Analytics, experimentation and AI systems for better digital decisions.
Conversion Works helps businesses improve digital performance through analytics, experimentation, AI-enabled workflows and data-quality tooling. Practical systems built by an operator — not strategy decks.
Two flagship tools, built for the AI era.
Conversion Works develops focused infrastructure for the work I do every day — data quality at one end of the stack, experimentation at the other.
Event data quality gateway for the AI era.
Conform infers your event schema from real traffic in 24 hours, blocks invalid events at the edge, and feeds developers fast, in-console feedback. Catch broken tracking before it pollutes analytics, marketing platforms, experimentation and AI.
conform.conversion-works.co.uk →Edge-first A/B testing on Cloudflare Workers.
Run JS, CSS, HTML, redirect and rewrite experiments at the edge — zero flicker, no heavy client SDK, and analytics-grade attribution. Built by experimenters for experimenters.
workers.conversion-works.co.uk →Where I do my best work.
The work spans analytics, experimentation, AI and the infrastructure that holds it all together. Engagements are usually a mix of consulting and building.
Experimentation strategy
A/B and multivariate testing programmes that produce decisions, not just dashboards. Hypothesis quality, statistical literacy, edge-first delivery.
Product analytics
Event models, funnels and retention measurement that the product team actually uses to ship better software.
Marketing analytics
Attribution, total-marketing ROAS, server-side tagging and incrementality so budgets follow real impact.
GA4 & Adobe Analytics
Implementation, audit, migration and remediation. Twenty years of getting clean numbers out of messy platforms.
AI workflow design
Codify the analytics, experimentation and ops work your team does daily into structured systems that AI agents can run.
Data quality & event governance
Schema validation, contracts and edge-native gateways so bad events never reach analytics, marketing or AI.
Featured articles.
Is AI the Harbinger of Real Personalisation?
Personalisation has been next year's big thing since about 2006. Most companies got as far as personas and stopped. AI removes the cost of building the experiences, but a behavioural segment of 20,000 visitors still needs eight months to conclude one test. The constraint moves to measurement.
Experimentation6 Ways to Improve the ROI of Your Experimentation Programme Using AI in 2026/27
Where AI changes the return on a corporate experimentation programme: three ways it cuts the cost of a decision and three ways it raises the value of one. Plus where it fails, including the review queue, the win factory and the 64% chance that a flat test 'wins' somewhere.
ExperimentationMulti-Armed Bandits and Return-User Bias: The Trade-Off Nobody Explains
Move traffic mid-test and an A/A experiment can read 15% apart. Why reassigning traffic biases a bandit when conversions take days, how GrowthBook and Optimizely handle it, and the cadence trade-off that decides whether a bandit is worth running.
Need someone who can both think and build it?
Whether it's an experimentation programme that needs technical lift, an analytics implementation that's drifting, or an AI workflow you want codified — let's talk.