn't always mean moving the backend. Shopify Hydrogen and BigCommerce Catalyst can both run on Vercel and connect to their own commerce APIs. A custom Next.js storefront gives teams control over rendering and hosting. If you switch commerce engines, you'll still need to adapt your catalog, cart, and checkout integrations.

Black Friday traffic can expose these limits. If product pages and checkout depend on the same backend service and it can't keep up, both can slow down. With either architecture, caching public catalog pages and load-testing checkout can help manage that risk.

Copy link to headingDecoupling turns page speed into conversion lift

With a separate storefront deployment, the frontend team chooses the framework, caching strategy, JavaScript budget, and infrastructure. The team controls how the storefront renders and can ship performance improvements on its own schedule.

Helly Hansen moved from an Adobe Commerce progressive web app to a Next.js storefront on Vercel. During Black Friday and Cyber Monday, it reported 80% year-over-year growth and a conversion rate twice that of the previous year. Load times fell from more than four seconds to between 1.8 and 2 seconds. Every Core Web Vital went from red to green. The team went from one A/B test a month to five or six per sprint.

"Conversion was so high because the site was so fast," said Anshika Jain, Helly Hansen's chief digital officer. "During Black Friday, we increased our performance marketing spend three times because it was so profitable."

The higher conversion rate let Helly Hansen spend more on paid marketing during the campaign.

In Rakuten 24's study, visitors with a faster Largest Contentful Paint (LCP) converted at rates up to 61.13% higher. In a separate month-long A/B test, Rakuten 24 improved several Web Vitals on a landing page. The updated page had a 33.13% higher conversion rate and 53.37% more revenue per visitor than the original. LCP measures how soon after navigation the largest qualifying image, text block, or video in view appears.

PAIGE reported a 76% higher conversion rate during Black Friday and Cyber Monday after moving to Shopify, Next.js, and Vercel. Ruggable reported 300% more unbranded organic clicks and 40% faster site speed after rebuilding its frontend while keeping Shopify.

Going headless doesn't guarantee a faster storefront. Too much client-side JavaScript, uncached API calls, or heavy reliance on client rendering can make a new storefront slower than the one it replaces. Separating the storefront gives teams control over rendering and caching. Measure how changes to either affect performance and conversion.

Copy link to headingCore Web Vitals are commerce metrics

Core Web Vitals describe how fast a page loads, how much it shifts, and how quickly it responds to shoppers.

  • Largest Contentful Paint (LCP) measures the time from the start of navigation until the largest qualifying image, text block, or video in view appears.

  • Cumulative Layout Shift (CLS) measures the largest burst of unexpected layout shifts during a visit. Late-loading content pushing a button out of place is one example of a shift.

  • Interaction to Next Paint (INP) measures a page's responsiveness based on its slowest qualifying interactions, such as clicks on filters, size selectors, and add-to-cart buttons.

Google uses Core Web Vitals in its ranking systems, but good scores alone don't guarantee higher rankings. Across the studies it reviewed, the Baymard Institute found an average cart abandonment rate of 70.22%. Track checkout completion alongside performance to see whether a faster, more stable storefront reduces drop-off.

Copy link to headingHow to maintain Core Web Vitals

Decide what to cache based on how often the data changes and how current it needs to be. You can often cache public product descriptions and category layouts for hours or days. Inventory and pricing may need shorter cache lifetimes, revalidation, or a fresh check when a request comes in. One cache policy for everything can leave shoppers with stale prices or make the storefront fetch data it already has. Choose a policy that fits what the shopper is doing.

JavaScript can take time to download and run. Analytics, personalization, A/B tests, and chat widgets can make a new storefront ship more JavaScript than the one it replaced. Render on the server where it helps, limit client-side JavaScript, and defer nonessential scripts. Measure INP on real devices to check whether key interactions stay responsive.

Keep measuring after launch. New features and third-party scripts can slow the storefront down. Vercel Speed Insights collects real-user performance data from preview and production deployments. Compare routes and releases once you have enough visits to make the comparison useful.

Copy link to headingIncremental migration preserves SEO and revenue

You can reduce migration risk by moving storefront routes in stages and keeping the commerce platform as the system of record. Start with a route where you can measure the results before rolling out to more pages.

Separate deployments can help teams ship storefront changes more often. Tray.ai cut its website build times from a full day to two minutes after updating its build infrastructure. During a storefront migration, measure how long it takes to build, review, and release a route change.

Focus on five things during the migration:

  • High-traffic route sequencing. Start with a route that gets enough traffic to measure results and that you can migrate with manageable risk. Helly Hansen started with checkout to test a critical flow without disrupting homepage marketing attribution. Product detail or collection pages can also work if you can test them in isolation.

  • URL structure and canonical parity. Keep existing URLs where possible and check each page's canonical tag. If a path changes, use a permanent redirect straight to the new URL. Check internal links, sitemap entries, and product structured data before switching traffic.

  • Session continuity across architectures. Shoppers may move between old and new routes during the migration. Use the commerce platform's supported cart and checkout flows. Test that shoppers stay signed in and keep their cart items as they move between routes.

  • Frontend API response caching. Two storefronts can put more load on the same backend. Track request volume during rollout and cache public catalog responses that don't need to be current on every request. Keep individual shoppers' cart and account responses out of shared caches.

  • Merchandiser workflow stability. Keep merchandisers and marketing teams in their current tools where possible. Use webhooks and APIs to pass updates to the new storefront. Check that both teams can still preview and publish changes.

Copy link to headingComposing the headless commerce stack on Vercel

Choose the Vercel feature that addresses your storefront's bottleneck, then measure the effect after you deploy.

Commerce problem

Vercel feature

How it works

Product catalog freshness without full rebuilds

Next.js Incremental Static Regeneration (ISR)

When product data changes, use on-demand revalidation to refresh the affected pages without rebuilding the site

Server-side rendering latency

Fluid compute

Choose a compute region near the commerce API. With Active CPU billing, you don't pay for CPU time spent waiting on I/O. You still pay for provisioned memory

Geographic personalization, A/B testing, locale routing

Routing Middleware

Apply locale or experiment rules with Routing Middleware before serving the route

Global page delivery

Vercel CDN

Serve cached storefront content from the Vercel CDN near shoppers

Durable order flows

Vercel Workflows

Use Vercel Workflows to save progress and retry failed steps, such as fulfillment or notifications. Make sure retries don't repeat side effects when writing to external systems

Shopping assistant responses

AI SDK

Stream tool results with the AI SDK and display them with storefront components using generative UI

Multi-provider model routing for AI features

AI Gateway

Use one endpoint for multiple providers. Configure model fallbacks and bring your own provider credentials if needed

Merchandiser and QA review without shared staging

Preview Deployments

Create preview URLs for eligible changes in connected Git repositories so teams can see how the changes look and work

Performance regression monitoring

Speed Insights

Track Core Web Vitals from real visits and investigate changes as data comes in

Rapid storefront prototyping

v0

Generate Next.js prototypes with v0, including shadcn/ui components and full-stack project starters

After traffic to Lift Foils doubled overnight, Commerce-UI used Routing Middleware to serve storefront pages tailored to each region. The agency also reviews changes in Preview Deployments before they go live.

Notion paired Statsig with Global Config to run experiments on its marketing sites and used Vercel's monitoring tools to track performance. Hotfix deployment time fell from one hour to 15 minutes, a 75% decrease. The sites averaged a CLS of 0.09 or better. Teams can review changes in previews before release and use production monitoring to spot issues afterward.

Copy link to headingHeadless storefronts prepare for agent-driven commerce

Shoppers browse and buy products through the storefront. Some also use AI assistants to find products, compare options, or start a purchase.

Shoppers using an AI assistant can compare options without opening a product page. The assistant can use supported commerce interfaces to look up product details, prices, and availability in a structured format. Page speed still matters when shoppers browse the store or follow a checkout link.

The same commerce backend that supplies product data to a React storefront can also supply it to a shopping agent. Teams still need to give the agent a way to find products, update the cart, and hand off to checkout.

A product page can render React components for shoppers and include structured product data for search engines. Agents can use separate tools. For example, Vercel Shop makes product search and cart tools discoverable through the Model Context Protocol (MCP). If you add an assistant to the storefront, you can use the AI SDK to show tool results with your existing design system components.

What you cache and where your code runs can affect response times for storefront pages and agent requests. Measure storefront performance and API latency separately to see what needs work.

Copy link to headingThe storefront you already built is the one agents will use

A headless migration can prepare your store for shopping agents. Agents can use the same product and cart APIs as the storefront if you expose the operations they need. Track browser metrics such as LCP for shoppers and API response times for agents.

On Vercel, Next.js can refresh catalog pages without a full rebuild using Incremental Static Regeneration. With Fluid compute, you can choose where server rendering runs relative to the commerce API. Routing Middleware can apply personalization rules before serving the page. Speed Insights reports Core Web Vitals from real visits.

Explore Vercel for retail to learn how to build and optimize a headless storefront.

Reference Architecture

Choose a stack and work through the decisions behind it, with curated documentation, guides, templates, and courses to help you build.

  • Agent stack: Build and operate agents with models, tools, and durable workflows.
  • Web applications: Choose rendering, data, authentication, and caching approaches.
  • Backends: Choose a runtime, connect services, and run background work.

Browse all stacks

Build with prompts

Build an app, extend an agent, or improve your project with ready-to-use prompts for your coding agent.

Ready to deploy?