Experimentation at the edge — by experimenters, for experimenters

Edge-first A/B testing on Cloudflare.

Run experiments faster, with more control and less performance overhead. Allocate variants at the edge, inject changes (JS/CSS/HTML), run redirects and rewrites, and measure results in your existing analytics.

JS CSS HTML Redirect Rewrite Mixed mode
Hosted in your tech stack Runs on your Cloudflare, Akamai or CloudFront.
No new source of truth Track in your current analytics or MI toolset.
Performance-first No clunky JavaScript, zero flicker.
Deploy in seconds No JS tagging required.

How it works

A straightforward flow your team can understand — and control.

  1. 1

    Deploy the Worker

    Install on the routes you want to experiment on — staging, production, or both. It runs on your own Cloudflare, so there's nothing new to scale and no vendor library slowing the page down.

  2. 2

    Define experiments

    Target URLs, set your splits, and define the actions: inject JS/CSS/HTML, redirect, or rewrite. Centralised config with a clear lifecycle — Build → QA → Live → Completed — that integrates with your staging sites, GitHub and AI coding agents.

  3. 3

    Track & learn

    Track exposure and conversions with out-of-the-box analytics integrations. Exposures and assignments flow into GA4 (or your stack) and you keep your existing source of truth — no complicated analytics set-ups.

Why teams choose Conversion Workers

Everything is built around speed, governance, and clean measurement — without heavyweight experimentation tooling.

⚡

Edge allocation

Assign visitors at the edge for fast, consistent variant delivery — milliseconds from the user, wherever they are.

🪶

Minimal performance overhead

Zero flicker, and no large client library loaded just to split traffic. Your Core Web Vitals stay intact.

🏠

Hosted on your infrastructure

Runs on your own Cloudflare — lower vendor lock-in and better value for money. Choose which features you want to use.

🧩

Mixed-mode variations

Inject JS + CSS + HTML together, and combine with redirects and rewrites when a test needs more than one lever.

📊

Clean measurement

Send exposures and assignments into GA4 (or your stack) and keep your existing source of truth. No complicated analytics set-ups.

🛡️

Experiment governance

Centralised config and a clear lifecycle — Build → QA → Live → Completed. Integrates with your staging sites, GitHub and AI coding agents.

What you can test

From a one-line copy tweak to a full-page redesign — usable alone or combined, all with zero flicker guaranteed.

🎨

UI changes

Restyle buttons, layouts and components via CSS + JS injection — no redeploy of your own app required.

✏️

Content swaps

Change headlines, offers and copy via HTML injection, targeted to the exact URLs you choose.

🔀

Landing page tests

Split traffic to alternate landing pages via edge redirects, with stable assignment per visitor.

📄

Full-page variants

Serve an entirely different page via rewrites while keeping the original URL in the address bar.

🧬

Mixed mode

Inject + redirect + rewrite together in a single experiment when a test needs more than one mechanism.

✨

Zero flicker

Because variants are resolved at the edge before the response reaches the browser, there's no flash of original content.

The key idea

Experimentation at the edge, measurement in your existing analytics.

No new source of truth, no heavyweight tooling, no performance tax — just faster experiments running on infrastructure you already own.

Request flow (simplified)

A predictable path from request to tracked exposure — assignment happens at the edge, measurement lands in your stack.

1
Request hits CloudflareThe visitor requests a targeted URL
→
2
Worker evaluatesURL rules + existing assignment cookie
→
3
Assign variantControl or Variation A/B by split rules
→
4
Apply actionsInject / redirect / rewrite per variant
→
5
TrackSend exposure event into GA4 (or your stack)

Feature highlights

Built for practical experimentation teams. Powerful enough for engineers, simple enough for CRO operators.

Targeting

Match by URL path rules. Keep targeting simple and predictable, so you always know which pages an experiment touches.

Persistence

Stable assignments via cookies, with sensible defaults and clear override rules — a returning visitor sees the same variant.

Variation types

JS/CSS/HTML injection, plus redirects and rewrites — usable alone or combined into a single mixed-mode experiment.

Staging → Production workflow

Run QA safely, then promote to Live when you're confident. Governance and lifecycle are built in, not bolted on.

Measurement, integrations & AI control

AI control

MCP server for AI agents

Conversion Workers ships a Model Context Protocol (MCP) server, so AI coding agents drive experimentation directly from your workflow. Describe a test in plain language and the agent builds the variation, targets the URLs, splits traffic, QAs it on staging and promotes it Build → QA → Live — spec-driven experimentation that fits AI-assisted development.

Analytics

Google Analytics integration

Exposures and assignments flow straight into GA4 as events, mapped to your existing dimensions and audiences. Analyse experiments in the reports, explorations and BigQuery export you already use — GA4 stays your single source of truth, with no parallel reporting layer to reconcile.

First-party

Onboard analytics — tagged & tagless

A first-party analytics platform runs at the edge alongside your experiments, so you can measure conversions without depending on a third-party tool. Tagged: fire a lightweight event from your existing setup. Tagless: capture conversions server-side at the edge with no client tags at all — resilient to ad-blockers, cookie loss and consent gaps. Flip the provider toggle between GA4 and first-party data whenever you want.

Want to see it running on your site?

Book a demo and we'll walk through setup, governance, and tracking — tailored to your stack.

Prefer email? hello@conversion-works.co.uk